# The Trawl Report · 2026-W41 Counted from the snapshot of 2026-10-12, set against the one from 2026-10-05. Method v4. No model wrote a word of this: every sentence is a template with a count in it. > The Cap'm counts what came up in the net and what went back over the side. Read [how we count](/method) before you quote any of it, the tool numbers most of all: they measure how loudly a tool's users say its name, not what anybody uses. ## The sea 26,622 repos seen in the last 90 days that say in public an AI tool wrote them, +4292 this week, from 21,763 owners. Counted from GitHub's search results alone: no star floor, no licence filter, no README read, no model asked, and nothing listed. This is the water the trough below is drawn from. - **81.1%** have no stars at all - **68.6%** carry no licence - **5.8%** would reach the judge under the trough's rules By stars: no stars 81.1%, 1 to 4 14.8%, 5 to 9 1.6%, 10 to 24 1.1%, 25 to 99 0.7%, 100 to 499 0.4%, 500 or more 0.2%. What stops the rest, by the first rule that stopped it: - **a licence we cannot quote from** — 16768 (63%), +1.3 points from 13780 (61.7%) - **outside the star window** — 4024 (15.1%), flat - **owned by an org, so nobody could claim or remove it** — 3695 (13.9%), −1.0 points from 3320 (14.9%) - **writing about vibe coding, not vibe-coded** — 453 (1.7%), flat - **no claim in its topics or description** — 130 (0.5%), flat - **no description to quote** — 14 (0.1%), flat - **not touched in 90 days** — 6 (0%), flat The sea is the trough's own searches with the ropes taken off, so it inherits [bias 1 and bias 2](/method) whole: still repos whose owners said so, still found by searches named after tools. What it takes off is the star floor and the licence filter, and the three lines above are how much of the population those two rules hold back. ## The trough 1,184 repos listed, +275 this week, from 1,120 owners. - trawled: 1,170 (+274) — dragged out of public GitHub; nobody submitted them - opted in: 14 (+1) — committed a slopscore.md and asked to be counted ## How much of it is software The judge has looked at 1,472 candidates so far (+398 new this week), and **69.3% were software somebody made**: an app, a game, a tool, a library, or something with hardware attached. The rest went back over the side. That is the number this bulletin exists for, and the reason the trawl keeps what it rejects. A leaderboard that keeps only what it accepted can tell you what its winners look like and nothing about the pile they came out of. The pile, largest bucket first: - **app** — 554 (37.6%), flat - **tool-for-ai-coding** — 315 (21.4%), flat - **tool** — 305 (20.7%), flat - **library** — 82 (5.6%), flat - **game** — 62 (4.2%), −0.7 points from 53 (4.9%) - **list-or-template** — 59 (4%), flat - **not-ai-made** — 55 (3.7%), flat - **hardware** — 17 (1.2%), flat These are a model's labels, picked from a closed list, not a maker's own words. They are the only judged numbers in the bulletin, and they say so wherever they appear. ## Which tool gets the credit 197 credits in the trawled sample. A repo can name more than one tool, so the shares are of credits, not of repos. - **claude-code** — 103 (52.3%), +2.3 points from 77 (50%) - **claude** — 52 (26.4%), −2.8 points from 45 (29.2%) - **codex** — 21 (10.7%), +1.6 points from 14 (9.1%) - **cursor** — 7 (3.6%), flat - **lovable** — 5 (2.5%), −0.7 points from 5 (3.2%) - **chatgpt** — 3 (1.5%), +0.9 points from 1 (0.6%) - **copilot** — 2 (1%), flat - **aider** — 1 (0.5%), flat **Read this as how loudly each tool's users say its name in public.** Most of the trawl's eight grounds are named after a tool ([method](/method), bias 2), so a tool whose users never tag or describe their repos is undercounted by construction. It is not market share and it is not usage. Biggest swings since last week: claude −2.8 points, claude-code +2.3 points, codex +1.6 points, chatgpt +0.9 points, lovable −0.7 points. ## What it is written in - **python** — 549 (16.2%), flat - **javascript** — 506 (15%), flat - **shell** — 479 (14.2%), flat - **html** — 443 (13.1%), flat - **css** — 379 (11.2%), flat - **typescript** — 310 (9.2%), flat - **dockerfile** — 148 (4.4%), flat - **makefile** — 138 (4.1%), flat Shares are of 3,379 language credits over the trawled sample. GitHub lists every language a repo uses, so one repo can count more than once. ## What the net threw back 12,841 candidates looked at in the last 30 days and not listed, by reason. The window rolls, so movement here is the mix of reasons shifting, not one week's intake. - **a licence we cannot quote from** — 7672 (59.7%), flat - **never claims a model wrote it** — 2022 (15.7%), flat - **owned by an org, so nobody could claim or remove it** — 1306 (10.2%), flat - **something else** — 720 (5.6%), −0.5 points from 632 (6.1%) - **no description to quote** — 380 (3%), +0.7 points from 236 (2.3%) - **a tool for people coding with AI** — 319 (2.5%), flat - **writing about vibe coding, not vibe-coded** — 294 (2.3%), flat - **a list, guide or template** — 62 (0.5%), flat ## The small print - Counted under method v4, frozen and versioned at [/method](/method). When a rule changes the version goes up and the change is dated; nothing is edited in place. - This site measures software whose author **says in public** that an AI tool wrote it. It does not measure AI-written software, and no number here can. - Tool credit was counted over the 17 tools the dictionary held that night: claude-code, claude, cursor, copilot, chatgpt, codex, gemini-cli, windsurf, aider, cline, roo, muse-code, lovable, bolt, v0, replit, antigravity. A tool not on that list was not looked for. [Which tools were added, and when](/method). - The sea is counted from GitHub search results alone. Nothing in it is read, judged, listed or named, and its facts are as of the last time the sounding passed. [How it is sounded](/method). - The numbers behind every sentence above: [`/report/2026-W41.json`](/report/2026-W41.json), frozen with the bulletin and still there after the nightly snapshot they came from has been pruned. - Spotted an error? [Say so](/contact). Corrections run in the next bulletin; nothing is edited in quietly.