Overviewall green2026-08-03 06:04 UTC · 135d history
Overview
Portfolio health across every product and automation channel.
Needs attention
AI Model Watch — Social distribution via this hub is TODO: no warmed Reddit/X/Bluesky accounts exist for this brand yet; must follow hub anti-detection + Reddit karma/age rules before any promo.; GSC/GA4 analytics pending (analytics_tool=none); Cloudflare Web Analytics is a privacy-friendly option.
FutPicks — GEO unblocked 2026-06-08 AND VERIFIED HOLDING 2026-06-10: all AI crawlers 200; llms.txt LIVE; /compare/rebelbetting + /compare/trademate LIVE (SSR, in both sitemaps, GSC submitted).
LicitaIA — status is 'building'
NbaPropLab — ALL-SPANISH pivot 2026-06-24 (GA4: the only engaged cohort is ES; site/engine/checkout already ES). Social flips EN->ES (X @StatLineNerd, Bluesky, YouTube, Discord). The @nbaproplab Telegram = the NBAv3 C# engine's Spanish board = the public free funnel (single emitter); the EN nba-telegram-drop is RETIRED and nba-discord-feed is PAUSED pending an ES builder. Site TODO: localize /players to ES + change the 21+ gate to 18+. ICP repointed to LatAm.
TPVTop — Google Ads pending: no activar hasta limpiar claims rojos (GTM compliance).
Vigía DPO — status is 'building'
Automation health
Routines
23
Healthy
23
Overdue
0
Stale
0
Disabled
0
Next run · Weekly audit of Sentiment Radar ingested data quality; reports precision issues, starved subjects, and improvement suggestions. in 58m
Daily NbaPropLab Discord free good-picks feed (Good-tier lead, shows misses, lint-gated). Server + #free-good-picks webhook EXIST and the sender (discord_post.py) is verified live (2026-06-23) — runs as a daily cron.At 06:35 PM, every day13h agoin 10hskip SKIP: no Good-tier picks 2026-08-02. NBA dark (offseason, 0 games); WNBA has 4 games (IND@MIN, LAS@POR, CON@DAL, TOR@GSV) with 202 picks published, but ratingDistribution is elite 0 / good 0 / marginal 15 / weak 111 / avoid 76. Top score of the day 62.4 (Charlisse Leger-Walker P+R+A u14.5, Marginal), below the Good threshold. NOT a pipeline gap: proplab_system_status healthy, morning_wnba succeeded 2026-08-02T10:28Z and recovery_wnba 09:43Z, and 202 picks are live. Backend: Good tier 3153 settled, 62.5 pct hit, +19.3 pct ROI, +609.7u, last settled 2026-08-01. No message built, no post sent. 2nd straight no-Good day.good_picks: 0games: 4picks: 202
SEO / GEO
Monthly per-product SEO audit refresh for the marketing hubAt 09:30 AM, on day 1 of the month1d agoin 29dok Audited 2 products (tpvtop, aimodelwatch); top issue: aimodelwatch sitemap has 434 URLs and zero lastmod, leaving 46% of the programmatic footprint (/compare, /migrate) invisibleproducts: 2
Weekly AI-crawler / GEO accessibility recheck for the marketing hubAt 09:19 AM, only on Monday2d agoin 1hok ai-meter bloqueado por robots gestionado de Cloudflare (GPTBot+ClaudeBot); futpicks ya abierto (200, sin disallow); resto de productos abiertosblocked: 1
Growth
Daily per-product metric snapshots for the marketing hubAt 08:01 AM, every day3m agoin 23hok Refreshed analytics for 9 products + deployedproducts: 9
Weekly cross-product growth review for the marketing hubAt 09:07 AM, only on Monday2d agoin 1hok Weekly review 2026-07-31 (corte parcial 4d): win TPVTop (GSC +4%, clics 12->14, Umami 32->59 visitors +84%) / laggard AI Model Watch (KPI 1/100 semana 5, bloqueo = captura no visibilidad)
Internal
NBAv3 daily production watchdog — product-path health + open-issue continuity (read-only; alerts on anomaly)At 04:13 PM, every day15h agoin 8hok Product path healthy (health 200, 6/6 containers healthy incl. telegram, 404/404 P(hit) on 08-02, settlement+odds fresh). TG-PUBLISH fix re-verified (no new 42703; 0 GOOD picks so no send due). Open: ADMIN-CH 8d, LOG-ROTATE 8d.gap_days: 0picks_latest: 404good_latest: 0
NBAv3 daily v5.0 shadow-rollout progress analysis (read-only; evaluates flip-readiness)At 04:45 PM, every day15h agoin 8hok Dormancy tick (Sunday, non-Monday): no restore-daily condition met; appended ledger line only. trainable=8350, validatable=77/300, AUC 0.4714, gbm_vm 48.9 vs GOOD 58.0.oos_n: 11076trainable: 8350validatable: 77
Daily forensic postmortem of every user-facing NBA+WNBA pick from the previous day — why each hit/missed (prediction vs reality) — written to pick-analysis/<date>/. Read-only on prod; Opus, high thinking.At 03:07 PM, every day16h agoin 7hok 2026-08-01 WNBA-only 4/10 (40%); loss column spined by mode #1 (peripheral Unders on alphas Wilson/Cardoso/Ionescu); migrated RECURRING.md to v2 + promoted defect cluster to backlog 999.6; POSTed 10 narratives + recap (all 200)picks: 10hits: 4hitrate_pct: 40
Daily: settle + mark + report the v1 paper experiment through the LOCAL pmpaper engine (idempotent, lock-protected; pilot_v0 kept separate).At 09:08 AM, every day22h agoin 1hok Daily ops OK: 2 settlements (v1.1 negeri-sembilan WIN +97.56; v1 moonshot-best-chinese-ai LOSS -108.67), 4 marks, 0 errors; v1.1 1/50 clusters, v1 3/50newly_settled: 2new_marks: 4errors: 0
Daily: research eligible Polymarket markets and place v1 paper decisions through the LOCAL pmpaper engine (evidence + cluster required; guards enforced in code).At 10:01 AM, every day22h agoin 1hok v1.1: 2 decisiones PLACED (JOLTS junio 7.4-7.5M NO $150 net edge 0.095; margen Clacton >=30% YES $150 net edge 0.037). CSU 9+ named storms rechazada REJECT_DEPTH_CAP (tope $18 < suelo $25) pese a ser la mejor tesis. 11 mercados investigados y descartados.placed: 2rejected: 2researched_skipped: 11
Weekly Sharpyard SEO/indexation check: confirms new blog posts get into Google, pings sitemap, auto-requests indexing for not-yet-indexed posts via GSCAt 10:04 AM, only on Tuesday2d agoin 1dok 18/18 posts del blog indexados (0 pendientes), sitemap resubmitido HTTP 204, 7d: 2 clicks / 30 impresiones (vs 0/20 semana previa)posts_total: 18indexed: 18not_indexed: 0
Weekly Sharpyard funnel review + single highest-impact next action (Reddit excluded as a channel)At 09:53 AM, only on Monday2d agoin 1hok STILL PARKED (sunset): 2a semana negativa consecutiva. GSC 275->261 imp (-5%), 11->8 clics (-27%); GA4 51->41 sesiones (-20%), sign_up=0; waitlist 0/150-250 (listas 4/5 literalmente vacias); 1 star / 30-50. Unico dato nuevo: 3 sesiones desde chatgpt.com (GEO pasivo) y vertical-slice-architecture-dotnet adelanta a claude-code-for-dotnet como top earner GSC. Decision sin cambio: cero gasto en superficies Sharpyard.gsc_impressions: 261gsc_clicks: 8ga4_sessions: 41
Weekly audit of Sentiment Radar ingested data quality; reports precision issues, starved subjects, and improvement suggestions.At 09:03 AM, only on Monday2d agoin 58mpartial Stack offline 3rd week; Docker Desktop uninstalled and replaced by native WSL Docker CE 29.6.1 - deploy-* containers and radar named volumes are gone, ingestion corpus absent on host. Audit blocked at step 1; report written with rebuild plan.outage_days: 17containers_present: 0radar_volumes_present: 0
Nightly (22:30) ContentEngine journal + watchdog: writes a complete per-run narrative of the day (published/held/blocked/failed/stuck, with links and reasons) to dashboard/engine-journal.md and flags anomalies (engine down, stale parked runs, missed cadence, dispatcher errors).At 10:33 PM, every day9h agoin 14hpartial Pulse 2026-08-02 (Sun): DEGRADED - engine healthy (api/v1/health 200, peek clean []), 2/2 due slots ran & published (nbaproplab-bluesky 12:00 -> bsky 3ms3tionlie2x; futpicks-telegram 17:50 -> t.me msg 675), 0 blocked/failed/stuck. Degrade: (1) nbaproplab-bluesky ran 62m58s vs 3-7m baseline - 2nd stall in 4 days (07-30 was 47m49s) and worsening; (2) dispatcher silent 5 days (last entry 2026-07-28T18:04). Angle->surface mismatch reopened (futpicks-telegram handed Bluesky-only channel-cross-promo, 5th time). Closed: local morning cluster recovered, journal ran 2 nights running.product: sharpyard
Fast-path ContentEngine health alarm (12:00/15:00/20:00 Madrid). Read-only check of health, stuck parked runs, Failed/Blocked runs, dispatcher errors, and missed-cadence; on anomaly pushes a Telegram alert via lib/notify.py and appends to dashboard/engine-watchdog-log.md. Silent (log-only) when healthy. Complements the nightly engine-daily-journal.At 12:09 PM, every day11h agoin 4hno result
Weekly (Mon) engagement learning loop: pulls per-post Bluesky metrics for @statlinenerd + @futprob, attributes angles via the ContentEngine ledger, updates playbooks/social-what-works.md, and tunes engine angle weights when there is real signal.At 10:44 AM, only on Monday2d agoin 2hok Followers flat 2/2 (5th review, 8wks). Raw +2 (NbaPropLab 07-28 repost), FutPicks +0. KEY: resolved all lifetime interactions via actor APIs - 100% non-human (bots @too-fee/@timxai.ai, deleted accts, @futprob self-replies, ghost reposts w/ empty getRepostedBy). 78 posts, 0 verified human engagement; prior angle/opener conclusions pruned as artefacts. Top angle: show-misses raw avg 1.00 vs pool 0.16 - REJECTED (both hits ghost reposts, 1 rotation apart). Hashtag experiment ran (6/7 tagged, first ever) -> no humans. Marquee+binary-question cross ran 07-29, failed; closed. Weight changes: none (no angle >=3 posts this 4-day window; verified dist uniformly 0). Ops: angle weights appear inert (strict 6-angle round-robin, all n=4 despite w=5-8); futpicks-bluesky 2/5 days.product: sharpyard
Daily self-improvement for AI Model Watch — runs locally on the Max subscription (no API key)At 10:04 AM, every day22h agoin 1hpartial Shipped ROADMAP top item: 3 Google rows (gemini-3.5-flash-lite, robotics-er-2 preview + streaming) unblocked both stale replacement pointers, drift detector now 0; corrected gemini-3.6-flash released date. Catalog 201->204, build green 438 pages, pushed + dataset synced. DEPLOY BLOCKED: Analytics Engine not enabled on the CF account fails the whole deployment (owner one-click); production still serves 2026-07-31/201.models: 204pages: 438drift_hits: 0
Daily FutPicks HQ Discord feed to #análisis - the EXPERIMENTAL WC model's predictions + odds + edge (matches the live site; F1-lint-gated, EXPERIMENTAL framing). Server + webhook live.At 10:01 AM, every day22h agoin 1hpartial SKIPPED: sin lecturas del modelo (picksToday=0, evaluate_match sin resultados) — no se publico nadamatches_upcoming: 24model_reads: 0product: futpicks
Daily FutPicks #general engagement poll (native Discord poll: ¿quién gana? / ¿+- 2.5?) via build_engagement_prompt. WC pick-free, F1-gated. Animador voice.At 04:04 PM, every day16h agoin 7hok Encuesta de jornada en #generalproduct: futpicks
Daily FutPicks #historial settled-results recap (hits AND misses, red accent on misses) via build_results_recap. WC pick-free, F1-gated. Animador voice.At 11:05 PM, every day8h agoin 15hok Recap de resultados en #historialproduct: futpicks
Weekly (Mon) FutPicks #informe-del-modelo report card (hit/ROI/settled + honest misses, slate accent) via build_report_card. WC pick-free, F1-gated. Animador voice.At 09:04 AM, only on Monday2d agoin 58mok Informe semanal del modelo en #informe-del-modeloproduct: futpicks
Weekly ES Bluesky teaser for DevAI Semanal (@devaisemanal.bsky.social) after the Tuesday newsletter publishesAt 12:06 PM, only on Tuesday2d agoin 1dok Teaser ES del numero MCP Apps: la UI embebida no es autorizacion, el servidor valida usuario/tenant/politica igual que ante un cliente HTTP
Bluesky posting for LicitaIA (@licitaia.org), 3x/week, promoting the free pliego analyzer + guides to drive free signups (auto-publish).At 02:04 PM, only on Monday2d agoin 5hpartial SKIP (no post): ya habia salido un post hoy 2026-07-31 a las 06:28 UTC (tool calculadora umbral SARA, rkey 3mrwd7eifvo23). La cuenta postea cada 2-3 dias (07-24, 07-27, 07-29, 07-31) y publicar de nuevo 5,6h despues romperia la cadencia. Sin prioridad de guia nueva: el ultimo commit a blog-data.ts es 07-24 (acuerdo marco), ya publicado. Proximo slot sugerido: 08-02, toca GUIA (van 2 tools seguidos 07-31 y 07-22 y la ultima guia sin destacar reciente seria una evergreen fuera de la ventana de 3 semanas).product: licitaia
Accounts
25 accounts driving promotion across every channel.
Daily NbaPropLab Discord free good-picks feed (Good-tier lead, shows misses, lint-gated). Server + #free-good-picks webhook EXIST and the sender (discord_post.py) is verified live (2026-06-23) — runs as a daily cron.37 runs
ScheduleAt 06:35 PM, every dayLast13h agoNextin 10h
13h agoskipSKIP: no Good-tier picks 2026-08-02. NBA dark (offseason, 0 games); WNBA has 4 games (IND@MIN, LAS@POR, CON@DAL, TOR@GSV) with 202 picks published, but ratingDistribution is elite 0 / good 0 / marginal 15 / weak 111 / avoid 76. Top score of the day 62.4 (Charlisse Leger-Walker P+R+A u14.5, Marginal), below the Good threshold. NOT a pipeline gap: proplab_system_status healthy, morning_wnba succeeded 2026-08-02T10:28Z and recovery_wnba 09:43Z, and 202 picks are live. Backend: Good tier 3153 settled, 62.5 pct hit, +19.3 pct ROI, +609.7u, last settled 2026-08-01. No message built, no post sent. 2nd straight no-Good day.good_picks: 0games: 4picks: 202
1d agoskipSKIP: no Good-tier picks 2026-08-01. NBA dark (offseason, 0 games); WNBA has 2 games (LVA@CHI, NYL@PHX) with 124 picks published, but ratingDistribution is elite 0 / good 0 / marginal 8 / weak 87 / avoid 29. Top score of the day 61.7 (Alyssa Thomas PTS o14.5, Marginal), below the Good threshold. NOT a pipeline gap: proplab_system_status healthy, morning_wnba succeeded 2026-08-01T10:34Z and recovery_wnba 16:14Z (21 min before run). Backend live: Good tier 3153 settled, 62.5 pct hit, +19.3 pct ROI, +609.7u, last settled 2026-07-31 (4 of 5 Good picks hit). No message built, no post sent.good_picks: 0games: 2picks: 124
2d agoskipSKIP: no Good-tier picks 2026-07-31. NBA dark (0 games, offseason); WNBA has 3 games scheduled (SEA@ATL, DAL@WAS, IND@POR) but picksCount=0 and ratingDistribution all zeros - the daily pick pipeline had not run yet at execution time (06:25Z; morning_wnba last success 2026-07-30T10:34Z, recovery_wnba 2026-07-31T02:16Z with 0 picks published). API health: healthy. Backend live: Good tier 3146 settled, 62.5 pct hit, +19.3 pct ROI, last settled 2026-07-29. No message built, no post sent.good_picks: 0games: 3picks: 0
6d agoskipSKIP: empty slate 2026-07-27 (6th dark day: 07-21, 07-23, 07-24, 07-25, 07-26, 07-27). NBA dark (offseason) and WNBA 0 games / 0 picks, WNBA break window. Verified NOT stale instrumentation: proplab_system_status healthy with morning 10:45Z, morning_wnba 10:29Z and recovery_wnba 09:30Z all succeeding TODAY, and proplab_games returns []. Backend live: Good tier 3141 settled, 62.5 pct hit, +19.3 pct ROI, last settled date 2026-07-22 (nothing has settled in 5 days). No message built, no post sent.good_picks: 0games: 0picks: 0
7d agoskipSKIP: empty slate 2026-07-26 (5th dark day: 07-21, 07-23, 07-24, 07-25, 07-26). NBA dark (offseason) and WNBA 0 games / 0 picks, WNBA All-Star break window. Verified NOT a stale pipeline: proplab_system_status shows morning_wnba succeeded 2026-07-26T10:29Z and recovery_wnba 08:14Z, and the public /api/v1/games returns []. Backend healthy: Good tier 347 settled, 62.2% hit, +18.9% ROI, last settled date 2026-07-22. No post sent.good_picks: 0games: 0picks: 0
8d agoskipSKIP: empty slate 2026-07-25 (3rd straight day). NBA dark (offseason) and WNBA 0 games / 0 picks - WNBA All-Star break window. Backend healthy: track_record live, Good tier 62.2% hit / 18.9% ROI / 347 settled through 2026-07-22. No post sent.good_picks: 0games: 0product: nbaproplab
9d agoskipNo Good-tier slate 2026-07-24: NBA dark (0 games) and WNBA empty (0 games/0 picks). SKIP per empty-slate rule, no post.good_picks: 0product: nbaproplab
10d agookSKIP: empty slate 2026-07-23 (NBA and WNBA dashboards both 0 games / 0 picks; WNBA All-Star break window). No post sent, expected summer behavior per SKILL.md.good_picks: 0skipped: 1product: nbaproplab
11d agookgood_picks: 1
12d agoskippedSKIP: empty slate entirely on 2026-07-21 - proplab_dashboard returns gamesCount:0 / picksCount:0 for BOTH nba and wnba (and unfiltered), proplab_games returns []. Not a zero-Good-tier day: there is no slate at all (NBA offseason + WNBA no games scheduled/published, likely All-Star break). Nothing posted. Model health unchanged: Good tier 3137 picks 62.5% hit / +19.4% ROI through 07-20.good_picks: 0games: 0picks: 0
13d agookgood_picks: 1
14d agoskipNo Good-tier picks on 2026-07-19 WNBA slate (3 games; dist good:0 elite:0 marginal:12 weak:128); no NBA (offseason). Skipped per Good-lead contract.good_picks: 0games: 3product: nbaproplab
2 routines · 12 logged executions in this channel.
Monthly per-product SEO audit refresh for the marketing hub2 runs
ScheduleAt 09:30 AM, on day 1 of the monthLast1d agoNextin 29d
2d agookAudited 2 products (tpvtop, aimodelwatch); top issue: aimodelwatch sitemap has 434 URLs and zero lastmod, leaving 46% of the programmatic footprint (/compare, /migrate) invisibleproducts: 2
32d agookAudited 2 products (futpicks, nbaproplab); both June blockers fixed, top issue: nbaproplab 1,521 programmatic pages earn ~0 impressions -> rank-or-pruneproducts: 2
Weekly AI-crawler / GEO accessibility recheck for the marketing hub10 runs
ScheduleAt 09:19 AM, only on MondayLast2d agoNextin 1h
2d agookai-meter bloqueado por robots gestionado de Cloudflare (GPTBot+ClaudeBot); futpicks ya abierto (200, sin disallow); resto de productos abiertosblocked: 1
6d agook9 productos probados con UA GPTBot/ClaudeBot + control navegador: 8 abiertos; ai-meter sigue bloqueado por robots.txt gestionado de Cloudflare (edge devuelve 200); futpicks abierto (200 ambos bots), sin cambios de estado desde 2026-07-20blocked: 1
13d agookai-meter blocked (robots.txt Disallow GPTBot+ClaudeBot); other 8 open incl futpicks (200, its #1 unblock already done)blocked: 1
20d agookAll 8 products open to AI crawlers (GPTBot+ClaudeBot 200); futpicks confirmed unblocked, not 403; no state changesblocked: 0
27d agookAll 8 products open to AI crawlers (GPTBot+ClaudeBot 200); futpicks confirmed unblocked; no state changes since 2026-06-29blocked: 0
34d agookAll 8 product sites open to GPTBot+ClaudeBot (200, robots clean); futpicks NOT blocked at edge — body-verified real homepage to GPTBot, contradicting standing task premiseblocked: 0
41d agookAll 6 products open to GPTBot+ClaudeBot (200); futpicks no longer Cloudflare-blocked, #1 action done in prior run; vigia-dpo baselinedblocked: 0
48d agookfutpicks UNBLOCKED (GPTBot/ClaudeBot now 200, was 403 at CF edge); all 5 products open to AI crawlersblocked: 0
55d agookfutpicks blocked at Cloudflare edge (GPTBot/ClaudeBot 403); tpvtop, nbaproplab, sharpyard, devai-semanal all open (200)blocked: 1
58d agookfutpicks AI crawlers blocked: GPTBot, ClaudeBot; other products openfutpicks_blocked: GPTBot,ClaudeBot
Growth
2 routines · 67 logged executions in this channel.
