- Updated: March 26, 2026
- 13 min read
Claude’s Code Adoption Soars Since Launch – Key Metrics Revealed
Interesting ObservationsA quick read on momentum, adoption, and where the current activity is clustering.Growth+8%W/WAcceleration-17.9ppDoubling Time61 daysEarly Adopter CommitsThese are the earliest observed public-era Claude Code commits we can verify after launch. Multiple same-day candidates may exist, so this is suggestive rather than definitive.Feb 24, 2025Earliest observed launch-day candidates for Claude Code adoptionmoinmir/ClashOfCansChange initial game setup to always have exactly 1 correct can This improves the starting condition by ensuring players always begin with exactly one can in the correct position, making the initial game state more consistent.🤖 Generated with Claude Code Co-Authored-By: Claude 1 of 5 candidatesRotates every 5sNew Repos This Week114,785Original repos (non-forks) with their first observed Claude Code commit in the last 7 daysNet Code Delta+30.7BLines added minus lines deleted for the current sliceTop 3 Languages01TypeScript34.8%7.25M commits02Python18.9%3.92M commits03JavaScript10.2%2.13M commitsTotal Commits20,807,124Global activityLines Added+50,442,790,274New code generatedLines Deleted-19,767,154,942Refactored codeActive Repos1,087,408Contributing projectsViewActivity Over TimeSystem Activity LogSince Launch · All · Mar 25, 9:12 PMdaf269a03/25/2026, 19:59:4603/25, 19:59standardbeagle/agntclaim: Mdkq6bztAPFj by beagle-ab2-1591895 Co-Authored-By: Claude Opus 4.6 (1M context) 1dddc1503/25/2026, 19:59:2503/25, 19:59bencrane/data-engine-x-apidocs: update all documentation to reflect completed Global Entity Model project Co-Authored-By: Claude Opus 4.6 998df9703/25/2026, 19:58:5703/25, 19:58it-mannetje/hogenesAuth: magic link login, allowed_admins whitelist, server-side RLSc4039a203/25/2026, 19:58:5603/25, 19:58it-mannetje/hogenesAuth: magic link login, allowed_admins whitelist, server-side RLS106a6e803/25/2026, 19:58:5503/25, 19:58it-mannetje/hogenesAuth: magic link login, allowed_admins whitelist, server-side RLScc0e4e603/25/2026, 19:58:3503/25, 19:58zseramnay/Formants9 principes orchestration, doublures par registre, commentaires registre 16 instruments, encadré profil vs carte376890303/25/2026, 19:57:0103/25, 19:57thefiredev-cloud/Protocol-Guidefix: enable adaptive threshold + add 22 bridge chunks for top failure protocols – Enable adaptive threshold retry for ALL agency searches (not just high-accuracy) – Add dense bridge chunks for top 20 failing protocols covering ~130 test failures – Protocols: 1203 1209 1210 1211 1212 1213 1215 1216 1219 1220 1222 1223 1225 1229 1232 1237 1241 1244 1302 510 518 519226418103/25/2026, 19:56:3903/25, 19:56gvinsot/PulsarTeamfeat: add project statistics with charts (bugs created vs resolved, resolution time evolution) – Backend: add getTimeSeries() to globalTaskStore for day-by-day created/resolved counts, resolution time evolution, and open tickets over time – Backend: add GET /api/agents/tasks/stats/timeseries?project=X&days=N endpoint – Frontend: new ProjectStats component with Chart.js graphs (created vs resolved bar chart, resolution time line chart, open tickets area chart, state durations) – Frontend: integrate ProjectStats into ProjectsView – clicking any project card opens the statistics panel with all charts – Frontend: add getProjectTaskStats and getProjectTimeSeries API methods (by CLAUDIO)8fbe3ab03/25/2026, 19:56:1603/25, 19:56ydesjardins200-coder/wavesfinancialfix: remove stray closing div breaking Support mega menu on apply pagee2082f903/25/2026, 19:56:0903/25, 19:56kuzmich89/AI-brokerUse NHTSA instead of AutoAstat for VIN data loading in CarCaseDetail Replace the removed fetchAutoAstat button with a new ‘Завантажити з NHTSA’ button that calls vinApi.decode(vin) and updates the case fields directly. Shows warning toast when mock data is returned (VIN not in NHTSA DB). https://claude.ai/code/session_01XNhJmFU2Jzmc5B1FZTwGWK950ddb403/25/2026, 19:55:5403/25, 19:55Dicklesworthstone/frankenscipyfeat(all): add DBSCAN clustering, FFT plan