Daily per-product metric snapshots for the marketing hub58 runs
ScheduleAt 08:01 AM, every dayLast3m agoNextin 23h
1d agookRefreshed analytics for 9 products + deployedproducts: 9
2d agookRefreshed analytics for 9 products + deployedproducts: 9
4d agookRefreshed analytics for 8 products + deployedproducts: 8
5d agookRefreshed analytics for 9 products + deployedproducts: 9
6d agookRefreshed analytics for 9 products + deployedproducts: 9
7d agookRefreshed analytics for 9 products + deployedproducts: 9
8d agookRefreshed analytics for 9 products + deployedproducts: 9
9d agookRefreshed analytics for 9 products + deployedproducts: 9
10d agookRefreshed analytics for 9 products + deployedproducts: 9
11d agookRefreshed analytics for 9 products + deployedproducts: 9
12d agookRefreshed analytics for 9 products + deployedproducts: 9
13d agookRefreshed analytics for 9 products + deployedproducts: 9
+ 46 earlier runs
Weekly cross-product growth review for the marketing hub9 runs
ScheduleAt 09:07 AM, only on MondayLast2d agoNextin 1h
ScheduleAt 04:45 PM, every dayLast15h agoNextin 8h
15h agookDormancy tick (Sunday, non-Monday): no restore-daily condition met; appended ledger line only. trainable=8350, validatable=77/300, AUC 0.4714, gbm_vm 48.9 vs GOOD 58.0.oos_n: 11076trainable: 8350validatable: 77
1d agookPARKED-DATA-GATED: logistic AUC 0.473 (sub-random, GOOD bin worst of four), NEW Premium 45.8% vs CURRENT GOOD 58.3% (-12.5pp); GBM vm 48.2% vs 58.0%; trainable 8218/3000 MET, validatable 77/300 short; pipeline healthyoos_settled: 10944auc: 0.4729trainable: 8218
3d agookOOS unfroze post-break (10520, +435); logistic AUC new low 0.4691, NEW Premium 44.9% vs GOOD 57.8% (-12.9pp); trainable 7794 MET, validatable 70/300 short; 21 PARKED - DATA-GATEDoos_settled: 10520auc: 0.4691new_premium_hit: 44.9
5d agookPARKED-DATA-GATED: OOS 10085 flat 6th run (All-Star break tail; 5 games tonight, unfreezes 07-29). Logistic AUC 0.4766, GOOD bin 0.475 worst of four. NEW Premium 45.5% vs CUR GOOD 57.8% (-12.3pp). trainable=7359/3000 MET, validatable=65/300 short. Pipeline healthy: 508 picks scored 100% P(hit), 0 GOOD.oos_settled: 10085auc: 0.4766trainable: 7359
Daily forensic postmortem of every user-facing NBA+WNBA pick from the previous day — why each hit/missed (prediction vs reality) — written to pick-analysis/<date>/. Read-only on prod; Opus, high thinking.6 runs
ScheduleAt 03:07 PM, every dayLast16h agoNextin 7h
16h agook2026-08-01 WNBA-only 4/10 (40%); loss column spined by mode #1 (peripheral Unders on alphas Wilson/Cardoso/Ionescu); migrated RECURRING.md to v2 + promoted defect cluster to backlog 999.6; POSTed 10 narratives + recap (all 200)picks: 10hits: 4hitrate_pct: 40
1d agook2026-07-31 postmortem: 10 WNBA picks, 7/10=70%. All 3 misses were game-script/total blindness (mode #3): low game 156 sank 2 scoring Overs, +22.5 shootout 210 sank an all-around Under. Buried contrary matchups (#5) were 4/4 right. 10 per-pick files + summary + RECURRING updated; 11/11 public narratives POSTed 200 OK.picks: 10hits: 7hitrate_pct: 70
2d agookDaily ops OK (3rd run of the day, duplicate_run=true): 0 new settlements, 0 new marks, 0 errors; v1 now 2/50 resolved clusters realized -374.61, v1.1 0/50 unrealized +15.46; regenerated stale v1 scorecardnewly_settled: 0new_marks: 0errors: 0
3d agopartialDaily pipeline corrio sin integrity error (v1.1 scorecard regenerado, 0/50 clusters) pero 0 settles y 0 marks: egress a Polymarket caido (HTTPS directo reseteado + SSH Hetzner decomisionado, clave inexistente)newly_settled: 0new_marks: 0errors: 6
Daily: research eligible Polymarket markets and place v1 paper decisions through the LOCAL pmpaper engine (evidence + cluster required; guards enforced in code).5 runs
ScheduleAt 10:01 AM, every dayLast22h agoNextin 1h
21h agookv1.1: 2 decisiones PLACED (JOLTS junio 7.4-7.5M NO $150 net edge 0.095; margen Clacton >=30% YES $150 net edge 0.037). CSU 9+ named storms rechazada REJECT_DEPTH_CAP (tope $18 < suelo $25) pese a ser la mejor tesis. 11 mercados investigados y descartados.placed: 2rejected: 2researched_skipped: 11
2d agookv1.1: 80 candidates, 11 researched, 2 decisions PLACED (BN Negeri Sembilan YES $250 net edge 0.105; Odyssey 3rd-weekend 52-57m NO $35 net edge 0.255), 1 REJECT_DEPTH_CAP resized and placed, 9 skipped with reasonscandidates: 80researched: 11placed: 2
3d agookFixed broken transport (stale ssh key path silently returned an empty universe); universe went 0 -> 80 candidates. Placed 2 v1.1 decisions: NO Spider-Man opening weekend >=350m (net edge 0.305, \) and NO Nordone wins SC first round (net edge 0.094, \). 7 clusters researched and skipped, incl. Sanders-endorsee bucket rejected on a computed coherence check.candidates: 80placed: 2researched_skips: 7
6d agook40 candidates, 1 researchable; NO on 2431240 (Alibaba best Chinese AI model) rejected REJECT_DEPTH_CAP - depth cap 37.89 now below the 100 minimum stake, market structurally uninvestablecandidates: 40researched: 1placed: 0
Weekly Sharpyard SEO/indexation check: confirms new blog posts get into Google, pings sitemap, auto-requests indexing for not-yet-indexed posts via GSC2 runs
ScheduleAt 10:04 AM, only on TuesdayLast2d agoNextin 1d
5d agook16/18 posts indexados; sitemap resubmitido (204); solicitada indexacion en GSC para los 2 EN 'Crawled - not indexed' (subagents-skills, dotnet-10-feature); 7d: 1 clic / 21 impresiones (vs 2/34)posts_total: 18indexed: 16not_indexed: 2
Weekly Sharpyard funnel review + single highest-impact next action (Reddit excluded as a channel)2 runs
ScheduleAt 09:53 AM, only on MondayLast2d agoNextin 1h
2d agookSTILL PARKED (sunset): 2a semana negativa consecutiva. GSC 275->261 imp (-5%), 11->8 clics (-27%); GA4 51->41 sesiones (-20%), sign_up=0; waitlist 0/150-250 (listas 4/5 literalmente vacias); 1 star / 30-50. Unico dato nuevo: 3 sesiones desde chatgpt.com (GEO pasivo) y vertical-slice-architecture-dotnet adelanta a claude-code-for-dotnet como top earner GSC. Decision sin cambio: cero gasto en superficies Sharpyard.gsc_impressions: 261gsc_clicks: 8ga4_sessions: 41
6d agookSharpyard PARKED: waitlist 0/150-250 y 1/30-50 stars sin cambio; segunda semana de decaimiento (GSC 305->275 imp, 15->11 clics; GA4 77->51; clones 14d 2). Accion unica: disparar los 8 crossposts Dev.to/Hashnode ya escritos (sin publicar desde 06-25/07-23) como ultimo test de coste cero.waitlist: 0stars: 1gsc_impressions: 275
Weekly audit of Sentiment Radar ingested data quality; reports precision issues, starved subjects, and improvement suggestions.2 runs
ScheduleAt 09:03 AM, only on MondayLast2d agoNextin 58m
2d agopartialStack offline 3rd week; Docker Desktop uninstalled and replaced by native WSL Docker CE 29.6.1 - deploy-* containers and radar named volumes are gone, ingestion corpus absent on host. Audit blocked at step 1; report written with rebuild plan.outage_days: 17containers_present: 0radar_volumes_present: 0
6d agopartialStack offline 13 dias (desde ~2026-07-14), segunda semana consecutiva bloqueada; causa raiz confirmada: docker-compose sin restart policy. Datos intactos.days_down: 13containers_stopped: 12volumes_intact: 3
Nightly (22:30) ContentEngine journal + watchdog: writes a complete per-run narrative of the day (published/held/blocked/failed/stuck, with links and reasons) to dashboard/engine-journal.md and flags anomalies (engine down, stale parked runs, missed cadence, dispatcher errors).27 runs
ScheduleAt 10:33 PM, every dayLast9h agoNextin 14h
9h agopartialPulse 2026-08-02 (Sun): DEGRADED - engine healthy (api/v1/health 200, peek clean []), 2/2 due slots ran & published (nbaproplab-bluesky 12:00 -> bsky 3ms3tionlie2x; futpicks-telegram 17:50 -> t.me msg 675), 0 blocked/failed/stuck. Degrade: (1) nbaproplab-bluesky ran 62m58s vs 3-7m baseline - 2nd stall in 4 days (07-30 was 47m49s) and worsening; (2) dispatcher silent 5 days (last entry 2026-07-28T18:04). Angle->surface mismatch reopened (futpicks-telegram handed Bluesky-only channel-cross-promo, 5th time). Closed: local morning cluster recovered, journal ran 2 nights running.product: sharpyard
1d agopartialPulse 2026-08-01 (Sat): DEGRADED - engine healthy (api/v1/health 200, peek clean), 3/3 due slots ran & published (nbaproplab-video DryRun 10:30, nbaproplab-bluesky 12:00, futpicks-telegram 17:50), 0 blocked/failed/skipped/stuck; BUT monitoring is down: this journal missed the night of 07-31 (4 of last 5 nights), dispatcher silent 4 days since 07-28T18:04, and the 06:20-06:50 local daily cluster did not run today (6 runs.jsonl entries vs 26 on 07-31). 07-31 (Fri) backfilled tonight: 4/4 ran, 3 published + 1 legit Declined (futpicks-video, WC angle post-WC).product: sharpyard
2d agopartialPulse 2026-07-30 (Thu): DEGRADED - engine healthy (api/v1/health 200, peek clean), 2/2 due slots ran & published, 0 blocked/failed/skipped/stuck; BUT the monitoring harness was down: this nightly journal missed 3 nights (07-28/29/30, backfilled on 07-31), dispatcher has logged nothing since 07-28 18:04, and local runs.jsonl collapsed 31->21->3->5/day. Root cause: user profile path moved C:\Users\ceja_ -> C:\Users\user; engine_routines.json still hardcodes the old path; C:\Users\ceja_ junction recreated 07-31 00:15. Publishing unaffected (engine runs server-side): 7/7 due slots published 07-28..07-30. Watch: nbaproplab-bluesky ran 47m49s on 07-30 vs 4m norm; futpicks-telegram angle-surface mismatch reopened 07-29.product: sharpyard
6d agookPulse: nominal - engine healthy, 3/3 due slots ran, 3 published, 0 blocked/failed/stuck, dispatcher 3 launches all rc=0. futpicks-video was NOT due (engine cron is Fri-only; task-file cadence sheet is stale).product: sharpyard
8d agookPulse: nominal - engine healthy, 3/3 due slots ran, 3 published, 0 blocked, 0 failed, 0 SkippedNotFresh, 0 stuck; dispatcher 3 launches all rc=0, zero rc=129.product: sharpyard
9d agookPulse nominal: engine healthy, 4/4 due slots ran, 3 published + 1 legit decline (futpicks-video no data), 0 stuck, dispatcher 4x rc=0 zero rc=129 zero churn. Tripwire Friday cleared BOTH open harness bugs: rc=129 launch-kill absent under the midday cluster, and the WC-drift video terminated as Declined on one launch (no relaunch livelock).product: sharpyard
10d agookPulse: nominal — engine healthy · 2/2 due slots ran · 2 published · 0 failed/blocked · 0 stuck. nbaproplab-bluesky (wnba-bridge) + futpicks-telegram (mito-vs-dato) both PUBLISHED. Dispatcher 2 launches both rc=0, zero rc=129, zero relaunch churn. Fri 07-24 is the rc=129 midday-cluster + first Fri-only video test.product: sharpyard
11d agookPulse: nominal - engine healthy (api/v1/health 200) - 3/3 due slots ran, 3 published, 0 failed/blocked/stuck - dispatcher 3/3 rc=0, zero rc=129, relaunch loop dead (0 video launches vs ~35/day) - futpicks-video now Fri-only per engine config (WC->club fix landed); futpicks-bluesky drew club angle and published: fix confirmed. Watch Fri 07-24: first video run on new cadence + midday collision window.product: sharpyard
12d agopartialPulse: degraded - engine reachable, 2/2 due slots ran & published, 0 stuck at close, but ~19h dispatcher blackout (07-20 18:18 to 07-21 13:03), 71 of 72 dispatcher launches wasted on the two 07-20 WC-drift runs until they were Declined at 22:02, and rc=129 launch-kill recurred once at 21:49 despite today's CREATE_NO_WINDOW fix.product: sharpyard
14d agookPulse: nominal - engine healthy, peek clean, 2/2 due Sunday slots ran, 2 published (nbaproplab-bluesky + futpicks-telegram), 0 failed/blocked/skipped/stuck, dispatcher rc=0 on both. rc=129 absent (solo-launch Sunday, not a clearing signal). 9th clean SkippedNotFresh-free day.product: sharpyard
+ 15 earlier runs
Fast-path ContentEngine health alarm (12:00/15:00/20:00 Madrid). Read-only check of health, stuck parked runs, Failed/Blocked runs, dispatcher errors, and missed-cadence; on anomaly pushes a Telegram alert via lib/notify.py and appends to dashboard/engine-watchdog-log.md. Silent (log-only) when healthy. Complements the nightly engine-daily-journal.
ScheduleAt 12:09 PM, every dayLast11h agoNextin 4h
no executions logged yet
Weekly (Mon) engagement learning loop: pulls per-post Bluesky metrics for @statlinenerd + @futprob, attributes angles via the ContentEngine ledger, updates playbooks/social-what-works.md, and tunes engine angle weights when there is real signal.5 runs
ScheduleAt 10:44 AM, only on MondayLast2d agoNextin 2h
2d agookFollowers flat 2/2 (5th review, 8wks). Raw +2 (NbaPropLab 07-28 repost), FutPicks +0. KEY: resolved all lifetime interactions via actor APIs - 100% non-human (bots @too-fee/@timxai.ai, deleted accts, @futprob self-replies, ghost reposts w/ empty getRepostedBy). 78 posts, 0 verified human engagement; prior angle/opener conclusions pruned as artefacts. Top angle: show-misses raw avg 1.00 vs pool 0.16 - REJECTED (both hits ghost reposts, 1 rotation apart). Hashtag experiment ran (6/7 tagged, first ever) -> no humans. Marquee+binary-question cross ran 07-29, failed; closed. Weight changes: none (no angle >=3 posts this 4-day window; verified dist uniformly 0). Ops: angle weights appear inert (strict 6-angle round-robin, all n=4 despite w=5-8); futpicks-bluesky 2/5 days.product: sharpyard
6d agookFollowers flat 4th review (statlinenerd 2, futprob 2, both +0). 14 posts this window, TOTAL ENGAGEMENT 0 - regression vs 2 non-zero posts last week. Top angle: none (all 8 attributed angles avg 0.0). Question-hook experiment did not replicate (3 more, all 0). Found 0 hashtags in 33 lifetime posts - new experiment is hashtag A/B. Ops: futpicks-bluesky published 2/7 days vs nbaproplab 7/7. Weight changes: none (no signal).product: sharpyard
13d agookFollowers flat 2/2 (3rd week); engagement +2 each (NbaPropLab 3->5, FutPicks 5->7). First-ever repost: nba 07-16 question-hook (score 2); fut 07-19 final preview 1 reply. Top opener: question-hook (0.5 avg, 1 repost of 4). No weight changes (near-zero baseline, no angle >=3 posts).product: sharpyard
20d agookFollowers flat 2/2 both accts (no growth WoW, 2nd wk); engagement ~0, all 13 attributed posts scored 0 except FutPicks group-projection (2); top angle group-projection (n=1 noise); contrast-opener hypothesis falsified; weight changes: none (no signal)product: sharpyard
27d agookFollowers flat 2/2 (first recorded baseline); top opener=market-vs-model contrast (FutPicks group-projection score 2, build-in-public score 3); NbaPropLab stat-slam openers flat 0; weight changes: none (no signal, all angles n=1)product: sharpyard
Daily self-improvement for AI Model Watch — runs locally on the Max subscription (no API key)5 runs
ScheduleAt 10:04 AM, every dayLast22h agoNextin 1h
21h agopartialShipped ROADMAP top item: 3 Google rows (gemini-3.5-flash-lite, robotics-er-2 preview + streaming) unblocked both stale replacement pointers, drift detector now 0; corrected gemini-3.6-flash released date. Catalog 201->204, build green 438 pages, pushed + dataset synced. DEPLOY BLOCKED: Analytics Engine not enabled on the CF account fails the whole deployment (owner one-click); production still serves 2026-07-31/201.models: 204pages: 438drift_hits: 0
2d agookRun #2 (same day): shipped scripts/check-lifecycle-drift.mjs — re-reads Google+Anthropic deprecation tables and diffs vs catalog. First run found 4 real Google gaps incl. a catalog note asserting Google publishes no retirement date for gemini-embedding-001 (Google states 2028-05-14). deprecations feed 77->79 rows; no changelog/alert (our omissions, not market events). Anthropic 0 drift.drift_hits_found: 4deprecation_rows: 79catalog_models: 201
2d agookDocumented the /api freshness contract (cache window, ETag/304, gzip) and caught Google re-pointing the Gemini 2.5/2.0 Flash migration path at 3.6 Flash — 2 stale rows fixed, gemini-3.6-flash added (200->201), first changelog event in 22 daysmodels: 201subscribers: 2pages: 435
5d agookAdded Amazon Nova 2 Omni (Preview) from the AWS Price List API — catalog 199->200, first row that both chats and generates images; fixed a stale feed freshness date; LEARNINGS #56/#57/#58models: 200pages: 434subscribers: 2
6d agookAmazon price-source debt 4 to 0 via the AWS Price List API: 3 rows confirmed to the cent, 3 new first-party values (nova-2-lite cache-read, nova-2-pro 1.25/10), 2 aggregator claims disproved. Build green 433 pages, deployed, live feed verified.flagged_rows_before: 24flagged_rows_after: 20models: 199
Daily FutPicks HQ Discord feed to #análisis - the EXPERIMENTAL WC model's predictions + odds + edge (matches the live site; F1-lint-gated, EXPERIMENTAL framing). Server + webhook live.20 runs
ScheduleAt 10:01 AM, every dayLast22h agoNextin 1h
22h agopartialSKIPPED: sin lecturas del modelo (picksToday=0, evaluate_match sin resultados) — no se publico nadamatches_upcoming: 24model_reads: 0product: futpicks
2d agopartialSKIPPED: 18 upcoming matches (Eliteserien/Allsvenskan/Superliga) but scoring engine returned 0 model reads (picksToday=0, evaluate_match results empty). Nothing posted.upcoming_matches: 18model_reads: 0posted: 0
2d agopartialSKIP: sin lecturas del modelo (0 odds snapshots hoy, picks board vacio); no se publico nada en #analisismatches_upcoming: 17model_reads: 0odds_snapshots_today: 0
5d agopartialSKIPPED: no model reads (day 7). Board empty for 2026-07-28 at 08:0xZ (all ratings + preview), flab_picks_today=[], flab_evaluate_match returns results:[] on UCL fixtures 2699/2507. NEW DIAGNOSIS: not just the T-3h window - the 07-26 board shows 8 pick records (AM+PM passes, settled with real results) where EVERY model field is zero (score/odds/edge/modelProbability/fairOdds/dataQuality/all 7 blocks, selection empty, blockDetails []). Engine runs and settles but emits null scoring, so rating always lands Marginal/Weak and Good can never repopulate. Track record unchanged: last settled Good pick 2026-05-17, 311 picks, 35.7 pct hit, ROI +4.35 pct all-time.matches_upcoming_3d: 23picks_today: 0board_today: 0
6d agopartialSKIPPED: no model reads at run time (08:03Z). Public Good board flab_picks_today=[] and matches_today picks=[] on all 3 fixtures, picksToday=0. NOT evidence the engine is dark: picks are scored in a T-3h preview window (today's kickoffs 17:00Z => ~14:00Z), and on 07-26 picksToday went 0 at 08:02Z -> 12 at ~16:01Z. This 08:0xZ slot is structurally upstream of scoring, so it will skip every day regardless of model health. Second gate: on 07-26 all 12 scored picks were internal Marginal/Weak, so the public Good board stayed empty anyway. Odds ingestion healthy (2349 snapshots, 7-8 books incl. Pinnacle). Nothing posted.picks_today: 0odds_snapshots: 2349upcoming_matches: 42
7d agopartialSKIPPED: no model reads. Club leagues back + 2429 odds snapshots today, but scoring engine returns empty (picksToday=0, picks board [], evaluate_match results=[] on 3 leagues). Last settled pick 2026-05-17. Nothing posted.picks_today: 0odds_snapshots: 2429upcoming_matches: 44
8d agookSKIP: model dark (picks board empty 07-25/07-26, flab_evaluate_match returns no markets on 6 fixtures across Allsvenskan/Eliteserien/Superliga/UCL-Q2); club season restarted + odds flowing, scoring engine still not emittingmatches_upcoming: 44model_reads: 0last_settled_pick: 2026-05-17
9d agookSKIP: model dark (empty picks board + evaluate returns no markets); no model reads to postproduct: futpicks
10d agookSKIP: no model reads - picks board empty for 2026-07-23; all 9 EL 2nd-qualifying matches returned empty evaluate results (no odds ingested); nothing postedproduct: futpicks
11d agookSKIP: no model reads - picks board empty (today+tomorrow, all ratings incl. preview), picksToday=0. WC over 07-19; football off-season (last settled Good pick 05-17). Only CL/EL qualifiers upcoming, unrated. Did NOT post per guardrail.reads: 0matches_upcoming: 12product: futpicks
14d agoskippedSkipped: no derivable model reads. WC Final (Spain v Argentina) is displayOnly; picks board/today empty; CL/EL Jul21-23 qualifiers return empty evaluate results. No post per SKIP rule.product: futpicks
15d agoskipSkipped: no model reads available (picks board empty; WC 3rd-place final + final are displayOnly with no model output; CL/EL qualifiers 3+ days out, results:[]). Skip clause honored, no post.product: futpicks
ScheduleAt 04:04 PM, every dayLast16h agoNextin 7h
15h agookEncuesta de jornada en #generalproduct: futpicks
1d agookEncuesta de jornada en #generalproduct: futpicks
2d agookEncuesta de jornada en #generalproduct: futpicks
5d agookEncuesta de jornada en #generalproduct: futpicks
6d agookEncuesta de jornada en #generalproduct: futpicks
7d agookEncuesta de jornada en #generalproduct: futpicks
8d agookEncuesta de jornada en #generalproduct: futpicks
9d agookEncuesta de jornada en #generalproduct: futpicks
10d agookEncuesta de jornada en #generalproduct: futpicks
11d agookEncuesta de jornada en #generalproduct: futpicks
12d agookEncuesta de jornada en #generalproduct: futpicks
13d agookEncuesta de jornada en #generalproduct: futpicks
+ 20 earlier runs
Daily FutPicks #historial settled-results recap (hits AND misses, red accent on misses) via build_results_recap. WC pick-free, F1-gated. Animador voice.23 runs
ScheduleAt 11:05 PM, every dayLast8h agoNextin 15h
8h agookRecap de resultados en #historialproduct: futpicks
1d agopartialSin resultados liquidados desde la ultima ejecucion (2026-07-31) - no se publica recapsettled_since_last_run: 0product: futpicks
2d agookRecap de resultados en #historialproduct: futpicks
6d agookRecap de resultados en #historialproduct: futpicks
7d agookRecap de resultados en #historialproduct: futpicks
8d agopartialSin resultados liquidados desde la ultima ejecucion (2026-07-24) - no se publica recapsettled_since_last_run: 0product: futpicks
9d agookRecap de resultados en #historialproduct: futpicks
10d agookRecap de resultados en #historialproduct: futpicks
11d agookRecap de resultados en #historialproduct: futpicks
13d agookRecap de resultados en #historialproduct: futpicks
17d agookRecap de resultados en #historialproduct: futpicks
21d agookRecap de resultados en #historialproduct: futpicks
ScheduleAt 09:04 AM, only on MondayLast2d agoNextin 58m
2d agookInforme semanal del modelo en #informe-del-modeloproduct: futpicks
4d agookInforme semanal del modelo en #informe-del-modeloproduct: futpicks
6d agookInforme semanal del modelo en #informe-del-modeloproduct: futpicks
13d agookInforme semanal del modelo en #informe-del-modeloproduct: futpicks
20d agookInforme semanal del modelo en #informe-del-modeloproduct: futpicks
27d agookInforme semanal del modelo en #informe-del-modeloproduct: futpicks
Weekly ES Bluesky teaser for DevAI Semanal (@devaisemanal.bsky.social) after the Tuesday newsletter publishes8 runs
ScheduleAt 12:06 PM, only on TuesdayLast2d agoNextin 1d
2d agookTeaser ES del numero MCP Apps: la UI embebida no es autorizacion, el servidor valida usuario/tenant/politica igual que ante un cliente HTTPpost ↗
5d agookTeaser ES del numero 2026-07-28: guia de contexto de Anthropic para Claude 5, patrones de Claude 3 ahora contraproducentespost ↗
12d agookTeaser ES del numero 2026-07-21: 0day de Cursor con full disclosure tras negarse el vendor a parchearpost ↗
19d agookTeaser issue 2026-07-14: overhead de tokens de Claude Code (33k vs 7k OpenCode = 4.7x coste)post ↗
26d agookTeaser 2026-07-07: prompt engineering no murio (VS Code+OpenAI bajan tool calls/tokens en GPT-5.5)post ↗
40d agookTeaser issue Copilot Spaces: mas contexto no mejora al agente, lo satura; Spaces como capa de mision no vertedero del repopost ↗
47d agookTeaser del articulo A2A Protocol (A2A vs MCP: capas distintas para agentes)post ↗
Bluesky posting for LicitaIA (@licitaia.org), 3x/week, promoting the free pliego analyzer + guides to drive free signups (auto-publish).14 runs
ScheduleAt 02:04 PM, only on MondayLast2d agoNextin 5h
2d agopartialSKIP (no post): ya habia salido un post hoy 2026-07-31 a las 06:28 UTC (tool calculadora umbral SARA, rkey 3mrwd7eifvo23). La cuenta postea cada 2-3 dias (07-24, 07-27, 07-29, 07-31) y publicar de nuevo 5,6h despues romperia la cadencia. Sin prioridad de guia nueva: el ultimo commit a blog-data.ts es 07-24 (acuerdo marco), ya publicado. Proximo slot sugerido: 08-02, toca GUIA (van 2 tools seguidos 07-31 y 07-22 y la ultima guia sin destacar reciente seria una evergreen fuera de la ventana de 3 semanas).product: licitaiapost-de-hoy ↗
2d agookTOOL rotation: calculadora umbral SARA (/calculadora-umbral-sara) tras 3 guias seguidas (07-24, 07-27, 07-29). Angulo nuevo vs el 07-10 (que fue valor estimado vs presupuesto base): que GANA el licitador si es SARA - plazos ampliados + recurso especial gratuito art. 44 LCSP. Nota: atproto faltaba en el entorno, reinstalado (requirements.txt:14).post ↗
6d agookGuia revision de precios (merged 07-18, nunca destacada: los slots 07-20/07-22 fueron tools y el 07-24 fue la guia mas nueva de acuerdo marco). Hook art.103 LCSP: no es subida automatica por inflacion; salvo energia hace falta 20% ejecutado + 1 anio desde formalizacion.post ↗
9d agookPosted NEW guide (merged today): acuerdo marco / contratos basados, arts. 219-221 LCSP, 4yr max duration, GO/NO-GO for pymespost ↗
11d agooksimulador baja temeraria tool: umbral no es % fijo, depende del nº de licitadores (art. 85 RGLCAP)post ↗
20d agookSimulador baja temeraria tool + freshly-deepened guide (art.149 LCSP: justificacion antes de excluir; garantia complementaria 5% extra)post ↗
23d agookFree TOOL rotation: calculadora umbral SARA (/calculadora-umbral-sara). Hook: SARA se mide sobre valor estimado (art.101 LCSP), no el presupuesto base; umbral 2026-2027 servicios/suministros 140.000 EUR (AGE) o 216.000 EUR (resto). First tool feature since 06-29 pliego analyzer; broke 5-post guide streak.product: licitaiapost ↗
25d agookGuia subcontratacion en contratos publicos (NUEVA guia, mergeada hoy): responsabilidad no se traslada + tareas criticas (art.215 LCSP)post ↗
27d agookNew guide (merged today): modificacion-contratos-publicos — 20% limit art. 204 LCSP, review the clause before biddingpost ↗
2026-07-31 (weekly): REVISION A CIEGAS -- el colector aimeter_local (SSH a produccion) devolvio null, asi que NO hay lectura de funnel nueva y el KPI 0/5 se ARRASTRA del 07-27 sin reverificar. ACCION (1) NUEVA y prioritaria: reparar/reintentar aimeter_local para no volver a revisar sin datos; (2) SIN CAMBIO: cero distribucion = cero lecturas nuevas, meter la auto-auditoria en conversaciones de alta intencion (umbral de lectura comercial 25 audits + 10 emails, sigue MUY lejos). || 2026-07-27 (weekly): funnel PLANO al 100% -- 1 audit, 1 email verificado, 2 email-qualified demand, 0 pro-intent, 0 leads de agencia = identico a 07-20; KPI 0/5. Ops sano (cola 0, 8/dia sin usar). DIAGNOSTICO: no es el producto, es que no hubo NINGUNA distribucion nueva -> cero lecturas nuevas. ACCION sin cambio pero con compromiso explicito: meter la auto-auditoria real en conversaciones de alta intencion esta semana (>=3) y recoger utilidad del informe + consentimiento de caso-estudio; sin eso el umbral de lectura comercial (25 audits + 10 emails) no llega nunca. || 2026-07-20 (weekly): funnel real (opctl m920q) -- 1 audit completado, 1 email verificado, 2 email-qualified demand, 0 pro-intent, 0 leads de agencia, 0 consentimientos caso-estudio; KPI 0/5 senales Pro. Ops sano (statusOk, cola 0, 8/dia). 1 visitante cualificado/30d, 21 exclusiones sinteticas. Sin GA4/GSC por diseno. Aun bajo el umbral de lectura comercial (25 audits + 10 emails). ACCION sin cambio: distribuir la auto-auditoria real solo en conversaciones de alta intencion; recoger utilidad del informe + consentimiento explicito de caso-estudio. || Distribute the real self-audit sample only inside high-intent conversations; collect report usefulness and explicit case-study consent.
2026-07-31 (audit mensual SEO): PRIMERA auditoria completa. Es la mejor superficie tecnica del hub y no hay nada que arreglar ahi (robots con Allow explicito a GPTBot/ClaudeBot/PerplexityBot/Google-Extended, llms.txt excelente y fechado 2026-07-29, Dataset con dateModified en la home, Product+AggregateOffer+FAQPage+BreadcrumbList en /models/*, titulos que ya llevan precio y contexto, canonicals correctos, TTFB 0.12s, enlazado interno completo: /compare enlaza sus 127 y /deprecations sus 74 /migrate). Mes 1 del breakout: imp 46->1931 (+4098%), primeros 12 clics. ACCION #1: el sitemap tiene 434 URLs y CERO lastmod/changefreq/priority -- para un producto cuyo valor es 'refrescado a diario' eso es la herida autoinfligida grande: Google no sabe que cambio hoy, asi que prioriza por grafo de enlaces y por eso /models/* (200 enlaces desde la home) cosecha 1908 de las 1931 imp mientras /compare/* rinde 5 de 127 paginas (4%) y /migrate/* 13 de 74 (18%), juntas 42 imp = 46% del sitio invisible. NO es calidad: cuando salen rankean muy alto (/compare/command-a-vs-kimi-k2-7-code pos 2.0, /migrate/pixtral-12b pos 3.8 con clic, /migrate/gemini-2-0-flash pos 4.0). El dato ya existe en el refresh diario de ../Autopilot -> emitir lastmod real por URL en el generador de sitemap y pinchar IndexNow solo con lo cambiado. Indicador: /compare/* pasa de 5 a 40+ paginas con impresiones. HALLAZGOS 2 y 3: los 5 /guides/* (cheapest-llm-for-chatbots/rag/coding/summarization/vision) sacaron 0 impresiones en 28d pese a ser queries comerciales de alta intencion y estar destacados en llms.txt; y www.aimodelwatch.dev devuelve 200 con la home completa en vez de 301 al apex (se autocanonicaliza, pero GSC ya indexa https://www.aimodelwatch.dev/migrate/command-light y hay fila aparte de http://). OJO al leer el CTR 0.62%: buena parte de las impresiones son queries de agente IA con comillas y booleanos (ej. 'llama 4 behemoth' meta (available or released or ...) 12 imp pos 8.1) que nunca van a hacer clic -- es senal GEO buena, no fallo de snippet. Menor: /deprecations es la pagina top (88 imp, pos 11.3, 2 clics) y solo lleva Organization + un FAQPage de UNA pregunta, sin ItemList sobre la tabla de retiradas. Informe: products/aimodelwatch/FULL-AUDIT-REPORT.md. Confirma desde el lado SEO la tesis del weekly: ya no hay problema de visibilidad, solo de captura (1/100 subs, semana 5). || 2026-07-31 (weekly): SE CONFIRMA la tesis del 07-27 -- la palanca es CAPTURA, no visibilidad. GSC lateral tras el aplanamiento (imp 1961->1931 -2%, clics 12 planos, pos 20.9 clavada) mientras Umami SIGUE subiendo 238->262 visitors (+10%, 521 pageviews/28d) y el KPI sigue 1/100 en SEMANA 5. /deprecations se consolida como pagina top (2 clics/88 imp, pos 11.3) por delante de /compare y /migrate. ACCION: (1) reescribir el CTA de alertas EN /deprecations, que es la pagina que ya recibe la demanda, no en la home; (2) NUEVO: cablear el colector listmonk en product.yaml -- hoy el KPI se arrastra a mano y el producto se revisa a ciegas. || 2026-07-27 (weekly): CAMBIA LA PALANCA. GSC se APLANA tras 3 semanas de breakout (imp 1942->1961 +1%, clics 11->12) mientras Umami SIGUE subiendo (183->238 visitors, +30%) y el KPI sigue 1/100 (semana 4). Con trafico creciendo y captura a cero, el bloqueo YA NO es visibilidad sino conversion. Ademas cambia el mix: /deprecations es la pagina top (2 clics/87 imp, pos 11.4) por delante de /compare y /migrate. ACCION NUEVA: poner el CTA de alertas por email en /deprecations y /models/* (donde aterrizan de verdad, no en /alerts) y medir su clic; la geo-optimizacion pasa a segundo plano. || 2026-07-20 (weekly): GSC sigue subiendo -- imp 1435->1942 (+35%), clics 9->11 (+22%, pos 20.9); Umami 134->183 visitors (+37%, 28d). KPI subs 1/100 plano (Listmonk lista 7). El footprint programatico (/compare /migrate /models) cosecha semana 3. ACCION sin cambio: geo-optimizar las paginas que YA impresionan + CTA de alertas en /models/*; cerrar la decision GA4-vs-Umami (probablemente NO cablear GA4). || 2026-07-13 (weekly): GSC BREAKOUT sostenido -- imp 376->1435 (+281% en 7d), primeros clics organicos aguantan 0->9 (pos 19.2), sobre las paginas programaticas /compare /migrate /models. Umami 134 visitors/173 visits (28d). KPI ground-truth 1/100. ACCION sin cambio: aplicar geo-optimization a las paginas que YA impresionan + CTA de alertas en /models/* (ahi aterrizan, no en /alerts); cerrar la decision GA4-vs-Umami (probablemente NO se cablea GA4). || 2026-07-11 (weekly): GSC TAKEOFF -- impressions 376->1244 (+231%) and FIRST organic clicks (0->7, pos 18.0), landing on the programmatic pages (/compare/*, /migrate/*, /models/*) -> the footprint is harvesting. Umami 121 visitors/156 visits (28d). KPI ground truth: Listmonk list 7 = 1 subscriber (the 3 Umami subscribe events are dupes/retries; 1/100 goal). ACTIONS: apply the geo-optimization playbook to the pages already impressing + surface an alerts CTA on /models/* pages (visitors land there, not on /alerts); decide whether GA4 is still needed (with Umami + subscribe event probably NOT -- close the gap and document it). github collector null (private repo). || 2026-07-06 (weekly): GSC now SERVING query data (376 impressions, 0 clicks, avg pos 19.1; up from 264 the day before) -- blocker #1 'verify GSC to read query data' is DONE. Top-impressing pages: /, /about, /alerts. NEXT primitives: (1) wire GA4 (still null -> no session/conversion measurement), (2) apply the geo-optimization playbook to the pages already impressing, (3) decide a warmed dev social presence (Reddit/X) per anti-detection before any promo. Growth stays SEO/GEO-led + autonomous in CI. || 2026-06-28: Launched. Live on aimodelwatch.dev (Cloudflare Pages), 181 models/12 providers, daily self-improving GitHub Actions loop (refresh count -> Claude ships next roadmap item -> build -> deploy -> IndexNow). Growth is SEO/GEO-led and fully autonomous in CI. NEXT (needs local SPAM hub or human): (1) verify GSC for the domain to read query data, (2) decide on a warmed dev social presence (Reddit/X) for value-first distribution per hub anti-detection rules, (3) apply geo-optimization playbook.
2026-07-31 (weekly): SE DESTAPA UN CIEGO -- lastIssueOpenRate 0.0 -> 0.2 (20%): el tracking de aperturas YA reporta, asi que el 0.0 de las ultimas semanas era instrumentacion y no una entrega muerta. KPI subs activos 5 PLANO tras el 4->5; sign_up 7/28d vs 5 miembros = la fuga alta->miembro se estrecha pero persiste. SEO RECUPERA el CTR que se erosionaba: imp 6252->6477 (+4%) y clics 80->82 (+2%, pos 9.74->9.60). ACCION: doblar en el cluster Claude Code (tutoriales-claude-code-aceptar 19 clics/1207 imp y claude-code-dotnet-csharp-guia 13 clics -- la cosecha de Sharpyard ya es la #2 del sitio) y atacar la fuga sign_up->miembro, que es el unico paso que separa trafico de KPI. || 2026-07-27 (weekly): EL KPI SE MUEVE -- subs activos 4->5 (+1), primer avance tras 3 semanas planas; la fuga alta->miembro se estrecha pero persiste (10 key events sign_up/28d vs 5 miembros). SEO en MESETA con CTR erosionando: imp 5998->6252 (+4%) pero clics 87->80 (-8%, pos 9.7) y GA4 293->268 (-9%). ACCION (2, en orden): (1) seguir cerrando la fuga alta->miembro-activo -- es el unico multiplicador real; (2) verificar el tracking de lastIssueOpenRate 0.0, que con 5 subs reales ya no se explica como lag de Ghost. Sigue el trasvase de Sharpyard: claude-code-dotnet-csharp-guia es ya la #2 pagina del sitio (12 clics). || 2026-07-20 (weekly): KPI subs activos PLANO en 4 (3a semana) PESE a 9 key events sign_up 7d -> la FUGA alta->miembro persiste (dupes/cancelaciones/confirm): DIAGNOSTICAR la fuga alta->miembro-activo antes que buscar mas trafico. Motor SEO sano (GSC 5493->5998 +9%, clics 75->87 +16% pos 10.0, GA4 293 plano). Superficies UTM sociales 0 sesiones (7a semana) -> RETIRAR y redoblar cluster copilot/claude-code + GEO. Vigilar lastIssueOpenRate 0.0. || 2026-07-13 (weekly): KPI subs PLANO en 4 (2a sem) PESE a 7 key events sign_up en 7d -> las altas NO cuajan en Ghost (dupes/cancelaciones/confirm): DIAGNOSTICAR la fuga alta->miembro-activo antes que buscar mas trafico. Motor SEO sano (GSC 4961->5493 +11%, clics 71->75, GA4 297). Superficies UTM sociales 0 sesiones 6a semana -> RETIRAR y redoblar cluster copilot/claude-code (40 clics/5 paginas) + GEO. Vigilar lastIssueOpenRate 0.0. || 2026-07-11 (weekly): KPI subs activos PLANO en 4 (el +1 del 07-06 no continuo; 2 key events sign_up en 7d ambos DIRECTOS pero Ghost sigue 4 -> vigilar cancelaciones/dupes). Motor SEO sano: GSC imp 4961->5516 (+11%), clics 71->73 (pos 9.77), GA4 302 (28d), Umami 234 visitors. ATRIBUCION 7d: CERO sesiones de las 6 UTM sociales (5a semana consecutiva) y CERO de los buckets UTM internos -> la conversion llega por directo/organico; considerar RETIRAR las superficies sociales UTM muertas y redoblar el cluster copilot/claude-code (5 paginas top = 39 clics GSC) + GEO (perplexity.ai 4 ses, hub.getaimeter.com 9). VIGILAR: lastIssueOpenRate 0.0 (lag de Ghost o issue sin tracking?). || 2026-07-06 (weekly): KPI (subs activos) 3 -> 4 -- PRIMER movimiento del KPI en 5+ semanas, tras el rework 06-29 de CTA+UTM internos en las 6 paginas top de GSC. GA4 28d 288->298 (+3.5%), GSC clics 66->71 (+7.6%, pos 9.58); sigue liderando el cluster github-copilot-ai-credits. El motor SEO/GEO esta sano y la conversion de alta POR FIN se mueve. FOCO: confirmar la atribucion del +1 en los buckets UTM internos (copilot_credits/claude_code_tutoriales) antes de declarar victoria; seguir cosechando el cluster copilot. || 2026-06-29: el trafico SEO RE-ACELERO tras la meseta (GA4 sessions 203->288 +42% WoW, GSC impresiones 4395->4833 +10%, clics 55->66 +20%), liderado por el cluster github-copilot-ai-credits. PERO el KPI (subs activos) sigue PLANO en 3 por 5a semana. DIAGNOSTICO RE-CONFIRMADO con run_report GA4 fresco (7d): CERO de las 6 superficies UTM (weekly_thread/weekly_teaser/latam_value/midudev_recursos/spanish_digest/es_creator_swap) genero NINGUNA sesion etiquetada esta semana -- todo es directo/organico/referral; solo 1 key event en 7d, atribuido a directo. O sea las superficies sociales NO generan clics medibles Y el embudo de alta no se puede atribuir. FOCO (instrumentacion primero): verificar/arreglar el evento sign_up en el portal Ghost via ga_admin.py (primitiva del KPI; sin el, el embudo es ciego), LUEGO reforzar CTA+UTM de enlace interno desde las paginas top de GSC (github-copilot-ai-credits-*, tutoriales-claude-code-*) al portal. GEO/LLM sigue siendo el unico canal con conversion atribuida (copilot.com ai-assistant esta semana). PENDIENTE HUMANO: login daily.dev + Flipboard (2/8 crossposts fallan), rotar claves Ghost/Mailgun. || ACTIONED 2026-06-29 (post-review): el 'instrumentacion primero' era FALSA ALARMA -- el evento sign_up esta SANO y contando (no era el bloqueo). DONE en vivo: (1) 6 paginas top de GSC reescritas con CTA+UTM por-cluster (copilot_credits/claude_code_tutoriales/cli_agentes/tokens_coste) -> las conversiones de embudo interno ya se atribuyen por tema; (2) nuevo articulo ES 'Claude Code para .NET y C#' publicado via pipeline evergreen (cosecha el trafico .NET+Claude de Sharpyard, que sigue parked). Proximo review: vigilar los nuevos buckets UTM internos.