caching, RBF/Akima interpolation, BFGS optimizer, and expand constants/special/stats Cluster (+126 lines): DBSCAN density-based clustering with configurable eps and min_samples, returning cluster labels with -1 for noise points.Constants (+108 lines): Comprehensive physical constants module (speed of light, Planck, Boltzmann, Avogadro, electron mass, proton mass, elementary charge, gravitational constant, etc.) matching scipy.constants surface. FFT (+94 lines): Plan caching for repeated FFT sizes, reducing planning overhead for iterative algorithms. Adds real-to-complex (rfft/irfft) optimized paths.Interpolate (+216 lines): RBF (radial basis function) interpolation with multiquadric/inverse_multiquadric/gaussian/linear/cubic kernels. Akima interpolation (subspline with reduced overshoot). RegularGridInterpolator for N-dimensional interpolation on regular grids. Linalg (+32/-32): Refactors matrix operation signatures for consistency, replacing ad-hoc parameter ordering with standardized (matrix, n, .) form. Ndimage (+8/-8): Minor cleanup of filter axis validation messages.Optimize (+128 lines): BFGS quasi-Newton optimizer with Wolfe line search, inverse Hessian approximation, and gradient convergence detection. Signal (+12/-12): Consistency fixes for window function parameter validation. Special: Simplifies Airy function implementation (+48/-48), fixes convenience function parameter passing, updates re-exports. Stats (+6/-6): Minor fixes to distribution parameter edge cases. Integrate (+4/-4): Quadrature tolerance consistency fixes. Co-Authored-By: Claude Opus 4.6 (1M context) db513f603/25/2026, 19:55:5003/25, 19:55prompted365/context-grapple-gunUpdate CogPR-57 doctrine: mark all three defects as fixed (tic 108)354f8ab03/25/2026, 19:55:1503/25, 19:55prompted365/context-grapple-gunCogPR-57 mandate lifecycle fix: race guard, concurrency guard, reconcile-first Three structural defect fixes (authorized at tic 107 review, not tech debt): 1. cgg-gate.sh: Re-validate mandate status before inline lightweight consumption — prevents double-consumption race with mogul-runner 2. review SKILL.md: Concurrency guard at steps 5.5 and 8.5 — check current.json status before writing mandates or spawning Mogul. Running/pending mandates are not overwritten. 3. session-restore.sh: Reconcile-first cycle computation — read previous mandate tic_context as primary schedule source, modulo as fallback only when no previous context exists.Estate snapshot can add cycles but not replace schedule. Eliminates recomputation drift. Runtime parity: both hooks synced to ~/.claude/hooks/ and verified.c92551703/25/2026, 19:54:0503/25, 19:54kuzmich89/AI-brokerDisable AutoAstat button in CarCaseDetail Remove the ‘Завантажити дані’ button that called fetch-autoastat from the case detail page. AutoAstat is not used for now. https://claude.ai/code/session_01XNhJmFU2Jzmc5B1FZTwGWKe9ac56f03/25/2026, 19:53:4703/25, 19:53ydesjardins200-coder/wavesfinancialfeat: rebuild apply funnel for conversion — repayment preview, decline screen, cleaner UX6992f8903/25/2026, 19:53:2003/25, 19:53A13XMC11/lanlabs-agentefeat: add Redis-backed multi-message buffering with whatsapp-agentkit skill – New agent/buffer.py: Manages message buffering with Redis – Groups multiple messages into coherent context – Configurable timeout (default 2.5s) – Automatic deduplication (webhook retries) – Backpressure handling (max 15 messages/buffer) – Age-based topic separation (5 min max) – Structured JSON logging – Updated agent/main.py: – Integrates buffer_manager in webhook handler – Single unified response instead of per-message – Connects to Redis on startup – Graceful degradation if Redis unavailable – Improved structured logging – Dependencies: Added redis + python-json-logger – Config: New REDIS_URL, BUFFER_TIMEOUT_MS, MAX_BUFFER_AGE_MS env vars Implements whatsapp-agentkit skill (RED-GREEN-REFACTOR tested)cc1f00c03/25/2026, 19:53:1203/25, 19:53james-axis/project-originfix: Correct untitledui icon imports – MapPin01→MarkerPin01, Search→SearchLg, RefreshCw→RefreshCw01, FileText01→File018188fd503/25/2026, 