2026-07-31 (weekly): el 90d REBOTA -14.52% -> -11.98% (101 picks, 34.7% hit, -12.1u) pero es la MISMA mecanica de ventana al reves, NO mejora del modelo: sin picks Good desde 2026-05-17 la ventana solo cubre 05-02..05-17 y se VACIA a mediados de agosto. All-time PLANO +4.35% (311 picks, +13.54u) = la cifra literal de los posts. Honest-body SIGUE ACTIVA (90d <= 0): ventana explicita en toda cifra ROI, ningun hook con ROI positivo; NO es flag de plataforma. Por mercado H2H +14.8% y H2HH1 +14.1% POSITIVOS, el dano esta en BTTS (-57%) y OU2.5 (-45%). Trafico: GSC imp 531->633 (+19%), clics 5->6, pos 48.1->46.0 desde base muy baja; GA4 82->76. ACCION: (1) angulos no-ROI y ventana explicita; (2) apoyar el hook en H2H, el unico mercado con proof positivo; (3) AVISO DE RELOJ: sin motor de scoring de vuelta con el club, a mediados de agosto el KPI 90d deja literalmente de existir -- decidir antes si el dashboard pasa a citar all-time. || 2026-07-27 (weekly): Good-picks ROI 90d -11.45% -> -14.52% (104 picks, 33.7% hit, -15.1u). Es MECANICA de ventana pura, no deriva del modelo: sin picks Good desde 2026-05-17, la ventana solo contiene 05-01..05-17 y se VACIA a mediados de agosto. All-time PLANO +4.35% (311 picks). Honest-body ACTIVA (verificado en footballlab): toda cifra nombra su ventana; ningun post lidera con ROI positivo mientras 90d<=0; NO es flag de plataforma, NO se pausa nada. Dano concentrado en OU2.5 (-48.9%) y BTTS (-62.4%); H2H sigue +14.8%. ACCION: el desbloqueo real NO es de marketing sino el reinicio del motor de scoring con el club (agosto) -- hasta entonces, Bluesky honest-proof + angulos no-ROI (metodo, H2H, lecturas de mercado) y track-record machine-readable. Trafico GA4 89->82, GSC 526->531 plano, go.bsky.app unico referral vivo. Telegram 3. || 2026-07-20 (weekly): Good-picks ROI 90d -3.14% -> -11.45% (190 picks/-21.75u) -- sigue cayendo EXACTAMENTE como predicho (rueda fuera de la semana rentable 04-07..04-11; 0 picks Good nuevos desde 05-17, hiato club hasta agosto post-WC). All-time PLANO +4.35% (311 picks) = cifra literal de Bluesky. Honest-body ACTIVA (verificado footballlab): toda cifra nombra ventana; ningun post lidera con ROI positivo mientras 90d<=0; NO es flag de plataforma. Trafico GA4 94->89, GSC imp 461->526 (+14%), go.bsky.app #1 referral. Telegram 3. Palanca sin cambio: Bluesky honest-proof + track-record machine-readable. || 2026-07-13 (weekly): Good-picks ROI 90d SIGUE negativo (-3.14%, 249 picks/-7.81u) -> reglas honest-body SE MANTIENEN (toda cifra ROI nombra ventana; ningun post lidera con ROI positivo mientras 90d<=0; citar all-time +4.35% SOLO con su etiqueta). Estructural: 0 picks Good nuevos desde 05-17 (hiato club hasta agosto) -> el 90d no recupera sin picks; NO es flag de plataforma. Trafico: go.bsky.app #1 referral (HOOK build-in-public), chatgpt.com primeras citas. StatLineNerd en HOLD hasta 07-25. Palanca sin cambio: Bluesky honest-proof + track-record machine-readable. || 2026-07-11b (user decision): NO pausar publicaciones -- flag LIMPIADO 07:57 UTC mismo dia (X FutProbLab/DevAI reanudados; StatLineNerd sigue en HOLD deliberado hasta 07-25 por su label). En su lugar: FIX DE CUERPOS en todos los generadores FutPicks (regla 1: toda cifra ROI nombra su ventana explicita; regla 2: mientras 90d <= 0, ningun post lidera con ROI positivo -- transparencia con AMBOS numeros o angulos no-ROI). Playbook actualizado (la regla de pausa queda sustituida); record-flag reservado para flags reales de plataforma. || 2026-07-11 (weekly): PAUSA-PROOF DISPARADA -- Good-picks ROI 90d +4.35 -> -0.3% (276 picks desde 04-12; regla <=0 cumplida) -> record-flag EJECUTADO 07-11 07:42 UTC (3 cuentas X pausadas 48h; es la pausa-proof pre-comprometida el 07-06, NO un flag de plataforma). Causa: TODO el profit vivo (+14.49u) vino de 04-07..04-11 (primeros 5 dias, 35 picks); ex-primera-semana el registro es plano/negativo (-0.95u en 276); sin picks nuevos hasta el club (agosto) el 90d NO recupera. All-time sigue +4.35% (311 picks, +13.54u) = lo que citan los posts de Bluesky (literalmente cierto pero apoyado en esa semana). PENDIENTE HUMANO: parar los posts proof-ROI de Bluesky (slot ~13:21 diario; ContentEngine canal fut / routine local) mientras el 90d sea <=0; corregir el KPI del routine (hoy escribio 4.35 = all-time mal etiquetado como 90d). Trafico: GA4 144->106 (-26%), GSC imp 342->443 (+30%), clics 6 pos 46; go.bsky.app #1 referral (41 ses vs google 6); chatgpt.com 3 ses (primeras citas LLM). Telegram 3 plano. || 2026-07-06 (weekly): Good-picks ROI 90d PLANO en +4.35% -> proof HOLDS, promo ON (si <=0 PAUSA total via lib/x_scheduler.py record-flag). GA4 134->144 (+7.5%), GSC imp 164->342 (+109%). go.bsky.app SIGUE #1 referral (53 sesiones > Google 6): el HOOK build-in-public de Bluesky es el canal medible. Modo Mundial pick-free (hasta 07-19). Palanca SIN CAMBIO: track-record machine-readable (Dataset schema) para convertir el desbloqueo de crawlers en citas LLM + doblar en Bluesky. || 2026-07-01 (SEO audit): the 2026-06 CRITICAL (robots.txt blocked all AI crawlers) is RESOLVED & verified holding -- catch-all Allow, llms.txt LIVE (cites +8.0% ROI / 65.9% vs Pinnacle), /compare pages SSR. Fresh GSC (28d): 4 clicks / 224 imp / avg pos 42.3 = indexed-not-ranking on the RIGHT queries (calculador de momios/parlay pos ~50, 'picks mundial hoy' pos 44). Top SEO levers now: (1) make the track-record machine-readable (Dataset/structured markup) to convert the GEO unblock into actual AI citations; (2) climb /calculadora + /world-cup off page 4 with on-page depth on the queries already impressing. CWV still poor (LCP 4.5s / CLS 0.225, ~205KiB unused JS) -- wire GOOGLE_API_KEY for field data. Full report: products/futpicks/FULL-AUDIT-REPORT.md || 2026-06-29: WC mode (06-11..07-19, pick-free per docs/wc-runbook.md) -- traffic accelerating WoW (GA4 sessions 103->134 +30%, GSC impressions 93->164 +76%, clicks 2->4). WIN: go.bsky.app is now the #1 referral source (49 sessions, ahead of Google) -> the Bluesky build-in-public HOOK is finally driving measurable traffic; double down there. WATCH: Good-picks ROI % (90d) slid 8.0->4.35 -- still POSITIVE so proof holds and promo stays ON, but the decay is structural (pick-free WC window: strong April picks age out of the rolling 90d, negative May remains). If it hits <=0, PAUSE promo across all surfaces (python lib/x_scheduler.py record-flag). Still-gating instrumentation (UNDONE): wire the GA4 purchase/pro_subscribe event so Stripe Pro conversions (primary KPI, 0 by design, keyEvents 0) become measurable. SEO still page-4 (pos 39.5), indexed-not-ranking -> push SERP CTR on /world-cup + /calculadora. Plan: docs/research/2026-06-10-futpicks-plan.md || ACTIONED 2026-06-29 (post-review): the 'wire the purchase event (UNDONE)' item was a STALE FALSE ALARM -- Stripe + the GA4 'purchase' key event are already wired & live; KPI=0 is genuinely zero paid customers, not a measurement gap. DONE: fixed the one real defect (purchase fired transaction_id only -> 0 revenue) by adding value/currency/items (resolved at runtime from /billing/config) in ../Furbov2 footballlab-web -- needs a normal CI deploy. Bluesky double-down LIVE (1 post + 6 queued daily, honest +4.35% ROI proof). Decision pending: trial-start vs first-charge revenue booking semantics.
2026-07-31 (weekly): RESPIRO tras el pico -- GSC imp 3446->3263 (-5%) pero clics 32 CLAVADOS y la pos media MEJORA 19.05->18.77. SIGUE siendo el LIDER del hub en clics organicos absolutos y la base ensanchada AGUANTA (home 14 clics/394 imp, plantilla-oferta-tecnica 3, como-leer-pliegos-pcap 2 en pos 8.0, garantias-avales 2). GA4 cookieless 22 plano, keyEvents 0. ACCION SIN CAMBIO pero MAS URGENTE: instrumentar el KPI de altas gratis activadas (#232). Con 5 semanas de SEO compounding y cero medicion del North Star, el trafico llega y no sabemos si activa -- es ya EL unico bloqueo serio del producto. || 2026-07-27 (weekly): GANADOR DEL HUB -- SEO compounding semana 5 y MEJOR semana hasta hoy: GSC imp 2813->3446 (+23%), clics 23->32 (+39%, pos 19.0). Y se ENSANCHA la base: entran al marcador plantilla-oferta-tecnica (3 clics/79 imp) y como-leer-pliegos-pcap (2 clics, pos 7.9) junto a home (14) y garantias-avales (2); primera cita LLM (chatgpt.com). ACCION: (1) doblar el patron que YA funciona -- plantillas y guias de lectura de pliegos, que es lo que acaba de empezar a convertir; (2) DESBLOQUEAR #232 (GA4 de altas): con este volumen entrando, el North Star ciego es ya el unico bloqueo serio del producto. || 2026-07-20 (weekly): SEO COMPOUNDING semana 4 -- GSC imp 2093->2813 (+34%), clics 15->23 (+53%, pos 19.5); GA4 18->24 (+33%, 100% Spain, bsky+go.bsky lideran). El fix de prerender (#259) sigue rindiendo. ACCION sin cambio: doblar clusters que YA impresionan (contratos-menores pos 6.1, garantias-avales) + profundizar baja-temeraria. PENDIENTE Fase 1 sin cambio: GA4 real de altas (#232, North Star ciego sin el), E-E-A-T (#233), Brevo (#234), LinkedIn (#237 manual). || 2026-07-13 (weekly): SEO COMPOUNDING semana 3 -- GSC imp 1275->2093 (+64%), clics 5->15 (+200%, pos 20.3); el fix de prerender (#259) sigue rindiendo. ACCION sin cambio: doblar contenido en clusters que YA impresionan (contratos-menores pos 6.1, garantias-avales) y profundizar baja-temeraria (193 imp, pos 42 = mayor upside). PENDIENTE Fase 1 sin cambio: GA4 real de altas (#232, sin el no se mide el North Star), E-E-A-T (#233), Brevo (#234), LinkedIn (#237 manual). || 2026-07-11 (weekly): SEO COMPOUNDING semana 2 -- GSC imp 1275->2011 (+58%), clics 5->14 (+180%, pos 20.3). contratos-menores-2026 pos 6.1 (229 imp) y garantias-avales (2 clics) lideran; baja-temeraria 179 imp pos 41.9 = el mayor upside on-page (profundizar ese contenido). Brand 'licitaia' 4 clics pos 1.8. GA4 cookieless 14 ses (bsky 5 + x 1 + google 3; 100% Spain). ACCION: doblar contenido en los clusters que YA impresionan. PENDIENTE Fase 1 sin cambio: GA4 real de altas (#232), E-E-A-T (#233), Brevo (#234), LinkedIn (#237 manual). Colector: gsc.py crash cp1252 en Windows ARREGLADO 07-11; github collector null (repo privado). || 2026-07-06 (weekly): BREAKTHROUGH SEO -- GSC impresiones 109 -> 1275 (~11x WoW), primeros clics organicos 0 -> 5. El fix de prerender (PR #259, HTML+JSON-LD a crawlers) no solo desbloqueo la indexacion sino que YA COMPOUNDEA: primer salto real de visibilidad. GA4 13 sesiones (bsky.app 4 + x 1 = la automatizacion social MWF empuja). El track A (contenido/SEO/GEO) esta funcionando; doblar en el contenido de los clusters que impresionan. PENDIENTE Fase 1 sin cambio: instrumentar GA4 real de altas (#232), autor E-E-A-T (#233), lifecycle Brevo (#234), LinkedIn (#237 manual) -- sin GA4 automatizado aun no se mide el North Star (altas gratis activadas). || 2026-06-26: ESTRATEGIA GTM aprobada (A contenido/SEO/GEO + B free-tools/PLG, C comunidad de contingencia). Spec: ../Spam/docs/superpowers/specs/2026-06-26-licitaia-gtm-strategy-design.md. Ejecución en issues/milestones de github.com/Khavel/licitaia (label gtm; Fases 1-4). Fase 1 en curso: ARREGLADO el bloqueador #1 -- las URL prerenderizadas (sin barra) servían el shell CSR vacío a los crawlers; ahora sirven HTML prerenderizado + JSON-LD (PR #259) -- CONFIRMADO 06-29: GSC impresiones 0->109 WoW, las paginas ya se sirven/indexan (clics 0, pos 15.2). Datos legales corregidos (umbrales SARA #257, baja temeraria art.149/85 #258). Referral surfaced (#261). PENDIENTE Fase 1: instrumentar analítica (#232), entidad de autor real E-E-A-T (#233, NECESITA identidad del fundador), lifecycle Brevo (#234), página LinkedIn (#237, MANUAL). North Star = altas gratis ACTIVADAS + retención, NO MRR. Constraint: ~2h/sem fundador, 0€, inbound.
2026-07-31 (weekly): proof ESTABLE -- Good all-time 19.3% ROI PLANO (3146 picks, 62.5% hit, +607.2u; +5 picks WNBA desde el 07-27). El slice 90d cede 17.2%->13.9%, otra vez MECANICA de ventana (entran dias flojos de julio, salen los buenos de finales de abril); el all-time, que es la cifra publicable, no se mueve. SEO: imp 128->140 (+9%) y el clic organico SE MANTIENE en 1 (3a lectura consecutiva, /nba-predictions 62 imp pos 30.5); la home ya impresiona en pos 4.7 (60 imp) sin llevarse el clic. ACCION: (1) PENDIENTE HUMANO arrastrado 2a revision -- el HOLD de X @StatLineNerd vencio el 2026-07-25 y el label SIGUE sin verificar: decidir release-hold o re-hold, no dejarlo colgado otra semana; (2) trabajar titulo/meta de la home, que rankea en pos 4.7 y no convierte. || 2026-07-27 (weekly): proof estable -- Good all-time 19.3% PLANO (3141 picks, 62.5%, +606.4u). El slice 90d salta 8.7% -> 17.2% pero es MECANICO (sale de la ventana el bloque perdedor 04-21..04-27: 172 picks Good, 44% hit); 90d solo-WNBA = 18.9% (347 picks) -> la WNBA sostiene el proof en off-season. GUARDRAIL: NO publicar ese salto como mejora del modelo -- es el mismo efecto de ventana que hunde a FutPicks, en sentido contrario; toda cifra nombra su ventana. SEO: imp 88->128 (+45%) y el clic organico AGUANTA en 1 (2a semana, /nba-predictions) = ya no fue casualidad. GA4 106->85 (uso propio via miniapp). ACCION sin cambio: rank-or-prune del footprint + proof-drip ES; no sobre-invertir en off-season. PENDIENTE HUMANO: el HOLD de @StatLineNerd vencio el 07-25 y el gate ya no bloquea, pero el label sigue sin verificar -> decidir release-hold o re-hold antes de reanudar X. || 2026-07-20 (weekly): proof FUERTE y estable -- Good all-time 19.5%->19.3% ROI (3134 picks, 62.5% hit, +605.8u; leve dilucion por volumen WNBA verano), slice 90d 8.0%->8.7%. HITO: PRIMER clic organico tras 9 semanas (0->1; GSC imp 68->88 +29%, pos 16.8) sobre /nba-predictions con query ES no-brand. GA4 130->106 (-18%; 26 usuarios/28d = uso repetido via miniapp), Umami 61 visitors. Telegram 3. Palanca sin cambio: rank-or-prune del footprint + proof-drip ES; StatLineNerd HOLD hasta 07-25 -> Bluesky; no sobre-invertir off-season. || 2026-07-13 (weekly): proof FUERTE -- Good all-time 19.5% ROI (3121 picks, 62.5% hit), slice 90d +8% (consolidando con WNBA). GA4 130 (plano; 26 usuarios/28d = uso repetido via miniapp). GSC clics 0 la 9a semana PERO impresionan queries ES no-brand (predicciones/prediccion/analisis nba) -> el footprint ES empieza a impresionar, aun lejos del clic. Telegram 3. Palanca SIN CAMBIO: rank-or-prune del footprint + proof-drip ES; StatLineNerd en HOLD hasta 07-25 -> Bluesky mientras; no sobre-invertir en off-season. || 2026-07-11 (weekly): proof FUERTE -- Good all-time 19.4% ROI (3114 picks, 62.5% hit), slice 90d +8.4% (era +5.9 -> mejorando con WNBA). GA4 127->125 (plano; solo 26 usuarios/28d -> mucho uso repetido/propio), Umami 59 visitors/124 visits. GSC imp 46->65, clics 0 (8a semana) PERO por primera vez impresionan queries ES no-brand ('predicciones nba' pos 40, 'prediccion nba' pos 33, 'analisis partidos nba') -> el footprint ES empieza a dar senal de impresion, aun lejos del clic. Telegram 3 plano. X: flag limpiado el mismo dia (decision usuario 07-11: no pausar publicaciones); @StatLineNerd sigue en HOLD deliberado hasta 07-25 por su label de X (lib/x_scheduler.py release-hold para reanudar antes) -> proof-drip via Bluesky mientras tanto. Palanca sin cambio: rank-or-prune + proof-drip ES; no sobre-invertir en off-season. || 2026-07-06 (weekly): Good-tier ROI FIRME en 19.2% (public API all-time; la prueba publica AGUANTA fuerte -- el slice 90d es +5.9%, mas blando por off-season pero positivo). GA4 sessions 117->127 (+8.5%) via WNBA-bridge; GSC clics SIGUEN 0 (7a semana, 37/46 imp = brand query 'proplab'). Palanca SIN CAMBIO: rank-or-prune del footprint programatico + proof-drip ES en @StatLineNerd/Bluesky; no sobre-invertir hasta que vuelva la temporada NBA. || 2026-07-01 (SEO audit): 2 of 3 June HIGH findings RESOLVED -- homepage canonical now present, title fixed to 'PropLab - Análisis inteligente de player props NBA', + new FAQPage schema. Fresh GSC (28d): 0 clicks / 36 imp / pos 9.1, but 29 of 36 imp = brand query 'proplab' -> the 1,521 programmatic /players/ pages earn ~ZERO impressions (indexed but never retrieved for prop queries, 6+ wks). #1 SEO fix: RANK-OR-PRUNE the footprint -- add ES data-driven analysis prose to the ~150-300 rotation-relevant players (citable, query-matchable) AND noindex the deep-bench long tail (< N games) to concentrate index signal. #2: add SoftwareApplication/Offer + Dataset schema (only Org/WebSite/FAQPage today). Full report: products/nbaproplab/FULL-AUDIT-REPORT.md || 2026-06-29: off-season, GA4 sessions +41% WoW (83->117) on WNBA-bridge content (top referral web.telegram.org, 17). GSC impressions flat & tiny (30->29) and clicks STILL 0 across 6 weeks -> programmatic prop pages are indexed but earn zero CTR; the only impression-getter is the brand query. Good-tier ROI 19.3% -> proof HOLDS, strongly positive (slight slip from 20.1; proplab MCP out of scope this run, KPI carried from the 06-28 collector). ACTION (unchanged lever, sharper): rewrite SERP titles/meta on the top programmatic player/prop pages to break the chronic 0-clicks (the ~1,521 indexed URLs are dead weight until they earn a single click) and keep the low-cost ES proof-drip on @StatLineNerd/Bluesky (Good-tier HOOK, show misses). Do NOT over-invest paid effort until the NBA season returns. Tweet ES picks via twitter_api_post.py --account StatLineNerd --pick-id.
2026-07-31 (weekly): STILL PARKED (sunset). SEGUNDA lectura negativa seguida -- GSC imp 275->261 (-5%), clics 11->8 (-27%), GA4 51->41 (-20%); waitlist SIGUE 0/0, 1 star, 3 clones/14d. Cambia el reparto interno: /blog/vertical-slice-architecture-dotnet pasa a ser la pagina top (3 clics/91 imp). El decaimiento sin refresh es EXACTAMENTE lo esperado en un dominio aparcado, NO una senal que exija accion. DECISION SIN CAMBIO: parked, sin build, sin re-pitch, cero gasto; el valor ya se trasvaso a devaisemanal (13 clics, #2 de ese sitio). || 2026-07-27 (weekly): STILL PARKED (sunset). PRIMERA semana negativa en todo: GSC imp 305->275 (-10%) y clics 15->11 (-27%) rompen 5 semanas de compounding pasivo; GA4 77->51 (-34%); clones 14d 12->2. Waitlist SIGUE 0, 1 star. El decaimiento sin refresh es lo ESPERADO en un dominio aparcado, no una senal nueva. DECISION SIN CAMBIO: cero gasto en superficies Sharpyard. El trasvase a devaisemanal ya esta hecho y rinde (claude-code-dotnet-csharp-guia = #2 pagina de DevAI con 12 clics). || 2026-07-20 (weekly): STILL PARKED (sunset). GSC compounding pasivo imp 275->305 (+11%), clics 14->15 (/blog/claude-code-for-dotnet lidera) pero GA4 93->77 (-17%) y waitlist SIGUE 0 (1 star, 9 cloners) = cero senal embudo/credibilidad. Decision sin cambio: cero gasto en superficies Sharpyard; cosechar el trafico .NET+Claude hacia el motor devaisemanal (ya rinde). || 2026-07-13 (weekly): STILL PARKED. GSC compounding pasivo (imp 171->275 +61%, clics 12->14; /blog/claude-code-for-dotnet 7 clics/130 imp lidera) pero GA4 120->93 (-23%) y waitlist SIGUE 0 (1 star, 9 cloners) = cero senal de embudo/credibilidad. Decision sin cambio: cero gasto en superficies Sharpyard; cosechar el trafico .NET+Claude hacia el motor devaisemanal (ya rinde 7 clics ES). || 2026-07-11 (weekly): STILL PARKED. GSC keeps compounding passively (imp 171->232 +36%, clicks 12->14; /blog/claude-code-for-dotnet 7 clicks/128 imp leads) while GA4 slid 120->103 (-14%); waitlist STILL 0, 1 star, 8 cloners -> no funnel/credibility signal. The harvest into devaisemanal ALREADY pays: the ES article claude-code-dotnet-csharp-guia earns 7 clicks/78 imp in DevAI's GSC. Zero spend on Sharpyard surfaces; decision unchanged. || 2026-07-06 (weekly): STILL PARKED, decision unchanged. WoW: GSC impressions 86->171 (+99%), clicks 5->12 (+140%) -- the .NET+Claude content keeps compounding in SEO (/blog/claude-code-for-dotnet 5 clicks/106 imp) -- but GA4 sessions 158->120 (-24%) and waitlist confirmed STILL 0 (1 star, 10 cloners) = no funnel/credibility signal. ONE ACTION unchanged: harvest the .NET+Claude traffic into the devaisemanal engine; no spend on Sharpyard surfaces. || 2026-06-29 STILL PARKED (no re-pitch). WoW: GSC impressions 36->86 (+139%), clicks 1->5, and the lead-magnet repo got its FIRST star (0->1) + 5 key events fired -- but waitlist confirmed is STILL 0 after 17 unique cloners, so the credibility/funnel signal is still absent (1 star != traction). Not enough to un-park. Only earner is /blog/claude-code-for-dotnet (4 clicks, 71 imp). DECISION UNCHANGED: domain parked, NO build, no new domain, no re-pitch. ONE ACTION: harvest the residual .NET+Claude content traffic into the devaisemanal GEO/content engine (cross-link / repurpose the claude-code-for-dotnet angle there) and stop spending on Sharpyard surfaces. Theses-dead reports: docs/research/2026-06-24-sharpyard-*.
2026-07-31 (audit mensual SEO): PRIMERA auditoria completa. Tecnicamente sano (robots abierto, GPTBot/ClaudeBot/PerplexityBot 200, www+http 301 al apex, SoftwareApplication+Offer en la home, 7/7 img con alt, TTFB 0.12s) y el mes revierte la tendencia: imp 1891->1924, clics 7->14 (+100%), pos media 38.1->30.3 (+7.8 puestos). ACCION #1 (SEO): consolidar y refrescar la pagina que YA convierte -- /blog/tpv-obligatorio-hosteleria-2026 tiene 582 imp y 7 de los 14 clics pero se atasca en pos 13.2 con tres problemas que se suman: (a) el titulo/URL dicen 2026 mientras el propio cuerpo y /blog/verifactu-aplazado-2027-restaurantes fechan los plazos en enero y julio de 2027; (b) canibalizacion con /blog/es-obligatorio-tener-tpv-bar (pos 9.0, solo 32 imp) que parte la query 'es obligatorio tener tpv en un bar' (66 imp, pos 19.1, 0 clics); (c) sin schema FAQPage pese a que casi todas las queries top son preguntas literales -> no se captura PAA ni cita de motores IA. Indicador: esa URL pasa de pos 13 a top-10 y de 7 a 25+ clics/28d. HALLAZGOS 2 y 3: 859 imp en paginas comerciales con CERO clics (/precios 413 imp pos 69.7, /tpv-franquicias-hosteleria 241 imp pos 15.7 -- esta ultima es puro problema de titulo/meta, no de ranking, y /tpv-pizzerias 205 imp pos 39.5); y NO hay llms.txt (404, geo:false) pese a que VeriFactu es una pregunta de busqueda-IA de manual en ES. Menor: /verifactu-bares solo lleva BreadcrumbList y saco 0 imp; el colector umami del hub NO lee eventos (el fix del sitio SI aterrizo: id fa7e47d9 + data-umami-event=whatsapp_click en el boton flotante), asi que el punto ciego de CTA es ahora del hub, no del producto; la service-account de Google da 403 en sc-domain:tpv-top.com (solo tira el token OAuth); tpvtop.com sin guion NO es nuestro y redirige a un tercero. Informe: products/tpvtop/FULL-AUDIT-REPORT.md. La prioridad de NEGOCIO sigue siendo la del weekly de hoy (cerrar go-lives via gestorias/AEHC + instrumentar el CTA); lo de arriba es la palanca SEO. || 2026-07-31 (weekly): MEJOR LECTURA DEL HUB -- GSC imp 1850->1924 (+4%, 3a subida seguida tras 5 semanas de caida), clics 12->14 (+17%, pos 31.2->30.3) y Umami SE DISPARA 32->59 visitors (+84%, 65 pageviews). El angulo VeriFactu ya arrastra una 2a pagina: tpv-obligatorio-hosteleria-2026 concentra 7 de los 14 clics (582 imp, pos 13.2) y ENTRA como-cambiar-de-tpv-sin-perder-datos con 2 clics en pos 7.1. PERO 0 eventos CTA en 28d: el trafico record no se atribuye a NI UN clic de WhatsApp/demo. ACCION: (1) SUBE A PRIORIDAD 1 -- arreglar el registro de eventos CTA (demo_click/email_click/WhatsApp) en ../TPVTop; con trafico ya real, seguir ciego es el desperdicio caro; (2) sin cambio: cerrar instalaciones via gestorias + AEHC, el CTR organico no es la palanca de este B2B. || 2026-07-27 (weekly): estabilizado -- GSC imp 1814->1850 (+2%, 2a semana al alza), clics 13->12, pos 31.2. /blog/tpv-obligatorio-hosteleria-2026 concentra 8 de los 12 clics (533 imp, pos 13.2): el angulo VeriFactu es EL unico convertidor. Umami 22->32 visitors (+45%) pero CERO eventos CTA en 28d -> con trafico ya real, el embudo sigue ciego. ACCION: (1) sin cambio de prioridad, cerrar instalaciones via gestorias + AEHC (el CTR organico no es la palanca de este B2B); (2) NUEVO: arreglar el registro de eventos CTA (demo_click/email_click/WhatsApp) en ../TPVTop -- ahora que llega trafico, seguir sin proxy de conversion es el desperdicio. || 2026-07-20 (weekly): GSC PRIMER repunte tras 5 semanas cayendo -- imp 1610->1814 (+13%), clics 11->13 (+18%, pos 33.3; tpv-obligatorio-hosteleria EL convertidor). Umami: el id publico fa7e47d9 ahora lee 22 visitors (era 6) = el fix data-website-id cosecha; 0 eventos CTA aun. KPI go-lives off-platform (sin dato). ACCION sin cambio: cerrar instalaciones via gestorias + AEHC; el CTR organico NO es la palanca de este B2B. || 2026-07-13 (weekly): LAGGARD -- GSC 5a semana cayendo (imp 1814->1610 -11%; clics 7->11 mejor semana, tpv-obligatorio-hosteleria EL convertidor). Umami: el colector lee el id publico fa7e47d9 y da solo 6 visitors/0 eventos CTA -> VERIFICAR que el fix data-website-id (fa7e47d9-bb74-45ee-a05b-c4e76eec5256) aterrizo en ../TPVTop; sigue casi ciego. KPI go-lives off-platform (sin dato). ACCION sin cambio: cerrar instalaciones via gestorias + AEHC; refresh cluster VeriFactu-2026 para frenar la caida. || 2026-07-11 (weekly): UMAMI MAL CABLEADO (hallazgo de hoy): la landing tpv-top.com envia eventos al website-id ADMIN (a17a69b2, dominio tpvtop-cloud-admin.vercel.app), no a tpvtop-public (fa7e47d9) que resuelve el colector por dominio -> el dashboard estaba CIEGO. Numeros reales (id admin, 28d): 56 visitors/67 visits (28d previos 81/119 = cayendo) y CERO eventos demo_click/email_click -> el proxy de clics WhatsApp/CTA no registra NADA en 28d. FIX en ../TPVTop: data-website-id -> fa7e47d9-bb74-45ee-a05b-c4e76eec5256 (layout Next.js). GSC: imp 1814->1669 (-8%, 4a semana cayendo) pero clics 7->11 (+57%, mejor semana = CTR sube; tpv-obligatorio-hosteleria-2026 6 clics/325 imp pos 10.7 sigue siendo EL convertidor). ACCION sin cambio: cerrar instalaciones via gestorias + asociaciones (AEHC); refresh del cluster VeriFactu-2026. || 2026-07-06 (weekly): SEO en TERCERA semana negativa consecutiva -- GSC impresiones 2080->1814 (-13% WoW), clics 6->7. Unico convertidor sigue /blog/tpv-obligatorio-hosteleria-2026 (274 imp, 4 clics), angulo VeriFactu. El CTR organico NO es la palanca de este B2B; KPI go-lives = off-platform, pull manual (sin dato esta semana). ACCION SIN CAMBIO: cerrar instalaciones via gestorias + asociaciones hosteleria (AEHC Cantabria); la caida sostenida justifica un refresh del cluster VeriFactu-2026, pero no es el embudo. || 2026-06-29: KPI = go-lives (off-platform, NO se mide en analytics; pull manual via WhatsApp/gestorias - sin dato automatizado esta semana). SEO en 2a semana NEGATIVA consecutiva: GSC impresiones 2630->2080 (-21% WoW, sigue cayendo desde la meseta), clics 7->6. El unico convertidor es /blog/tpv-obligatorio-hosteleria-2026 (273 imp, 2 clics) -> el angulo VeriFactu-obligatorio. PERO el CTR organico NO es la palanca de este negocio B2B. ACCION (sin cambio de prioridad): cerrar instalaciones via gestorias + asociaciones de hosteleria (AEHC Cantabria), canales de alta prioridad; refrescar/ampliar el cluster de contenido VeriFactu-2026 para frenar la caida de impresiones. NO activar Google Ads hasta limpiar claims rojos.
2026-07-31 (weekly): SIN CAMBIO estructural -- GSC imp 235->237 plano, los 3 clics organicos SE MANTIENEN (4a lectura seguida) y la pos media sigue mejorando 13.74->13.36. TODO el clic sale de requerimiento-aepd-10-dias (3 clics/87 imp, pos 3.8): el angulo de urgencia AEPD es el que engancha, ya confirmado. GA4 0 sesiones (congelado POR DISENO), Listmonk 1 unconfirmed. KPI LOI/depositos 0 porque la puerta NO se ha abierto -- no es fallo de marketing. ACCION humana sin cambio: 15 entrevistas diagnostico + preventa fria exigiendo 8-10 LOI ANTES de construir (../reg-desk/KIT-VALIDACION.md). || 2026-07-27 (weekly): SIN CAMBIO estructural -- GSC imp 209->235 (+12%) y los 3 clics organicos SE MANTIENEN (pos media 17.5->13.7). TODO el clic sale de una sola pagina: requerimiento-aepd-10-dias (3 clics/87 imp, pos 3.8) = el angulo de urgencia AEPD es el que engancha; empiezan a impresionar queries EN de AI Act. GA4 0 sesiones (congelado POR DISENO), Listmonk 1 unconfirmed. KPI LOI/depositos = 0 porque la puerta NO se ha abierto: no es un fallo de marketing. ACCION humana identica: 15 entrevistas diagnostico + preventa fria exigiendo 8-10 LOI ANTES de construir. Kit: ../reg-desk/KIT-VALIDACION.md. || 2026-07-20 (weekly): SIN CAMBIO estructural -- GSC imp 124->209 (+69%, dominio joven compounding) y PRIMEROS clics organicos 0->3 (pos 17.5; paginas AEPD requerimiento-aepd-10-dias pos 3.8, necesita-dpo-art-34 pos 7.4 prometen). GA4 lectura anomala 0 (el 07-19 marcaba 14; trafico congelado por diseno). Listmonk 1 unconfirmed. KPI LOI/depositos = 0 porque la puerta NO se ha abierto. ACCION humana identica: 15 entrevistas diagnostico + preventa fria exigiendo 8-10 LOI ANTES de construir. || 2026-07-13 (weekly): SIN CAMBIO -- GA4 plano en 14 (por diseno; engagement 7% = trafico frio), GSC imp 37->124 (+235%, dominio joven compounding; requerimiento-aepd-10-dias pos 3.8 y necesita-dpo-art-34 pos 7.4 = las paginas AEPD prometen), clics 0. Listmonk 1 unconfirmed. LOI/depositos = 0 porque la puerta NO se ha abierto. ACCION humana identica: 15 entrevistas diagnostico + preventa fria exigiendo 8-10 LOI ANTES de construir. || 2026-07-11 (weekly): SIN CAMBIO -- GA4 plano en 14 (por diseno; engagement 7% = trafico frio), GSC imp 37->107 (+189%, dominio joven compounding; requerimiento-aepd-10-dias pos 3.75 y necesita-dpo-art-34 pos 7.5 = las paginas AEPD prometen), clics 0. Listmonk 1 unconfirmed. LOI/depositos = 0 porque la puerta NO se ha abierto. ACCION humana identica: 15 entrevistas diagnostico + preventa fria exigiendo 8-10 LOI ANTES de construir. || 2026-07-06 (weekly): SIN CAMBIO -- analitica plana POR DISENO (GA4 14 sesiones), GSC imp 23->37 (+61%, dominio joven sirviendo /en/blog/data-breach-72-hours-dpo). LOI/depositos = 0 porque la puerta NO se ha abierto. ACCION humana/off-analytics: 15 entrevistas diagnostico a DPOs + preventa en frio exigiendo 8-10 LOI ANTES de construir. || 2026-06-25: PUERTA DE VALIDACION INVERTIDA (Paso A), sin cambio -- la analitica es plana POR DISENO y NO es la senal de validacion (GA4 sessions planas en 14 desde el deploy 06-20; GSC empieza a servir, 0->2 impresiones, /en/blog/data-breach-72-hours-dpo pos 8.5). LOI/depositos = 0 porque la puerta aun NO se ha abierto. ACCION (humana, fuera de analytics): 15 entrevistas diagnostico a DPOs + pre-venta del producto de pago a frio (LinkedIn + registro AEPD), exigir 8-10 depositos/LOI ANTES de construir el pipeline. NO medir altas gratis como exito (red-team: la AEPD regala el boletin). Kit: ../reg-desk/KIT-VALIDACION.md. Plan: ../reg-desk/BUSINESS-PLAN.md.
2026-07-31 (weekly): REVISION A CIEGAS -- el colector aimeter_local (SSH a produccion) devolvio null, asi que NO hay lectura de funnel nueva y el KPI 0/5 se ARRASTRA del 07-27 sin reverificar. ACCION (1) NUEVA y prioritaria: reparar/reintentar aimeter_local para no volver a revisar sin datos; (2) SIN CAMBIO: cero distribucion = cero lecturas nuevas, meter la auto-auditoria en conversaciones de alta intencion (umbral de lectura comercial 25 audits + 10 emails, sigue MUY lejos). || 2026-07-27 (weekly): funnel PLANO al 100% -- 1 audit, 1 email verificado, 2 email-qualified demand, 0 pro-intent, 0 leads de agencia = identico a 07-20; KPI 0/5. Ops sano (cola 0, 8/dia sin usar). DIAGNOSTICO: no es el producto, es que no hubo NINGUNA distribucion nueva -> cero lecturas nuevas. ACCION sin cambio pero con compromiso explicito: meter la auto-auditoria real en conversaciones de alta intencion esta semana (>=3) y recoger utilidad del informe + consentimiento de caso-estudio; sin eso el umbral de lectura comercial (25 audits + 10 emails) no llega nunca. || 2026-07-20 (weekly): funnel real (opctl m920q) -- 1 audit completado, 1 email verificado, 2 email-qualified demand, 0 pro-intent, 0 leads de agencia, 0 consentimientos caso-estudio; KPI 0/5 senales Pro. Ops sano (statusOk, cola 0, 8/dia). 1 visitante cualificado/30d, 21 exclusiones sinteticas. Sin GA4/GSC por diseno. Aun bajo el umbral de lectura comercial (25 audits + 10 emails). ACCION sin cambio: distribuir la auto-auditoria real solo en conversaciones de alta intencion; recoger utilidad del informe + consentimiento explicito de caso-estudio. || Distribute the real self-audit sample only inside high-intent conversations; collect report usefulness and explicit case-study consent.
Routines
No product-specific routines — covered by shared hub automation.
2026-07-31 (audit mensual SEO): PRIMERA auditoria completa. Es la mejor superficie tecnica del hub y no hay nada que arreglar ahi (robots con Allow explicito a GPTBot/ClaudeBot/PerplexityBot/Google-Extended, llms.txt excelente y fechado 2026-07-29, Dataset con dateModified en la home, Product+AggregateOffer+FAQPage+BreadcrumbList en /models/*, titulos que ya llevan precio y contexto, canonicals correctos, TTFB 0.12s, enlazado interno completo: /compare enlaza sus 127 y /deprecations sus 74 /migrate). Mes 1 del breakout: imp 46->1931 (+4098%), primeros 12 clics. ACCION #1: el sitemap tiene 434 URLs y CERO lastmod/changefreq/priority -- para un producto cuyo valor es 'refrescado a diario' eso es la herida autoinfligida grande: Google no sabe que cambio hoy, asi que prioriza por grafo de enlaces y por eso /models/* (200 enlaces desde la home) cosecha 1908 de las 1931 imp mientras /compare/* rinde 5 de 127 paginas (4%) y /migrate/* 13 de 74 (18%), juntas 42 imp = 46% del sitio invisible. NO es calidad: cuando salen rankean muy alto (/compare/command-a-vs-kimi-k2-7-code pos 2.0, /migrate/pixtral-12b pos 3.8 con clic, /migrate/gemini-2-0-flash pos 4.0). El dato ya existe en el refresh diario de ../Autopilot -> emitir lastmod real por URL en el generador de sitemap y pinchar IndexNow solo con lo cambiado. Indicador: /compare/* pasa de 5 a 40+ paginas con impresiones. HALLAZGOS 2 y 3: los 5 /guides/* (cheapest-llm-for-chatbots/rag/coding/summarization/vision) sacaron 0 impresiones en 28d pese a ser queries comerciales de alta intencion y estar destacados en llms.txt; y www.aimodelwatch.dev devuelve 200 con la home completa en vez de 301 al apex (se autocanonicaliza, pero GSC ya indexa https://www.aimodelwatch.dev/migrate/command-light y hay fila aparte de http://). OJO al leer el CTR 0.62%: buena parte de las impresiones son queries de agente IA con comillas y booleanos (ej. 'llama 4 behemoth' meta (available or released or ...) 12 imp pos 8.1) que nunca van a hacer clic -- es senal GEO buena, no fallo de snippet. Menor: /deprecations es la pagina top (88 imp, pos 11.3, 2 clics) y solo lleva Organization + un FAQPage de UNA pregunta, sin ItemList sobre la tabla de retiradas. Informe: products/aimodelwatch/FULL-AUDIT-REPORT.md. Confirma desde el lado SEO la tesis del weekly: ya no hay problema de visibilidad, solo de captura (1/100 subs, semana 5). || 2026-07-31 (weekly): SE CONFIRMA la tesis del 07-27 -- la palanca es CAPTURA, no visibilidad. GSC lateral tras el aplanamiento (imp 1961->1931 -2%, clics 12 planos, pos 20.9 clavada) mientras Umami SIGUE subiendo 238->262 visitors (+10%, 521 pageviews/28d) y el KPI sigue 1/100 en SEMANA 5. /deprecations se consolida como pagina top (2 clics/88 imp, pos 11.3) por delante de /compare y /migrate. ACCION: (1) reescribir el CTA de alertas EN /deprecations, que es la pagina que ya recibe la demanda, no en la home; (2) NUEVO: cablear el colector listmonk en product.yaml -- hoy el KPI se arrastra a mano y el producto se revisa a ciegas. || 2026-07-27 (weekly): CAMBIA LA PALANCA. GSC se APLANA tras 3 semanas de breakout (imp 1942->1961 +1%, clics 11->12) mientras Umami SIGUE subiendo (183->238 visitors, +30%) y el KPI sigue 1/100 (semana 4). Con trafico creciendo y captura a cero, el bloqueo YA NO es visibilidad sino conversion. Ademas cambia el mix: /deprecations es la pagina top (2 clics/87 imp, pos 11.4) por delante de /compare y /migrate. ACCION NUEVA: poner el CTA de alertas por email en /deprecations y /models/* (donde aterrizan de verdad, no en /alerts) y medir su clic; la geo-optimizacion pasa a segundo plano. || 2026-07-20 (weekly): GSC sigue subiendo -- imp 1435->1942 (+35%), clics 9->11 (+22%, pos 20.9); Umami 134->183 visitors (+37%, 28d). KPI subs 1/100 plano (Listmonk lista 7). El footprint programatico (/compare /migrate /models) cosecha semana 3. ACCION sin cambio: geo-optimizar las paginas que YA impresionan + CTA de alertas en /models/*; cerrar la decision GA4-vs-Umami (probablemente NO cablear GA4). || 2026-07-13 (weekly): GSC BREAKOUT sostenido -- imp 376->1435 (+281% en 7d), primeros clics organicos aguantan 0->9 (pos 19.2), sobre las paginas programaticas /compare /migrate /models. Umami 134 visitors/173 visits (28d). KPI ground-truth 1/100. ACCION sin cambio: aplicar geo-optimization a las paginas que YA impresionan + CTA de alertas en /models/* (ahi aterrizan, no en /alerts); cerrar la decision GA4-vs-Umami (probablemente NO se cablea GA4). || 2026-07-11 (weekly): GSC TAKEOFF -- impressions 376->1244 (+231%) and FIRST organic clicks (0->7, pos 18.0), landing on the programmatic pages (/compare/*, /migrate/*, /models/*) -> the footprint is harvesting. Umami 121 visitors/156 visits (28d). KPI ground truth: Listmonk list 7 = 1 subscriber (the 3 Umami subscribe events are dupes/retries; 1/100 goal). ACTIONS: apply the geo-optimization playbook to the pages already impressing + surface an alerts CTA on /models/* pages (visitors land there, not on /alerts); decide whether GA4 is still needed (with Umami + subscribe event probably NOT -- close the gap and document it). github collector null (private repo). || 2026-07-06 (weekly): GSC now SERVING query data (376 impressions, 0 clicks, avg pos 19.1; up from 264 the day before) -- blocker #1 'verify GSC to read query data' is DONE. Top-impressing pages: /, /about, /alerts. NEXT primitives: (1) wire GA4 (still null -> no session/conversion measurement), (2) apply the geo-optimization playbook to the pages already impressing, (3) decide a warmed dev social presence (Reddit/X) per anti-detection before any promo. Growth stays SEO/GEO-led + autonomous in CI. || 2026-06-28: Launched. Live on aimodelwatch.dev (Cloudflare Pages), 181 models/12 providers, daily self-improving GitHub Actions loop (refresh count -> Claude ships next roadmap item -> build -> deploy -> IndexNow). Growth is SEO/GEO-led and fully autonomous in CI. NEXT (needs local SPAM hub or human): (1) verify GSC for the domain to read query data, (2) decide on a warmed dev social presence (Reddit/X) for value-first distribution per hub anti-detection rules, (3) apply geo-optimization playbook.