19:51:4203/25, 19:51it-mannetje/hogenesStamboek: generation management, person editing, collapsible SVG tree, lightbox, photo management3e6808703/25/2026, 19:51:4103/25, 19:51it-mannetje/hogenesStamboek: generation management, person editing, collapsible SVG tree, lightbox, photo management3d21ba903/25/2026, 19:51:4003/25, 19:51it-mannetje/hogenesStamboek: generation management, person editing, collapsible SVG tree, lightbox, photo managementc9a991603/25/2026, 19:51:4003/25, 19:51it-mannetje/hogenesStamboek: generation management, person editing, collapsible SVG tree, lightbox, photo management8553e2403/25/2026, 19:51:3903/25, 19:51it-mannetje/hogenesStamboek: generation management, person editing, collapsible SVG tree, lightbox, photo managementf7890b203/25/2026, 19:51:1403/25, 19:51kuzmich89/AI-brokerFix NHTSA VIN loading and add Consignor/Transport/PreviousDocs sections – Add dedicated GET /api/vin/decode/{vin} endpoint that calls NHTSA directly without requiring case creation first, eliminating the two-step create-then-fetch flow that caused the loading error – Update NewCarCase.tsx: use vinApi.decode() for VIN lookup (cleaner, more specific error messages, shows brand/model/year on success) – Add Consignor section: sender name, country, address, EORI number – Add Transport section: type (sea/road/rail/air), vessel/vehicle ID, flag country, crossing point, expected date, document type & number – Add PreviousDocs (Graph 44) section: dynamic list of previous customs documents (T1/T2/TIR/EUR-1/CMR/MRN) with type, number, date, issuer – Add backend model fields + migration 014 for all new columns – Update CarCaseCreate/Update/Response schemas with new fields – Add vinApi to frontend api.ts and new fields to types/index.ts https://claude.ai/code/session_01XNhJmFU2Jzmc5B1FZTwGWK74220f103/25/2026, 19:50:4103/25, 19:50thefiredev-cloud/Protocol-Guidefix: prevent CDN caching of empty search results Empty results were getting cached at CDN (Fastly) for 1 hour via Cache-Control: public, max-age=3600. After deploying threshold/dictionary fixes, old NO RESULTS responses kept being served from CDN cache.Now: empty results get Cache-Control: no-storee0aa36a03/25/2026, 19:50:0403/25, 19:50ZhimingMei/finance-paper-reader-skillAdd README.md for repo and quick reference header to SKILL.md61ac80903/25/2026, 19:48:4503/25, 19:48Endymion1236/centre-equestre-agonfix: forfait annuel ne crée plus de payment doublon + badge impayés corrigé Bug — Forfait 1x crée un payment doublon : Avant : l’inscription forfait créait les échéances (payments) PUIS onEnroll créait un DEUXIÈME payment pour le même créneau → 2x ‘galop d’or’ à 25€ dans les impayés = 50€ parasite Après : onEnroll est appelé avec skipPayment:true pour les forfaits annuels → seules les échéances sont créées.Badge Impayés — exclut les échéances : Avant : le badge rouge affichait 11 (10 échéances + 1 commande) Après : le badge n’affiche que les commandes normales (les échéances ont leur propre onglet) Build compilé OK.651a5b803/25/2026, 19:48:2203/25, 19:48ayeshakhalid192007-dev/humanoid-ai-studiofix(backend): add Dockerfile with python:3.11-slim to reduce image size Replaces Nixpacks build (5.7GB) with explicit Dockerfile using slim base image + CPU-only PyTorch to stay under Railway’s 4GB trial limit. Co-Authored-By: Claude Sonnet 4.6 7dfe42803/25/2026, 19:47:3703/25, 19:47ZhimingMei/finance-paper-reader-skillInitial commit: finance-paper-reader skill02fec5703/25/2026, 19:44:5603/25, 19:44EmpoweredVote/Civic-Trivia-Championshipsfix: add missing New York State and Queens NY locale config files These were registered in generate-locale-questions.ts in a prior commit but the actual config files were never staged, causing TS2307 module not found errors on Render. Co-Authored-By: Claude Sonnet 4.6 78f9c0e03/25/2026, 19:43:0403/25, 19:43thefiredev-cloud/Protocol-Guidefix: lower similarity and quality thresholds for better recall 254/506 failures were all NO RESULTS (zero wrong-protocol). Precision is solid — recall is the bottleneck. – similarity thresholds: 0.58-0.63 → 0.45-0.50 – quality minimumThreshold: 