Routines
No product-specific routines — covered by shared hub automation.
2026-07-31 (weekly): SE DESTAPA UN CIEGO -- lastIssueOpenRate 0.0 -> 0.2 (20%): el tracking de aperturas YA reporta, asi que el 0.0 de las ultimas semanas era instrumentacion y no una entrega muerta. KPI subs activos 5 PLANO tras el 4->5; sign_up 7/28d vs 5 miembros = la fuga alta->miembro se estrecha pero persiste. SEO RECUPERA el CTR que se erosionaba: imp 6252->6477 (+4%) y clics 80->82 (+2%, pos 9.74->9.60). ACCION: doblar en el cluster Claude Code (tutoriales-claude-code-aceptar 19 clics/1207 imp y claude-code-dotnet-csharp-guia 13 clics -- la cosecha de Sharpyard ya es la #2 del sitio) y atacar la fuga sign_up->miembro, que es el unico paso que separa trafico de KPI. || 2026-07-27 (weekly): EL KPI SE MUEVE -- subs activos 4->5 (+1), primer avance tras 3 semanas planas; la fuga alta->miembro se estrecha pero persiste (10 key events sign_up/28d vs 5 miembros). SEO en MESETA con CTR erosionando: imp 5998->6252 (+4%) pero clics 87->80 (-8%, pos 9.7) y GA4 293->268 (-9%). ACCION (2, en orden): (1) seguir cerrando la fuga alta->miembro-activo -- es el unico multiplicador real; (2) verificar el tracking de lastIssueOpenRate 0.0, que con 5 subs reales ya no se explica como lag de Ghost. Sigue el trasvase de Sharpyard: claude-code-dotnet-csharp-guia es ya la #2 pagina del sitio (12 clics). || 2026-07-20 (weekly): KPI subs activos PLANO en 4 (3a semana) PESE a 9 key events sign_up 7d -> la FUGA alta->miembro persiste (dupes/cancelaciones/confirm): DIAGNOSTICAR la fuga alta->miembro-activo antes que buscar mas trafico. Motor SEO sano (GSC 5493->5998 +9%, clics 75->87 +16% pos 10.0, GA4 293 plano). Superficies UTM sociales 0 sesiones (7a semana) -> RETIRAR y redoblar cluster copilot/claude-code + GEO. Vigilar lastIssueOpenRate 0.0. || 2026-07-13 (weekly): KPI subs PLANO en 4 (2a sem) PESE a 7 key events sign_up en 7d -> las altas NO cuajan en Ghost (dupes/cancelaciones/confirm): DIAGNOSTICAR la fuga alta->miembro-activo antes que buscar mas trafico. Motor SEO sano (GSC 4961->5493 +11%, clics 71->75, GA4 297). Superficies UTM sociales 0 sesiones 6a semana -> RETIRAR y redoblar cluster copilot/claude-code (40 clics/5 paginas) + GEO. Vigilar lastIssueOpenRate 0.0. || 2026-07-11 (weekly): KPI subs activos PLANO en 4 (el +1 del 07-06 no continuo; 2 key events sign_up en 7d ambos DIRECTOS pero Ghost sigue 4 -> vigilar cancelaciones/dupes). Motor SEO sano: GSC imp 4961->5516 (+11%), clics 71->73 (pos 9.77), GA4 302 (28d), Umami 234 visitors. ATRIBUCION 7d: CERO sesiones de las 6 UTM sociales (5a semana consecutiva) y CERO de los buckets UTM internos -> la conversion llega por directo/organico; considerar RETIRAR las superficies sociales UTM muertas y redoblar el cluster copilot/claude-code (5 paginas top = 39 clics GSC) + GEO (perplexity.ai 4 ses, hub.getaimeter.com 9). VIGILAR: lastIssueOpenRate 0.0 (lag de Ghost o issue sin tracking?). || 2026-07-06 (weekly): KPI (subs activos) 3 -> 4 -- PRIMER movimiento del KPI en 5+ semanas, tras el rework 06-29 de CTA+UTM internos en las 6 paginas top de GSC. GA4 28d 288->298 (+3.5%), GSC clics 66->71 (+7.6%, pos 9.58); sigue liderando el cluster github-copilot-ai-credits. El motor SEO/GEO esta sano y la conversion de alta POR FIN se mueve. FOCO: confirmar la atribucion del +1 en los buckets UTM internos (copilot_credits/claude_code_tutoriales) antes de declarar victoria; seguir cosechando el cluster copilot. || 2026-06-29: el trafico SEO RE-ACELERO tras la meseta (GA4 sessions 203->288 +42% WoW, GSC impresiones 4395->4833 +10%, clics 55->66 +20%), liderado por el cluster github-copilot-ai-credits. PERO el KPI (subs activos) sigue PLANO en 3 por 5a semana. DIAGNOSTICO RE-CONFIRMADO con run_report GA4 fresco (7d): CERO de las 6 superficies UTM (weekly_thread/weekly_teaser/latam_value/midudev_recursos/spanish_digest/es_creator_swap) genero NINGUNA sesion etiquetada esta semana -- todo es directo/organico/referral; solo 1 key event en 7d, atribuido a directo. O sea las superficies sociales NO generan clics medibles Y el embudo de alta no se puede atribuir. FOCO (instrumentacion primero): verificar/arreglar el evento sign_up en el portal Ghost via ga_admin.py (primitiva del KPI; sin el, el embudo es ciego), LUEGO reforzar CTA+UTM de enlace interno desde las paginas top de GSC (github-copilot-ai-credits-*, tutoriales-claude-code-*) al portal. GEO/LLM sigue siendo el unico canal con conversion atribuida (copilot.com ai-assistant esta semana). PENDIENTE HUMANO: login daily.dev + Flipboard (2/8 crossposts fallan), rotar claves Ghost/Mailgun. || ACTIONED 2026-06-29 (post-review): el 'instrumentacion primero' era FALSA ALARMA -- el evento sign_up esta SANO y contando (no era el bloqueo). DONE en vivo: (1) 6 paginas top de GSC reescritas con CTA+UTM por-cluster (copilot_credits/claude_code_tutoriales/cli_agentes/tokens_coste) -> las conversiones de embudo interno ya se atribuyen por tema; (2) nuevo articulo ES 'Claude Code para .NET y C#' publicado via pipeline evergreen (cosecha el trafico .NET+Claude de Sharpyard, que sigue parked). Proximo review: vigilar los nuevos buckets UTM internos.
Routines
Weekly ES Bluesky teaser for DevAI Semanal (@devaisemanal.bsky.social) after the Tuesday newsletter publishes8 runs
ScheduleAt 12:06 PM, only on TuesdayLast2d agoNextin 1d
2d agookTeaser ES del numero MCP Apps: la UI embebida no es autorizacion, el servidor valida usuario/tenant/politica igual que ante un cliente HTTPpost ↗
5d agookTeaser ES del numero 2026-07-28: guia de contexto de Anthropic para Claude 5, patrones de Claude 3 ahora contraproducentespost ↗
12d agookTeaser ES del numero 2026-07-21: 0day de Cursor con full disclosure tras negarse el vendor a parchearpost ↗
19d agookTeaser issue 2026-07-14: overhead de tokens de Claude Code (33k vs 7k OpenCode = 4.7x coste)post ↗
26d agookTeaser 2026-07-07: prompt engineering no murio (VS Code+OpenAI bajan tool calls/tokens en GPT-5.5)post ↗
2026-07-31 (weekly): el 90d REBOTA -14.52% -> -11.98% (101 picks, 34.7% hit, -12.1u) pero es la MISMA mecanica de ventana al reves, NO mejora del modelo: sin picks Good desde 2026-05-17 la ventana solo cubre 05-02..05-17 y se VACIA a mediados de agosto. All-time PLANO +4.35% (311 picks, +13.54u) = la cifra literal de los posts. Honest-body SIGUE ACTIVA (90d <= 0): ventana explicita en toda cifra ROI, ningun hook con ROI positivo; NO es flag de plataforma. Por mercado H2H +14.8% y H2HH1 +14.1% POSITIVOS, el dano esta en BTTS (-57%) y OU2.5 (-45%). Trafico: GSC imp 531->633 (+19%), clics 5->6, pos 48.1->46.0 desde base muy baja; GA4 82->76. ACCION: (1) angulos no-ROI y ventana explicita; (2) apoyar el hook en H2H, el unico mercado con proof positivo; (3) AVISO DE RELOJ: sin motor de scoring de vuelta con el club, a mediados de agosto el KPI 90d deja literalmente de existir -- decidir antes si el dashboard pasa a citar all-time. || 2026-07-27 (weekly): Good-picks ROI 90d -11.45% -> -14.52% (104 picks, 33.7% hit, -15.1u). Es MECANICA de ventana pura, no deriva del modelo: sin picks Good desde 2026-05-17, la ventana solo contiene 05-01..05-17 y se VACIA a mediados de agosto. All-time PLANO +4.35% (311 picks). Honest-body ACTIVA (verificado en footballlab): toda cifra nombra su ventana; ningun post lidera con ROI positivo mientras 90d<=0; NO es flag de plataforma, NO se pausa nada. Dano concentrado en OU2.5 (-48.9%) y BTTS (-62.4%); H2H sigue +14.8%. ACCION: el desbloqueo real NO es de marketing sino el reinicio del motor de scoring con el club (agosto) -- hasta entonces, Bluesky honest-proof + angulos no-ROI (metodo, H2H, lecturas de mercado) y track-record machine-readable. Trafico GA4 89->82, GSC 526->531 plano, go.bsky.app unico referral vivo. Telegram 3. || 2026-07-20 (weekly): Good-picks ROI 90d -3.14% -> -11.45% (190 picks/-21.75u) -- sigue cayendo EXACTAMENTE como predicho (rueda fuera de la semana rentable 04-07..04-11; 0 picks Good nuevos desde 05-17, hiato club hasta agosto post-WC). All-time PLANO +4.35% (311 picks) = cifra literal de Bluesky. Honest-body ACTIVA (verificado footballlab): toda cifra nombra ventana; ningun post lidera con ROI positivo mientras 90d<=0; NO es flag de plataforma. Trafico GA4 94->89, GSC imp 461->526 (+14%), go.bsky.app #1 referral. Telegram 3. Palanca sin cambio: Bluesky honest-proof + track-record machine-readable. || 2026-07-13 (weekly): Good-picks ROI 90d SIGUE negativo (-3.14%, 249 picks/-7.81u) -> reglas honest-body SE MANTIENEN (toda cifra ROI nombra ventana; ningun post lidera con ROI positivo mientras 90d<=0; citar all-time +4.35% SOLO con su etiqueta). Estructural: 0 picks Good nuevos desde 05-17 (hiato club hasta agosto) -> el 90d no recupera sin picks; NO es flag de plataforma. Trafico: go.bsky.app #1 referral (HOOK build-in-public), chatgpt.com primeras citas. StatLineNerd en HOLD hasta 07-25. Palanca sin cambio: Bluesky honest-proof + track-record machine-readable. || 2026-07-11b (user decision): NO pausar publicaciones -- flag LIMPIADO 07:57 UTC mismo dia (X FutProbLab/DevAI reanudados; StatLineNerd sigue en HOLD deliberado hasta 07-25 por su label). En su lugar: FIX DE CUERPOS en todos los generadores FutPicks (regla 1: toda cifra ROI nombra su ventana explicita; regla 2: mientras 90d <= 0, ningun post lidera con ROI positivo -- transparencia con AMBOS numeros o angulos no-ROI). Playbook actualizado (la regla de pausa queda sustituida); record-flag reservado para flags reales de plataforma. || 2026-07-11 (weekly): PAUSA-PROOF DISPARADA -- Good-picks ROI 90d +4.35 -> -0.3% (276 picks desde 04-12; regla <=0 cumplida) -> record-flag EJECUTADO 07-11 07:42 UTC (3 cuentas X pausadas 48h; es la pausa-proof pre-comprometida el 07-06, NO un flag de plataforma). Causa: TODO el profit vivo (+14.49u) vino de 04-07..04-11 (primeros 5 dias, 35 picks); ex-primera-semana el registro es plano/negativo (-0.95u en 276); sin picks nuevos hasta el club (agosto) el 90d NO recupera. All-time sigue +4.35% (311 picks, +13.54u) = lo que citan los posts de Bluesky (literalmente cierto pero apoyado en esa semana). PENDIENTE HUMANO: parar los posts proof-ROI de Bluesky (slot ~13:21 diario; ContentEngine canal fut / routine local) mientras el 90d sea <=0; corregir el KPI del routine (hoy escribio 4.35 = all-time mal etiquetado como 90d). Trafico: GA4 144->106 (-26%), GSC imp 342->443 (+30%), clics 6 pos 46; go.bsky.app #1 referral (41 ses vs google 6); chatgpt.com 3 ses (primeras citas LLM). Telegram 3 plano. || 2026-07-06 (weekly): Good-picks ROI 90d PLANO en +4.35% -> proof HOLDS, promo ON (si <=0 PAUSA total via lib/x_scheduler.py record-flag). GA4 134->144 (+7.5%), GSC imp 164->342 (+109%). go.bsky.app SIGUE #1 referral (53 sesiones > Google 6): el HOOK build-in-public de Bluesky es el canal medible. Modo Mundial pick-free (hasta 07-19). Palanca SIN CAMBIO: track-record machine-readable (Dataset schema) para convertir el desbloqueo de crawlers en citas LLM + doblar en Bluesky. || 2026-07-01 (SEO audit): the 2026-06 CRITICAL (robots.txt blocked all AI crawlers) is RESOLVED & verified holding -- catch-all Allow, llms.txt LIVE (cites +8.0% ROI / 65.9% vs Pinnacle), /compare pages SSR. Fresh GSC (28d): 4 clicks / 224 imp / avg pos 42.3 = indexed-not-ranking on the RIGHT queries (calculador de momios/parlay pos ~50, 'picks mundial hoy' pos 44). Top SEO levers now: (1) make the track-record machine-readable (Dataset/structured markup) to convert the GEO unblock into actual AI citations; (2) climb /calculadora + /world-cup off page 4 with on-page depth on the queries already impressing. CWV still poor (LCP 4.5s / CLS 0.225, ~205KiB unused JS) -- wire GOOGLE_API_KEY for field data. Full report: products/futpicks/FULL-AUDIT-REPORT.md || 2026-06-29: WC mode (06-11..07-19, pick-free per docs/wc-runbook.md) -- traffic accelerating WoW (GA4 sessions 103->134 +30%, GSC impressions 93->164 +76%, clicks 2->4). WIN: go.bsky.app is now the #1 referral source (49 sessions, ahead of Google) -> the Bluesky build-in-public HOOK is finally driving measurable traffic; double down there. WATCH: Good-picks ROI % (90d) slid 8.0->4.35 -- still POSITIVE so proof holds and promo stays ON, but the decay is structural (pick-free WC window: strong April picks age out of the rolling 90d, negative May remains). If it hits <=0, PAUSE promo across all surfaces (python lib/x_scheduler.py record-flag). Still-gating instrumentation (UNDONE): wire the GA4 purchase/pro_subscribe event so Stripe Pro conversions (primary KPI, 0 by design, keyEvents 0) become measurable. SEO still page-4 (pos 39.5), indexed-not-ranking -> push SERP CTR on /world-cup + /calculadora. Plan: docs/research/2026-06-10-futpicks-plan.md || ACTIONED 2026-06-29 (post-review): the 'wire the purchase event (UNDONE)' item was a STALE FALSE ALARM -- Stripe + the GA4 'purchase' key event are already wired & live; KPI=0 is genuinely zero paid customers, not a measurement gap. DONE: fixed the one real defect (purchase fired transaction_id only -> 0 revenue) by adding value/currency/items (resolved at runtime from /billing/config) in ../Furbov2 footballlab-web -- needs a normal CI deploy. Bluesky double-down LIVE (1 post + 6 queued daily, honest +4.35% ROI proof). Decision pending: trial-start vs first-charge revenue booking semantics.
Routines
Daily FutPicks HQ Discord feed to #análisis - the EXPERIMENTAL WC model's predictions + odds + edge (matches the live site; F1-lint-gated, EXPERIMENTAL framing). Server + webhook live.20 runs
ScheduleAt 10:01 AM, every dayLast22h agoNextin 1h
22h agopartialSKIPPED: sin lecturas del modelo (picksToday=0, evaluate_match sin resultados) — no se publico nadamatches_upcoming: 24model_reads: 0product: futpicks
2d agopartialSKIPPED: 18 upcoming matches (Eliteserien/Allsvenskan/Superliga) but scoring engine returned 0 model reads (picksToday=0, evaluate_match results empty). Nothing posted.upcoming_matches: 18model_reads: 0posted: 0
2d agopartialSKIP: sin lecturas del modelo (0 odds snapshots hoy, picks board vacio); no se publico nada en #analisismatches_upcoming: 17model_reads: 0odds_snapshots_today: 0
5d agopartialSKIPPED: no model reads (day 7). Board empty for 2026-07-28 at 08:0xZ (all ratings + preview), flab_picks_today=[], flab_evaluate_match returns results:[] on UCL fixtures 2699/2507. NEW DIAGNOSIS: not just the T-3h window - the 07-26 board shows 8 pick records (AM+PM passes, settled with real results) where EVERY model field is zero (score/odds/edge/modelProbability/fairOdds/dataQuality/all 7 blocks, selection empty, blockDetails []). Engine runs and settles but emits null scoring, so rating always lands Marginal/Weak and Good can never repopulate. Track record unchanged: last settled Good pick 2026-05-17, 311 picks, 35.7 pct hit, ROI +4.35 pct all-time.matches_upcoming_3d: 23picks_today: 0board_today: 0
6d agopartialSKIPPED: no model reads at run time (08:03Z). Public Good board flab_picks_today=[] and matches_today picks=[] on all 3 fixtures, picksToday=0. NOT evidence the engine is dark: picks are scored in a T-3h preview window (today's kickoffs 17:00Z => ~14:00Z), and on 07-26 picksToday went 0 at 08:02Z -> 12 at ~16:01Z. This 08:0xZ slot is structurally upstream of scoring, so it will skip every day regardless of model health. Second gate: on 07-26 all 12 scored picks were internal Marginal/Weak, so the public Good board stayed empty anyway. Odds ingestion healthy (2349 snapshots, 7-8 books incl. Pinnacle). Nothing posted.picks_today: 0odds_snapshots: 2349upcoming_matches: 42
7d agopartialSKIPPED: no model reads. Club leagues back + 2429 odds snapshots today, but scoring engine returns empty (picksToday=0, picks board [], evaluate_match results=[] on 3 leagues). Last settled pick 2026-05-17. Nothing posted.picks_today: 0odds_snapshots: 2429upcoming_matches: 44
8d agookSKIP: model dark (picks board empty 07-25/07-26, flab_evaluate_match returns no markets on 6 fixtures across Allsvenskan/Eliteserien/Superliga/UCL-Q2); club season restarted + odds flowing, scoring engine still not emittingmatches_upcoming: 44model_reads: 0last_settled_pick: 2026-05-17
9d agookSKIP: model dark (empty picks board + evaluate returns no markets); no model reads to postproduct: futpicks
10d agookSKIP: no model reads - picks board empty for 2026-07-23; all 9 EL 2nd-qualifying matches returned empty evaluate results (no odds ingested); nothing postedproduct: futpicks
11d agookSKIP: no model reads - picks board empty (today+tomorrow, all ratings incl. preview), picksToday=0. WC over 07-19; football off-season (last settled Good pick 05-17). Only CL/EL qualifiers upcoming, unrated. Did NOT post per guardrail.reads: 0matches_upcoming: 12product: futpicks
14d agoskippedSkipped: no derivable model reads. WC Final (Spain v Argentina) is displayOnly; picks board/today empty; CL/EL Jul21-23 qualifiers return empty evaluate results. No post per SKIP rule.product: futpicks
15d agoskipSkipped: no model reads available (picks board empty; WC 3rd-place final + final are displayOnly with no model output; CL/EL qualifiers 3+ days out, results:[]). Skip clause honored, no post.product: futpicks
ScheduleAt 04:04 PM, every dayLast16h agoNextin 7h
15h agookEncuesta de jornada en #generalproduct: futpicks
1d agookEncuesta de jornada en #generalproduct: futpicks
2d agookEncuesta de jornada en #generalproduct: futpicks
5d agookEncuesta de jornada en #generalproduct: futpicks
6d agookEncuesta de jornada en #generalproduct: futpicks
7d agookEncuesta de jornada en #generalproduct: futpicks
8d agookEncuesta de jornada en #generalproduct: futpicks
9d agookEncuesta de jornada en #generalproduct: futpicks
10d agookEncuesta de jornada en #generalproduct: futpicks
11d agookEncuesta de jornada en #generalproduct: futpicks
12d agookEncuesta de jornada en #generalproduct: futpicks
13d agookEncuesta de jornada en #generalproduct: futpicks
+ 20 earlier runs
Daily FutPicks #historial settled-results recap (hits AND misses, red accent on misses) via build_results_recap. WC pick-free, F1-gated. Animador voice.23 runs
ScheduleAt 11:05 PM, every dayLast8h agoNextin 15h
8h agookRecap de resultados en #historialproduct: futpicks
1d agopartialSin resultados liquidados desde la ultima ejecucion (2026-07-31) - no se publica recapsettled_since_last_run: 0product: futpicks
2d agookRecap de resultados en #historialproduct: futpicks
6d agookRecap de resultados en #historialproduct: futpicks
7d agookRecap de resultados en #historialproduct: futpicks
8d agopartialSin resultados liquidados desde la ultima ejecucion (2026-07-24) - no se publica recapsettled_since_last_run: 0product: futpicks
9d agookRecap de resultados en #historialproduct: futpicks
10d agookRecap de resultados en #historialproduct: futpicks
11d agookRecap de resultados en #historialproduct: futpicks
13d agookRecap de resultados en #historialproduct: futpicks
17d agookRecap de resultados en #historialproduct: futpicks
21d agookRecap de resultados en #historialproduct: futpicks
2026-07-31 (weekly): RESPIRO tras el pico -- GSC imp 3446->3263 (-5%) pero clics 32 CLAVADOS y la pos media MEJORA 19.05->18.77. SIGUE siendo el LIDER del hub en clics organicos absolutos y la base ensanchada AGUANTA (home 14 clics/394 imp, plantilla-oferta-tecnica 3, como-leer-pliegos-pcap 2 en pos 8.0, garantias-avales 2). GA4 cookieless 22 plano, keyEvents 0. ACCION SIN CAMBIO pero MAS URGENTE: instrumentar el KPI de altas gratis activadas (#232). Con 5 semanas de SEO compounding y cero medicion del North Star, el trafico llega y no sabemos si activa -- es ya EL unico bloqueo serio del producto. || 2026-07-27 (weekly): GANADOR DEL HUB -- SEO compounding semana 5 y MEJOR semana hasta hoy: GSC imp 2813->3446 (+23%), clics 23->32 (+39%, pos 19.0). Y se ENSANCHA la base: entran al marcador plantilla-oferta-tecnica (3 clics/79 imp) y como-leer-pliegos-pcap (2 clics, pos 7.9) junto a home (14) y garantias-avales (2); primera cita LLM (chatgpt.com). ACCION: (1) doblar el patron que YA funciona -- plantillas y guias de lectura de pliegos, que es lo que acaba de empezar a convertir; (2) DESBLOQUEAR #232 (GA4 de altas): con este volumen entrando, el North Star ciego es ya el unico bloqueo serio del producto. || 2026-07-20 (weekly): SEO COMPOUNDING semana 4 -- GSC imp 2093->2813 (+34%), clics 15->23 (+53%, pos 19.5); GA4 18->24 (+33%, 100% Spain, bsky+go.bsky lideran). El fix de prerender (#259) sigue rindiendo. ACCION sin cambio: doblar clusters que YA impresionan (contratos-menores pos 6.1, garantias-avales) + profundizar baja-temeraria. PENDIENTE Fase 1 sin cambio: GA4 real de altas (#232, North Star ciego sin el), E-E-A-T (#233), Brevo (#234), LinkedIn (#237 manual). || 2026-07-13 (weekly): SEO COMPOUNDING semana 3 -- GSC imp 1275->2093 (+64%), clics 5->15 (+200%, pos 20.3); el fix de prerender (#259) sigue rindiendo. ACCION sin cambio: doblar contenido en clusters que YA impresionan (contratos-menores pos 6.1, garantias-avales) y profundizar baja-temeraria (193 imp, pos 42 = mayor upside). PENDIENTE Fase 1 sin cambio: GA4 real de altas (#232, sin el no se mide el North Star), E-E-A-T (#233), Brevo (#234), LinkedIn (#237 manual). || 2026-07-11 (weekly): SEO COMPOUNDING semana 2 -- GSC imp 1275->2011 (+58%), clics 5->14 (+180%, pos 20.3). contratos-menores-2026 pos 6.1 (229 imp) y garantias-avales (2 clics) lideran; baja-temeraria 179 imp pos 41.9 = el mayor upside on-page (profundizar ese contenido). Brand 'licitaia' 4 clics pos 1.8. GA4 cookieless 14 ses (bsky 5 + x 1 + google 3; 100% Spain). ACCION: doblar contenido en los clusters que YA impresionan. PENDIENTE Fase 1 sin cambio: GA4 real de altas (#232), E-E-A-T (#233), Brevo (#234), LinkedIn (#237 manual). Colector: gsc.py crash cp1252 en Windows ARREGLADO 07-11; github collector null (repo privado). || 2026-07-06 (weekly): BREAKTHROUGH SEO -- GSC impresiones 109 -> 1275 (~11x WoW), primeros clics organicos 0 -> 5. El fix de prerender (PR #259, HTML+JSON-LD a crawlers) no solo desbloqueo la indexacion sino que YA COMPOUNDEA: primer salto real de visibilidad. GA4 13 sesiones (bsky.app 4 + x 1 = la automatizacion social MWF empuja). El track A (contenido/SEO/GEO) esta funcionando; doblar en el contenido de los clusters que impresionan. PENDIENTE Fase 1 sin cambio: instrumentar GA4 real de altas (#232), autor E-E-A-T (#233), lifecycle Brevo (#234), LinkedIn (#237 manual) -- sin GA4 automatizado aun no se mide el North Star (altas gratis activadas). || 2026-06-26: ESTRATEGIA GTM aprobada (A contenido/SEO/GEO + B free-tools/PLG, C comunidad de contingencia). Spec: ../Spam/docs/superpowers/specs/2026-06-26-licitaia-gtm-strategy-design.md. Ejecución en issues/milestones de github.com/Khavel/licitaia (label gtm; Fases 1-4). Fase 1 en curso: ARREGLADO el bloqueador #1 -- las URL prerenderizadas (sin barra) servían el shell CSR vacío a los crawlers; ahora sirven HTML prerenderizado + JSON-LD (PR #259) -- CONFIRMADO 06-29: GSC impresiones 0->109 WoW, las paginas ya se sirven/indexan (clics 0, pos 15.2). Datos legales corregidos (umbrales SARA #257, baja temeraria art.149/85 #258). Referral surfaced (#261). PENDIENTE Fase 1: instrumentar analítica (#232), entidad de autor real E-E-A-T (#233, NECESITA identidad del fundador), lifecycle Brevo (#234), página LinkedIn (#237, MANUAL). North Star = altas gratis ACTIVADAS + retención, NO MRR. Constraint: ~2h/sem fundador, 0€, inbound.
Routines
Bluesky posting for LicitaIA (@licitaia.org), 3x/week, promoting the free pliego analyzer + guides to drive free signups (auto-publish).14 runs
ScheduleAt 02:04 PM, only on MondayLast2d agoNextin 5h
2d agopartialSKIP (no post): ya habia salido un post hoy 2026-07-31 a las 06:28 UTC (tool calculadora umbral SARA, rkey 3mrwd7eifvo23). La cuenta postea cada 2-3 dias (07-24, 07-27, 07-29, 07-31) y publicar de nuevo 5,6h despues romperia la cadencia. Sin prioridad de guia nueva: el ultimo commit a blog-data.ts es 07-24 (acuerdo marco), ya publicado. Proximo slot sugerido: 08-02, toca GUIA (van 2 tools seguidos 07-31 y 07-22 y la ultima guia sin destacar reciente seria una evergreen fuera de la ventana de 3 semanas).product: licitaiapost-de-hoy ↗
2d agookTOOL rotation: calculadora umbral SARA (/calculadora-umbral-sara) tras 3 guias seguidas (07-24, 07-27, 07-29). Angulo nuevo vs el 07-10 (que fue valor estimado vs presupuesto base): que GANA el licitador si es SARA - plazos ampliados + recurso especial gratuito art. 44 LCSP. Nota: atproto faltaba en el entorno, reinstalado (requirements.txt:14).post ↗
6d agookGuia revision de precios (merged 07-18, nunca destacada: los slots 07-20/07-22 fueron tools y el 07-24 fue la guia mas nueva de acuerdo marco). Hook art.103 LCSP: no es subida automatica por inflacion; salvo energia hace falta 20% ejecutado + 1 anio desde formalizacion.post ↗
9d agookPosted NEW guide (merged today): acuerdo marco / contratos basados, arts. 219-221 LCSP, 4yr max duration, GO/NO-GO for pymespost ↗
11d agooksimulador baja temeraria tool: umbral no es % fijo, depende del nº de licitadores (art. 85 RGLCAP)post ↗
20d agookSimulador baja temeraria tool + freshly-deepened guide (art.149 LCSP: justificacion antes de excluir; garantia complementaria 5% extra)post ↗
23d agookFree TOOL rotation: calculadora umbral SARA (/calculadora-umbral-sara). Hook: SARA se mide sobre valor estimado (art.101 LCSP), no el presupuesto base; umbral 2026-2027 servicios/suministros 140.000 EUR (AGE) o 216.000 EUR (resto). First tool feature since 06-29 pliego analyzer; broke 5-post guide streak.product: licitaiapost ↗
25d agookGuia subcontratacion en contratos publicos (NUEVA guia, mergeada hoy): responsabilidad no se traslada + tareas criticas (art.215 LCSP)post ↗
27d agookNew guide (merged today): modificacion-contratos-publicos — 20% limit art. 204 LCSP, review the clause before biddingpost ↗
2026-07-31 (weekly): proof ESTABLE -- Good all-time 19.3% ROI PLANO (3146 picks, 62.5% hit, +607.2u; +5 picks WNBA desde el 07-27). El slice 90d cede 17.2%->13.9%, otra vez MECANICA de ventana (entran dias flojos de julio, salen los buenos de finales de abril); el all-time, que es la cifra publicable, no se mueve. SEO: imp 128->140 (+9%) y el clic organico SE MANTIENE en 1 (3a lectura consecutiva, /nba-predictions 62 imp pos 30.5); la home ya impresiona en pos 4.7 (60 imp) sin llevarse el clic. ACCION: (1) PENDIENTE HUMANO arrastrado 2a revision -- el HOLD de X @StatLineNerd vencio el 2026-07-25 y el label SIGUE sin verificar: decidir release-hold o re-hold, no dejarlo colgado otra semana; (2) trabajar titulo/meta de la home, que rankea en pos 4.7 y no convierte. || 2026-07-27 (weekly): proof estable -- Good all-time 19.3% PLANO (3141 picks, 62.5%, +606.4u). El slice 90d salta 8.7% -> 17.2% pero es MECANICO (sale de la ventana el bloque perdedor 04-21..04-27: 172 picks Good, 44% hit); 90d solo-WNBA = 18.9% (347 picks) -> la WNBA sostiene el proof en off-season. GUARDRAIL: NO publicar ese salto como mejora del modelo -- es el mismo efecto de ventana que hunde a FutPicks, en sentido contrario; toda cifra nombra su ventana. SEO: imp 88->128 (+45%) y el clic organico AGUANTA en 1 (2a semana, /nba-predictions) = ya no fue casualidad. GA4 106->85 (uso propio via miniapp). ACCION sin cambio: rank-or-prune del footprint + proof-drip ES; no sobre-invertir en off-season. PENDIENTE HUMANO: el HOLD de @StatLineNerd vencio el 07-25 y el gate ya no bloquea, pero el label sigue sin verificar -> decidir release-hold o re-hold antes de reanudar X. || 2026-07-20 (weekly): proof FUERTE y estable -- Good all-time 19.5%->19.3% ROI (3134 picks, 62.5% hit, +605.8u; leve dilucion por volumen WNBA verano), slice 90d 8.0%->8.7%. HITO: PRIMER clic organico tras 9 semanas (0->1; GSC imp 68->88 +29%, pos 16.8) sobre /nba-predictions con query ES no-brand. GA4 130->106 (-18%; 26 usuarios/28d = uso repetido via miniapp), Umami 61 visitors. Telegram 3. Palanca sin cambio: rank-or-prune del footprint + proof-drip ES; StatLineNerd HOLD hasta 07-25 -> Bluesky; no sobre-invertir off-season. || 2026-07-13 (weekly): proof FUERTE -- Good all-time 19.5% ROI (3121 picks, 62.5% hit), slice 90d +8% (consolidando con WNBA). GA4 130 (plano; 26 usuarios/28d = uso repetido via miniapp). GSC clics 0 la 9a semana PERO impresionan queries ES no-brand (predicciones/prediccion/analisis nba) -> el footprint ES empieza a impresionar, aun lejos del clic. Telegram 3. Palanca SIN CAMBIO: rank-or-prune del footprint + proof-drip ES; StatLineNerd en HOLD hasta 07-25 -> Bluesky mientras; no sobre-invertir en off-season. || 2026-07-11 (weekly): proof FUERTE -- Good all-time 19.4% ROI (3114 picks, 62.5% hit), slice 90d +8.4% (era +5.9 -> mejorando con WNBA). GA4 127->125 (plano; solo 26 usuarios/28d -> mucho uso repetido/propio), Umami 59 visitors/124 visits. GSC imp 46->65, clics 0 (8a semana) PERO por primera vez impresionan queries ES no-brand ('predicciones nba' pos 40, 'prediccion nba' pos 33, 'analisis partidos nba') -> el footprint ES empieza a dar senal de impresion, aun lejos del clic. Telegram 3 plano. X: flag limpiado el mismo dia (decision usuario 07-11: no pausar publicaciones); @StatLineNerd sigue en HOLD deliberado hasta 07-25 por su label de X (lib/x_scheduler.py release-hold para reanudar antes) -> proof-drip via Bluesky mientras tanto. Palanca sin cambio: rank-or-prune + proof-drip ES; no sobre-invertir en off-season. || 2026-07-06 (weekly): Good-tier ROI FIRME en 19.2% (public API all-time; la prueba publica AGUANTA fuerte -- el slice 90d es +5.9%, mas blando por off-season pero positivo). GA4 sessions 117->127 (+8.5%) via WNBA-bridge; GSC clics SIGUEN 0 (7a semana, 37/46 imp = brand query 'proplab'). Palanca SIN CAMBIO: rank-or-prune del footprint programatico + proof-drip ES en @StatLineNerd/Bluesky; no sobre-invertir hasta que vuelva la temporada NBA. || 2026-07-01 (SEO audit): 2 of 3 June HIGH findings RESOLVED -- homepage canonical now present, title fixed to 'PropLab - Análisis inteligente de player props NBA', + new FAQPage schema. Fresh GSC (28d): 0 clicks / 36 imp / pos 9.1, but 29 of 36 imp = brand query 'proplab' -> the 1,521 programmatic /players/ pages earn ~ZERO impressions (indexed but never retrieved for prop queries, 6+ wks). #1 SEO fix: RANK-OR-PRUNE the footprint -- add ES data-driven analysis prose to the ~150-300 rotation-relevant players (citable, query-matchable) AND noindex the deep-bench long tail (< N games) to concentrate index signal. #2: add SoftwareApplication/Offer + Dataset schema (only Org/WebSite/FAQPage today). Full report: products/nbaproplab/FULL-AUDIT-REPORT.md || 2026-06-29: off-season, GA4 sessions +41% WoW (83->117) on WNBA-bridge content (top referral web.telegram.org, 17). GSC impressions flat & tiny (30->29) and clicks STILL 0 across 6 weeks -> programmatic prop pages are indexed but earn zero CTR; the only impression-getter is the brand query. Good-tier ROI 19.3% -> proof HOLDS, strongly positive (slight slip from 20.1; proplab MCP out of scope this run, KPI carried from the 06-28 collector). ACTION (unchanged lever, sharper): rewrite SERP titles/meta on the top programmatic player/prop pages to break the chronic 0-clicks (the ~1,521 indexed URLs are dead weight until they earn a single click) and keep the low-cost ES proof-drip on @StatLineNerd/Bluesky (Good-tier HOOK, show misses). Do NOT over-invest paid effort until the NBA season returns. Tweet ES picks via twitter_api_post.py --account StatLineNerd --pick-id.