0.50 → 0.35 – quality warnThreshold: 0.60 → 0.50 – minQualityResults: 2 → 12b6a99603/25/2026, 19:41:1203/25, 19:41EmpoweredVote/Civic-Trivia-Championshipsfix: replace smart apostrophes in PA/Philly collection descriptions Curly quotes in single-quoted TypeScript strings caused unterminated string literal errors in the build introduced by the Pennsylvania collections commit. Co-Authored-By: Claude Sonnet 4.6 655cf3603/25/2026, 19:40:0503/25, 19:40grimmolf/grimm-workshopAdd blog post: We’re measuring productivity wrong Capacity vs throughput framework for employee value and AI deployment. Includes Red Hat TAM program and Open Organization as case study.b9afde803/25/2026, 19:39:2903/25, 19:39james-axis/project-originfix: Fix Softphone NodeJS.Timeout type issue044ac5a03/25/2026, 19:38:4103/25, 19:38mdegans/agora-agentsagents: Add moderator agents, reassign lfm2, expand communities – Add 5 hand-crafted moderator agents (sentinel-mod, arbiter-mod, watchdog-mod, guardian-mod, inquisitor-mod) across cogito:14b and gpt-oss:20b for moderation pipeline testing – Reassign 111 lfm2:24b agents to working models (gpt-oss, cogito, gemma3, mistral-small, gemma3n) — lfm2 had near-total parse failure – Expand epidemius to 9 communities for broader moderation testing – Add community memberships for ~92 agents across 24 new communities – Include natural soul mutations from seed run (31 mutations at 3% rate) – Include 2 forced epidemius mutations from think:false testing Co-Authored-By: Claude Opus 4.6 (1M context) 404a5e303/25/2026, 19:38:2703/25, 19:38mdegans/agora-agentsserver: Add mutation-chance, agent-filter flags; fix thinking models; log Ollama stats – Add –mutation-chance CLI flag to override deep mutation probability (0-100) – Add –agent-filter CLI flag to run a subset of agents by name – Disable thinking mode (think: false) in Ollama requests to prevent reasoning models from refusing soul mutations – Log Ollama inference stats (prompt tokens, eval tokens, tok/s) with WARN threshold at 32k tokens for KV cache swap detection – Improve mutation failure logging (warn level with response preview) – Update system prompt: prefer comments over posts, list valid community slugs, add new communities Co-Authored-By: Claude Opus 4.6 (1M context) 68bc32603/25/2026, 19:38:2203/25, 19:38james-axis/project-originfix: Add practice_id support to call-flows routesb0230a303/25/2026, 19:37:4803/25, 19:37james-axis/project-originfeat: Add Phone System wizard with Twilio subaccount management – Add Stage 0: Practice/Adviser subaccount management routes – Each practice gets its own Twilio subaccount for isolated billing – Add 6-step setup wizard: Practice → Address → Compliance → Voice → Number → Routing – Full AU regulatory compliance flow (address, bundle, approval) – TwiML App creation per practice – Phone number search and purchase – Call flow/IVR configuration – Updated schema with twilio_practices and twilio_twiml_apps tables – Comprehensive frontend wizard UI with stepper – Real-time API integration with backend792fccb03/25/2026, 19:37:4003/25, 19:37snobol4ever/.githubPlanning session 3 archive — Icon gap analysis, GMR scoped340d1c003/25/2026, 19:36:5303/25, 19:36michalbaturko-lang/webflipfix(i18n): translate Czech strings in CoreWebVitals component Co-Authored-By: Claude Opus 4.6 94028ee03/25/2026, 19:36:1903/25, 19:36thefiredev-cloud/Protocol-Guidefix: regenerate pnpm-lock.yaml for Railway deployf7d46c803/25/2026, 19:36:0603/25, 19:36ydesjardins200-coder/wavesfinancialReplace nav with mega menu across all 19 pages — Products / Company / Support dropdowns, mobile drawer, hover+click, keyboard accessibledceecbc03/25/2026, 19:35:4603/25, 19:35ErikEvenson/galaxyfeat: ZEM/ZEV powered descent guidance — deorbit→ZEM/ZEV flow (#981) Architecture: plane_change → coast_to_deorbit → deorbit → powered_descent The deorbit burn puts the ship on a suborbital trajectory, then ZEM/ZEV guidance handles the powered descent to the pad. Changes: – zemzev_guidance_math.py: correct formula (no double-counting gravity) – zemzev_guidance.py: tick handler with safety override – maneuver_landing.py: deorbit transitions to powered_descent for targeted landings; coast transitions restored to deorbit phase – Harness: place_ship_above_target positions at 10km with 50 m/s descent Test results (117 total): – 115 passed, 2 failed (Io precision — harness propagation issue) – Luna precision at 10m: PASSES – All existing landing/unit tests: PASS (no regression) – 19 ZEM/ZEV math unit tests: PASS Remaining: Io precision tests need investigation — the harness Keplerian body propagation (Io around Jupiter at 17 km/s) introduces drift that doesn’t exist in the real N-body game. The ZEM/ZEV guidance itself is correct; the test harness needs better Io frame handling. Co-Authored-By: Claude Opus 4.6 (1M context) 06ea6bf03/25/2026, 19:34:1303/25, 19:34joshdataresources/climate-studioUpdate globals.css5e79a8f03/25/2026, 19:33:4703/25, 19:33prompted365/context-grapple-gunTic 107 review: promote CogPR-73/74/76/77/78 to CGG CLAUDE.md – CogPR-73: Governance label accuracy (observer-first → governed-at-boundaries) – CogPR-74: Arena velocity guard (convergence faster than evidence) – CogPR-76: Same-model convergence discount – CogPR-77: Concession cascade detection – CogPR-78: showClearContextOnPlanAccept must be true (cadence handoff chain)5cfcc9403/25/2026, 19:32:2903/25, 19:32mearley24/AI-Serverexpand: monitor top 25 wallets instead of 1083e390703/25/2026, 19:32:2803/25, 19:32ayeshakhalid192007-dev/humanoid-ai-studiofix(ci): remove deploy hook step — Railway deploys via GitHub integration Railway’s ‘Wait for CI’ toggle handles deployment automatically after CI passes. No deploy hook or token needed. Workflows now only run tests/audit as the CI gate Railway watches. Co-Authored-By: Claude Sonnet 4.6 1fc110003/25/2026, 19:29:1803/25, 19:29ydesjardins200-coder/wavesfinancialAdd privacy-policy.html — 11 sections, sticky TOC, plain-language summaries, data chips, third-party table, cookie grid, rights grid + update footer links across all pagesa3a412303/25/2026, 19:28:1703/25, 19:28u2giants/mission-controlfeat: initial scaffold — full Mission Control platform – Dockerfile, docker-compose, nginx, entrypoint, configure – Express API: missions, tasks, agents, logs (SSE), config routes – SQLite schema with WAL mode – Agent soul file management – Model registry (shared format with gateway, 9 models) – Multi-provider worker engine (Google/Anthropic/OpenAI unified adapter) – Worker pool with 2s polling, dependency-aware task assignment – Tool implementations: bash, http, file r/w, git, coolify, cloudflare, spawn_subtask – React UI: Dashboard, Missions kanban, Task detail with live SSE logs, Agents, Settings – Settings page includes model registry editor with inline editing + Push to Gateway – Smoke tested: healthz OK, Agent Alpha seeded, 9 models loaded Co-Authored-By: Claude Sonnet 4.6 c0af10103/25/2026, 19:27:1503/25, 19:27gvinsot/PulsarTeamfix: allow manual task column/step changes by skipping on_enter workflow transitions When a user manually drags or changes a task’s status, on_enter workflow transitions were firing immediately (e.g. change_status actions), reverting the move before the frontend could refresh. Now _checkAutoRefine receives the `by` parameter and skips on_enter transitions when by=’user’.- Auto-assign by column role still works for manual moves – Condition-based transitions still evaluated by periodic checker – Workflow-initiated changes (agents, jira-sync) still trigger on_enter – Added error handling in frontend handleDrop and handleStatusChange Co-Authored-By: Claude Opus 4.6 f92e28a03/25/2026, 19:26:2703/25, 19:26ayeshakhalid192007-dev/humanoid-ai-studiofix(backend): install CPU-only PyTorch to reduce Railway image size CUDA PyTorch is ~1.5GB causing Railway image push to fail. CPU-only build is ~200MB, well within trial account limits. sentence-transformers only needs CPU inference for embeddings. Co-Authored-By: Claude Sonnet 4.6 Top RepositoriesSince Launch · All · ranked by commitsCommitsAddedDeletedLanguagesCode distributionShowing total commits by languageStarsAll2+100+1000+Window7D30D90D1YSince Launch