Routines
NBAv3 daily production watchdog — product-path health + open-issue continuity (read-only; alerts on anomaly)6 runs
ScheduleAt 04:13 PM, every dayLast15h agoNextin 8h
15h agookProduct path healthy (health 200, 6/6 containers healthy incl. telegram, 404/404 P(hit) on 08-02, settlement+odds fresh). TG-PUBLISH fix re-verified (no new 42703; 0 GOOD picks so no send due). Open: ADMIN-CH 8d, LOG-ROTATE 8d.gap_days: 0picks_latest: 404good_latest: 0
1d agookHealthy: health 200, scoring+postgres Up, 248/248 picks with P(hit) on 2026-08-01, calibration current (4686 settled). Telegram container unhealthy - noted, not escalated.picks_latest: 248phit_coverage_pct: 100settled_picks: 4686
3d agookAll checks green: API 200, postgres/api/scoring Up healthy, 364/364 picks with shadow P(hit) on 2026-07-30, calibration snapshot fresh (settled 4327, Brier 0.2584, AUC 0.4691 on parked logistic = expected)picks_latest_rundate: 364picks_with_phit: 364settled_picks: 4327
ScheduleAt 04:45 PM, every dayLast15h agoNextin 8h
15h agookDormancy tick (Sunday, non-Monday): no restore-daily condition met; appended ledger line only. trainable=8350, validatable=77/300, AUC 0.4714, gbm_vm 48.9 vs GOOD 58.0.oos_n: 11076trainable: 8350validatable: 77
1d agookPARKED-DATA-GATED: logistic AUC 0.473 (sub-random, GOOD bin worst of four), NEW Premium 45.8% vs CURRENT GOOD 58.3% (-12.5pp); GBM vm 48.2% vs 58.0%; trainable 8218/3000 MET, validatable 77/300 short; pipeline healthyoos_settled: 10944auc: 0.4729trainable: 8218
3d agookOOS unfroze post-break (10520, +435); logistic AUC new low 0.4691, NEW Premium 44.9% vs GOOD 57.8% (-12.9pp); trainable 7794 MET, validatable 70/300 short; 21 PARKED - DATA-GATEDoos_settled: 10520auc: 0.4691new_premium_hit: 44.9
5d agookPARKED-DATA-GATED: OOS 10085 flat 6th run (All-Star break tail; 5 games tonight, unfreezes 07-29). Logistic AUC 0.4766, GOOD bin 0.475 worst of four. NEW Premium 45.5% vs CUR GOOD 57.8% (-12.3pp). trainable=7359/3000 MET, validatable=65/300 short. Pipeline healthy: 508 picks scored 100% P(hit), 0 GOOD.oos_settled: 10085auc: 0.4766trainable: 7359
Daily forensic postmortem of every user-facing NBA+WNBA pick from the previous day — why each hit/missed (prediction vs reality) — written to pick-analysis/<date>/. Read-only on prod; Opus, high thinking.6 runs
ScheduleAt 03:07 PM, every dayLast16h agoNextin 7h
16h agook2026-08-01 WNBA-only 4/10 (40%); loss column spined by mode #1 (peripheral Unders on alphas Wilson/Cardoso/Ionescu); migrated RECURRING.md to v2 + promoted defect cluster to backlog 999.6; POSTed 10 narratives + recap (all 200)picks: 10hits: 4hitrate_pct: 40
1d agook2026-07-31 postmortem: 10 WNBA picks, 7/10=70%. All 3 misses were game-script/total blindness (mode #3): low game 156 sank 2 scoring Overs, +22.5 shootout 210 sank an all-around Under. Buried contrary matchups (#5) were 4/4 right. 10 per-pick files + summary + RECURRING updated; 11/11 public narratives POSTed 200 OK.picks: 10hits: 7hitrate_pct: 70
6d agook2026-07-26 = 4th consecutive WNBA All-Star break off-day (0 games, empty topPicks, 0 fetch failures); no picks to postmortem, no narratives POSTed. Wrote off-day _summary.md + RECURRING.md row. Season resumes 07-28 (already ingested/scored).picks_analyzed: 0games: 0dashboard_calls: 3
Daily NbaPropLab Discord free good-picks feed (Good-tier lead, shows misses, lint-gated). Server + #free-good-picks webhook EXIST and the sender (discord_post.py) is verified live (2026-06-23) — runs as a daily cron.37 runs
ScheduleAt 06:35 PM, every dayLast13h agoNextin 10h
13h agoskipSKIP: no Good-tier picks 2026-08-02. NBA dark (offseason, 0 games); WNBA has 4 games (IND@MIN, LAS@POR, CON@DAL, TOR@GSV) with 202 picks published, but ratingDistribution is elite 0 / good 0 / marginal 15 / weak 111 / avoid 76. Top score of the day 62.4 (Charlisse Leger-Walker P+R+A u14.5, Marginal), below the Good threshold. NOT a pipeline gap: proplab_system_status healthy, morning_wnba succeeded 2026-08-02T10:28Z and recovery_wnba 09:43Z, and 202 picks are live. Backend: Good tier 3153 settled, 62.5 pct hit, +19.3 pct ROI, +609.7u, last settled 2026-08-01. No message built, no post sent. 2nd straight no-Good day.good_picks: 0games: 4picks: 202
1d agoskipSKIP: no Good-tier picks 2026-08-01. NBA dark (offseason, 0 games); WNBA has 2 games (LVA@CHI, NYL@PHX) with 124 picks published, but ratingDistribution is elite 0 / good 0 / marginal 8 / weak 87 / avoid 29. Top score of the day 61.7 (Alyssa Thomas PTS o14.5, Marginal), below the Good threshold. NOT a pipeline gap: proplab_system_status healthy, morning_wnba succeeded 2026-08-01T10:34Z and recovery_wnba 16:14Z (21 min before run). Backend live: Good tier 3153 settled, 62.5 pct hit, +19.3 pct ROI, +609.7u, last settled 2026-07-31 (4 of 5 Good picks hit). No message built, no post sent.good_picks: 0games: 2picks: 124
2d agoskipSKIP: no Good-tier picks 2026-07-31. NBA dark (0 games, offseason); WNBA has 3 games scheduled (SEA@ATL, DAL@WAS, IND@POR) but picksCount=0 and ratingDistribution all zeros - the daily pick pipeline had not run yet at execution time (06:25Z; morning_wnba last success 2026-07-30T10:34Z, recovery_wnba 2026-07-31T02:16Z with 0 picks published). API health: healthy. Backend live: Good tier 3146 settled, 62.5 pct hit, +19.3 pct ROI, last settled 2026-07-29. No message built, no post sent.good_picks: 0games: 3picks: 0
6d agoskipSKIP: empty slate 2026-07-27 (6th dark day: 07-21, 07-23, 07-24, 07-25, 07-26, 07-27). NBA dark (offseason) and WNBA 0 games / 0 picks, WNBA break window. Verified NOT stale instrumentation: proplab_system_status healthy with morning 10:45Z, morning_wnba 10:29Z and recovery_wnba 09:30Z all succeeding TODAY, and proplab_games returns []. Backend live: Good tier 3141 settled, 62.5 pct hit, +19.3 pct ROI, last settled date 2026-07-22 (nothing has settled in 5 days). No message built, no post sent.good_picks: 0games: 0picks: 0
7d agoskipSKIP: empty slate 2026-07-26 (5th dark day: 07-21, 07-23, 07-24, 07-25, 07-26). NBA dark (offseason) and WNBA 0 games / 0 picks, WNBA All-Star break window. Verified NOT a stale pipeline: proplab_system_status shows morning_wnba succeeded 2026-07-26T10:29Z and recovery_wnba 08:14Z, and the public /api/v1/games returns []. Backend healthy: Good tier 347 settled, 62.2% hit, +18.9% ROI, last settled date 2026-07-22. No post sent.good_picks: 0games: 0picks: 0
8d agoskipSKIP: empty slate 2026-07-25 (3rd straight day). NBA dark (offseason) and WNBA 0 games / 0 picks - WNBA All-Star break window. Backend healthy: track_record live, Good tier 62.2% hit / 18.9% ROI / 347 settled through 2026-07-22. No post sent.good_picks: 0games: 0product: nbaproplab
9d agoskipNo Good-tier slate 2026-07-24: NBA dark (0 games) and WNBA empty (0 games/0 picks). SKIP per empty-slate rule, no post.good_picks: 0product: nbaproplab
10d agookSKIP: empty slate 2026-07-23 (NBA and WNBA dashboards both 0 games / 0 picks; WNBA All-Star break window). No post sent, expected summer behavior per SKILL.md.good_picks: 0skipped: 1product: nbaproplab
11d agookgood_picks: 1
12d agoskippedSKIP: empty slate entirely on 2026-07-21 - proplab_dashboard returns gamesCount:0 / picksCount:0 for BOTH nba and wnba (and unfiltered), proplab_games returns []. Not a zero-Good-tier day: there is no slate at all (NBA offseason + WNBA no games scheduled/published, likely All-Star break). Nothing posted. Model health unchanged: Good tier 3137 picks 62.5% hit / +19.4% ROI through 07-20.good_picks: 0games: 0picks: 0
13d agookgood_picks: 1
14d agoskipNo Good-tier picks on 2026-07-19 WNBA slate (3 games; dist good:0 elite:0 marginal:12 weak:128); no NBA (offseason). Skipped per Good-lead contract.good_picks: 0games: 3product: nbaproplab
2026-07-31 (weekly): STILL PARKED (sunset). SEGUNDA lectura negativa seguida -- GSC imp 275->261 (-5%), clics 11->8 (-27%), GA4 51->41 (-20%); waitlist SIGUE 0/0, 1 star, 3 clones/14d. Cambia el reparto interno: /blog/vertical-slice-architecture-dotnet pasa a ser la pagina top (3 clics/91 imp). El decaimiento sin refresh es EXACTAMENTE lo esperado en un dominio aparcado, NO una senal que exija accion. DECISION SIN CAMBIO: parked, sin build, sin re-pitch, cero gasto; el valor ya se trasvaso a devaisemanal (13 clics, #2 de ese sitio). || 2026-07-27 (weekly): STILL PARKED (sunset). PRIMERA semana negativa en todo: GSC imp 305->275 (-10%) y clics 15->11 (-27%) rompen 5 semanas de compounding pasivo; GA4 77->51 (-34%); clones 14d 12->2. Waitlist SIGUE 0, 1 star. El decaimiento sin refresh es lo ESPERADO en un dominio aparcado, no una senal nueva. DECISION SIN CAMBIO: cero gasto en superficies Sharpyard. El trasvase a devaisemanal ya esta hecho y rinde (claude-code-dotnet-csharp-guia = #2 pagina de DevAI con 12 clics). || 2026-07-20 (weekly): STILL PARKED (sunset). GSC compounding pasivo imp 275->305 (+11%), clics 14->15 (/blog/claude-code-for-dotnet lidera) pero GA4 93->77 (-17%) y waitlist SIGUE 0 (1 star, 9 cloners) = cero senal embudo/credibilidad. Decision sin cambio: cero gasto en superficies Sharpyard; cosechar el trafico .NET+Claude hacia el motor devaisemanal (ya rinde). || 2026-07-13 (weekly): STILL PARKED. GSC compounding pasivo (imp 171->275 +61%, clics 12->14; /blog/claude-code-for-dotnet 7 clics/130 imp lidera) pero GA4 120->93 (-23%) y waitlist SIGUE 0 (1 star, 9 cloners) = cero senal de embudo/credibilidad. Decision sin cambio: cero gasto en superficies Sharpyard; cosechar el trafico .NET+Claude hacia el motor devaisemanal (ya rinde 7 clics ES). || 2026-07-11 (weekly): STILL PARKED. GSC keeps compounding passively (imp 171->232 +36%, clicks 12->14; /blog/claude-code-for-dotnet 7 clicks/128 imp leads) while GA4 slid 120->103 (-14%); waitlist STILL 0, 1 star, 8 cloners -> no funnel/credibility signal. The harvest into devaisemanal ALREADY pays: the ES article claude-code-dotnet-csharp-guia earns 7 clicks/78 imp in DevAI's GSC. Zero spend on Sharpyard surfaces; decision unchanged. || 2026-07-06 (weekly): STILL PARKED, decision unchanged. WoW: GSC impressions 86->171 (+99%), clicks 5->12 (+140%) -- the .NET+Claude content keeps compounding in SEO (/blog/claude-code-for-dotnet 5 clicks/106 imp) -- but GA4 sessions 158->120 (-24%) and waitlist confirmed STILL 0 (1 star, 10 cloners) = no funnel/credibility signal. ONE ACTION unchanged: harvest the .NET+Claude traffic into the devaisemanal engine; no spend on Sharpyard surfaces. || 2026-06-29 STILL PARKED (no re-pitch). WoW: GSC impressions 36->86 (+139%), clicks 1->5, and the lead-magnet repo got its FIRST star (0->1) + 5 key events fired -- but waitlist confirmed is STILL 0 after 17 unique cloners, so the credibility/funnel signal is still absent (1 star != traction). Not enough to un-park. Only earner is /blog/claude-code-for-dotnet (4 clicks, 71 imp). DECISION UNCHANGED: domain parked, NO build, no new domain, no re-pitch. ONE ACTION: harvest the residual .NET+Claude content traffic into the devaisemanal GEO/content engine (cross-link / repurpose the claude-code-for-dotnet angle there) and stop spending on Sharpyard surfaces. Theses-dead reports: docs/research/2026-06-24-sharpyard-*.
Routines
Weekly Sharpyard SEO/indexation check: confirms new blog posts get into Google, pings sitemap, auto-requests indexing for not-yet-indexed posts via GSC2 runs
ScheduleAt 10:04 AM, only on TuesdayLast2d agoNextin 1d
5d agook16/18 posts indexados; sitemap resubmitido (204); solicitada indexacion en GSC para los 2 EN 'Crawled - not indexed' (subagents-skills, dotnet-10-feature); 7d: 1 clic / 21 impresiones (vs 2/34)posts_total: 18indexed: 16not_indexed: 2
Weekly Sharpyard funnel review + single highest-impact next action (Reddit excluded as a channel)2 runs
ScheduleAt 09:53 AM, only on MondayLast2d agoNextin 1h
2d agookSTILL PARKED (sunset): 2a semana negativa consecutiva. GSC 275->261 imp (-5%), 11->8 clics (-27%); GA4 51->41 sesiones (-20%), sign_up=0; waitlist 0/150-250 (listas 4/5 literalmente vacias); 1 star / 30-50. Unico dato nuevo: 3 sesiones desde chatgpt.com (GEO pasivo) y vertical-slice-architecture-dotnet adelanta a claude-code-for-dotnet como top earner GSC. Decision sin cambio: cero gasto en superficies Sharpyard.gsc_impressions: 261gsc_clicks: 8ga4_sessions: 41
6d agookSharpyard PARKED: waitlist 0/150-250 y 1/30-50 stars sin cambio; segunda semana de decaimiento (GSC 305->275 imp, 15->11 clics; GA4 77->51; clones 14d 2). Accion unica: disparar los 8 crossposts Dev.to/Hashnode ya escritos (sin publicar desde 06-25/07-23) como ultimo test de coste cero.waitlist: 0stars: 1gsc_impressions: 275
2026-07-31 (audit mensual SEO): PRIMERA auditoria completa. Tecnicamente sano (robots abierto, GPTBot/ClaudeBot/PerplexityBot 200, www+http 301 al apex, SoftwareApplication+Offer en la home, 7/7 img con alt, TTFB 0.12s) y el mes revierte la tendencia: imp 1891->1924, clics 7->14 (+100%), pos media 38.1->30.3 (+7.8 puestos). ACCION #1 (SEO): consolidar y refrescar la pagina que YA convierte -- /blog/tpv-obligatorio-hosteleria-2026 tiene 582 imp y 7 de los 14 clics pero se atasca en pos 13.2 con tres problemas que se suman: (a) el titulo/URL dicen 2026 mientras el propio cuerpo y /blog/verifactu-aplazado-2027-restaurantes fechan los plazos en enero y julio de 2027; (b) canibalizacion con /blog/es-obligatorio-tener-tpv-bar (pos 9.0, solo 32 imp) que parte la query 'es obligatorio tener tpv en un bar' (66 imp, pos 19.1, 0 clics); (c) sin schema FAQPage pese a que casi todas las queries top son preguntas literales -> no se captura PAA ni cita de motores IA. Indicador: esa URL pasa de pos 13 a top-10 y de 7 a 25+ clics/28d. HALLAZGOS 2 y 3: 859 imp en paginas comerciales con CERO clics (/precios 413 imp pos 69.7, /tpv-franquicias-hosteleria 241 imp pos 15.7 -- esta ultima es puro problema de titulo/meta, no de ranking, y /tpv-pizzerias 205 imp pos 39.5); y NO hay llms.txt (404, geo:false) pese a que VeriFactu es una pregunta de busqueda-IA de manual en ES. Menor: /verifactu-bares solo lleva BreadcrumbList y saco 0 imp; el colector umami del hub NO lee eventos (el fix del sitio SI aterrizo: id fa7e47d9 + data-umami-event=whatsapp_click en el boton flotante), asi que el punto ciego de CTA es ahora del hub, no del producto; la service-account de Google da 403 en sc-domain:tpv-top.com (solo tira el token OAuth); tpvtop.com sin guion NO es nuestro y redirige a un tercero. Informe: products/tpvtop/FULL-AUDIT-REPORT.md. La prioridad de NEGOCIO sigue siendo la del weekly de hoy (cerrar go-lives via gestorias/AEHC + instrumentar el CTA); lo de arriba es la palanca SEO. || 2026-07-31 (weekly): MEJOR LECTURA DEL HUB -- GSC imp 1850->1924 (+4%, 3a subida seguida tras 5 semanas de caida), clics 12->14 (+17%, pos 31.2->30.3) y Umami SE DISPARA 32->59 visitors (+84%, 65 pageviews). El angulo VeriFactu ya arrastra una 2a pagina: tpv-obligatorio-hosteleria-2026 concentra 7 de los 14 clics (582 imp, pos 13.2) y ENTRA como-cambiar-de-tpv-sin-perder-datos con 2 clics en pos 7.1. PERO 0 eventos CTA en 28d: el trafico record no se atribuye a NI UN clic de WhatsApp/demo. ACCION: (1) SUBE A PRIORIDAD 1 -- arreglar el registro de eventos CTA (demo_click/email_click/WhatsApp) en ../TPVTop; con trafico ya real, seguir ciego es el desperdicio caro; (2) sin cambio: cerrar instalaciones via gestorias + AEHC, el CTR organico no es la palanca de este B2B. || 2026-07-27 (weekly): estabilizado -- GSC imp 1814->1850 (+2%, 2a semana al alza), clics 13->12, pos 31.2. /blog/tpv-obligatorio-hosteleria-2026 concentra 8 de los 12 clics (533 imp, pos 13.2): el angulo VeriFactu es EL unico convertidor. Umami 22->32 visitors (+45%) pero CERO eventos CTA en 28d -> con trafico ya real, el embudo sigue ciego. ACCION: (1) sin cambio de prioridad, cerrar instalaciones via gestorias + AEHC (el CTR organico no es la palanca de este B2B); (2) NUEVO: arreglar el registro de eventos CTA (demo_click/email_click/WhatsApp) en ../TPVTop -- ahora que llega trafico, seguir sin proxy de conversion es el desperdicio. || 2026-07-20 (weekly): GSC PRIMER repunte tras 5 semanas cayendo -- imp 1610->1814 (+13%), clics 11->13 (+18%, pos 33.3; tpv-obligatorio-hosteleria EL convertidor). Umami: el id publico fa7e47d9 ahora lee 22 visitors (era 6) = el fix data-website-id cosecha; 0 eventos CTA aun. KPI go-lives off-platform (sin dato). ACCION sin cambio: cerrar instalaciones via gestorias + AEHC; el CTR organico NO es la palanca de este B2B. || 2026-07-13 (weekly): LAGGARD -- GSC 5a semana cayendo (imp 1814->1610 -11%; clics 7->11 mejor semana, tpv-obligatorio-hosteleria EL convertidor). Umami: el colector lee el id publico fa7e47d9 y da solo 6 visitors/0 eventos CTA -> VERIFICAR que el fix data-website-id (fa7e47d9-bb74-45ee-a05b-c4e76eec5256) aterrizo en ../TPVTop; sigue casi ciego. KPI go-lives off-platform (sin dato). ACCION sin cambio: cerrar instalaciones via gestorias + AEHC; refresh cluster VeriFactu-2026 para frenar la caida. || 2026-07-11 (weekly): UMAMI MAL CABLEADO (hallazgo de hoy): la landing tpv-top.com envia eventos al website-id ADMIN (a17a69b2, dominio tpvtop-cloud-admin.vercel.app), no a tpvtop-public (fa7e47d9) que resuelve el colector por dominio -> el dashboard estaba CIEGO. Numeros reales (id admin, 28d): 56 visitors/67 visits (28d previos 81/119 = cayendo) y CERO eventos demo_click/email_click -> el proxy de clics WhatsApp/CTA no registra NADA en 28d. FIX en ../TPVTop: data-website-id -> fa7e47d9-bb74-45ee-a05b-c4e76eec5256 (layout Next.js). GSC: imp 1814->1669 (-8%, 4a semana cayendo) pero clics 7->11 (+57%, mejor semana = CTR sube; tpv-obligatorio-hosteleria-2026 6 clics/325 imp pos 10.7 sigue siendo EL convertidor). ACCION sin cambio: cerrar instalaciones via gestorias + asociaciones (AEHC); refresh del cluster VeriFactu-2026. || 2026-07-06 (weekly): SEO en TERCERA semana negativa consecutiva -- GSC impresiones 2080->1814 (-13% WoW), clics 6->7. Unico convertidor sigue /blog/tpv-obligatorio-hosteleria-2026 (274 imp, 4 clics), angulo VeriFactu. El CTR organico NO es la palanca de este B2B; KPI go-lives = off-platform, pull manual (sin dato esta semana). ACCION SIN CAMBIO: cerrar instalaciones via gestorias + asociaciones hosteleria (AEHC Cantabria); la caida sostenida justifica un refresh del cluster VeriFactu-2026, pero no es el embudo. || 2026-06-29: KPI = go-lives (off-platform, NO se mide en analytics; pull manual via WhatsApp/gestorias - sin dato automatizado esta semana). SEO en 2a semana NEGATIVA consecutiva: GSC impresiones 2630->2080 (-21% WoW, sigue cayendo desde la meseta), clics 7->6. El unico convertidor es /blog/tpv-obligatorio-hosteleria-2026 (273 imp, 2 clics) -> el angulo VeriFactu-obligatorio. PERO el CTR organico NO es la palanca de este negocio B2B. ACCION (sin cambio de prioridad): cerrar instalaciones via gestorias + asociaciones de hosteleria (AEHC Cantabria), canales de alta prioridad; refrescar/ampliar el cluster de contenido VeriFactu-2026 para frenar la caida de impresiones. NO activar Google Ads hasta limpiar claims rojos.
Routines
No product-specific routines — covered by shared hub automation.
2026-07-31 (weekly): SIN CAMBIO estructural -- GSC imp 235->237 plano, los 3 clics organicos SE MANTIENEN (4a lectura seguida) y la pos media sigue mejorando 13.74->13.36. TODO el clic sale de requerimiento-aepd-10-dias (3 clics/87 imp, pos 3.8): el angulo de urgencia AEPD es el que engancha, ya confirmado. GA4 0 sesiones (congelado POR DISENO), Listmonk 1 unconfirmed. KPI LOI/depositos 0 porque la puerta NO se ha abierto -- no es fallo de marketing. ACCION humana sin cambio: 15 entrevistas diagnostico + preventa fria exigiendo 8-10 LOI ANTES de construir (../reg-desk/KIT-VALIDACION.md). || 2026-07-27 (weekly): SIN CAMBIO estructural -- GSC imp 209->235 (+12%) y los 3 clics organicos SE MANTIENEN (pos media 17.5->13.7). TODO el clic sale de una sola pagina: requerimiento-aepd-10-dias (3 clics/87 imp, pos 3.8) = el angulo de urgencia AEPD es el que engancha; empiezan a impresionar queries EN de AI Act. GA4 0 sesiones (congelado POR DISENO), Listmonk 1 unconfirmed. KPI LOI/depositos = 0 porque la puerta NO se ha abierto: no es un fallo de marketing. ACCION humana identica: 15 entrevistas diagnostico + preventa fria exigiendo 8-10 LOI ANTES de construir. Kit: ../reg-desk/KIT-VALIDACION.md. || 2026-07-20 (weekly): SIN CAMBIO estructural -- GSC imp 124->209 (+69%, dominio joven compounding) y PRIMEROS clics organicos 0->3 (pos 17.5; paginas AEPD requerimiento-aepd-10-dias pos 3.8, necesita-dpo-art-34 pos 7.4 prometen). GA4 lectura anomala 0 (el 07-19 marcaba 14; trafico congelado por diseno). Listmonk 1 unconfirmed. KPI LOI/depositos = 0 porque la puerta NO se ha abierto. ACCION humana identica: 15 entrevistas diagnostico + preventa fria exigiendo 8-10 LOI ANTES de construir. || 2026-07-13 (weekly): SIN CAMBIO -- GA4 plano en 14 (por diseno; engagement 7% = trafico frio), GSC imp 37->124 (+235%, dominio joven compounding; requerimiento-aepd-10-dias pos 3.8 y necesita-dpo-art-34 pos 7.4 = las paginas AEPD prometen), clics 0. Listmonk 1 unconfirmed. LOI/depositos = 0 porque la puerta NO se ha abierto. ACCION humana identica: 15 entrevistas diagnostico + preventa fria exigiendo 8-10 LOI ANTES de construir. || 2026-07-11 (weekly): SIN CAMBIO -- GA4 plano en 14 (por diseno; engagement 7% = trafico frio), GSC imp 37->107 (+189%, dominio joven compounding; requerimiento-aepd-10-dias pos 3.75 y necesita-dpo-art-34 pos 7.5 = las paginas AEPD prometen), clics 0. Listmonk 1 unconfirmed. LOI/depositos = 0 porque la puerta NO se ha abierto. ACCION humana identica: 15 entrevistas diagnostico + preventa fria exigiendo 8-10 LOI ANTES de construir. || 2026-07-06 (weekly): SIN CAMBIO -- analitica plana POR DISENO (GA4 14 sesiones), GSC imp 23->37 (+61%, dominio joven sirviendo /en/blog/data-breach-72-hours-dpo). LOI/depositos = 0 porque la puerta NO se ha abierto. ACCION humana/off-analytics: 15 entrevistas diagnostico a DPOs + preventa en frio exigiendo 8-10 LOI ANTES de construir. || 2026-06-25: PUERTA DE VALIDACION INVERTIDA (Paso A), sin cambio -- la analitica es plana POR DISENO y NO es la senal de validacion (GA4 sessions planas en 14 desde el deploy 06-20; GSC empieza a servir, 0->2 impresiones, /en/blog/data-breach-72-hours-dpo pos 8.5). LOI/depositos = 0 porque la puerta aun NO se ha abierto. ACCION (humana, fuera de analytics): 15 entrevistas diagnostico a DPOs + pre-venta del producto de pago a frio (LinkedIn + registro AEPD), exigir 8-10 depositos/LOI ANTES de construir el pipeline. NO medir altas gratis como exito (red-team: la AEPD regala el boletin). Kit: ../reg-desk/KIT-VALIDACION.md. Plan: ../reg-desk/BUSINESS-PLAN.md.
Routines
No product-specific routines — covered by shared hub automation.