SlopScore
The Trawl Report · 2026-W38

Week 38 of 2026 · 14 to 20 September

Written 2026-09-21T00:05:43Z · method v4 · counted, never generated

listed?Every repo on the site right now, trawled and opted in together.468+254 this weekacross 431 owners
trawled?Found by the Cap'm's search of public GitHub. Nobody submitted these: the owner said in public that a model wrote the code, and the trawl took them at their word. Within that label it is close to a random sample.456+254 dragged out of public GitHub; nobody submitted them
opted in?The owner committed a slopscore.md to the repo and asked to be counted. Self-selected, so a fact about volunteers.12±0 committed a slopscore.md and asked to be counted
the judge has seen?Everything the trawl paid a small model to sort: what it let through and what it threw back, one population. A model's label from a closed list, and the only judged number in the bulletin.416+313 this week70.9% was software somebody made

Biggest swings this week

never claims a model wrote it?Nothing in the description or README says, in the past tense, that an AI tool wrote this. That claim is the whole population; without it the repo is not in it. · thrown back−15.2 pts
app?The judge's verdict: an application somebody made for a purpose. One of the five verdicts that gets a repo listed. · verdict−13.7 pts
a licence we cannot quote from?The licence is not on the permissive list (MIT, Apache-2.0, BSD, ISC, 0BSD, Unlicense, CC0), or there is none. The Cap'm's paperwork quotes the README, so the trawl lists only what it is allowed to quote. · thrown back+12.7 pts
tool-for-ai-coding?A verdict that throws the candidate back. Its users are people coding with AI: agent frameworks, rules and prompt packs, skills, editor add-ons, MCP servers for coding agents, usage dashboards. A tool for vibe coders is not vibe-coded software. · verdict+12.2 pts
something else?A reason too rare to earn a bucket of its own. · thrown back+6.2 pts
game?The judge's verdict: a game. One of the five verdicts that gets a repo listed. · verdict−3.9 pts
codex?OpenAI's Codex coding agent. · tool−2.4 pts
library?The judge's verdict: a library or package for other code to use. One of the five verdicts that gets a repo listed. · verdict+2.4 pts
Share points?The gap between this week's share and last week's, in percentage points. A section that grew moves every count; share points show who gained ground and who gave it up. gained or lost against last week's snapshot, across every counted section. Pink went up, purple went down; the number is the move, not the size. A big swing on a small section is the mix changing, not the world.

Counted from the snapshot of 2026-09-21, set against the one from 2026-09-14. 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 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

7,302 repos seen in the last 90 days that say in public an AI tool wrote them, from 6,217 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.

  • 74.9% have no stars at all
  • 59.7% carry no licence
  • 9.4% would reach the judge under the trough's rules

By stars: no stars 74.9%, 1 to 4 18.2%, 5 to 9 2.4%, 10 to 24 1.8%, 25 to 99 1.3%, 100 to 499 0.8%, 500 or more 0.6%.

What stops the rest, by the first rule that stopped it:

  • a licence we cannot quote from?The licence is not on the permissive list (MIT, Apache-2.0, BSD, ISC, 0BSD, Unlicense, CC0), or there is none. The Cap'm's paperwork quotes the README, so the trawl lists only what it is allowed to quote. — 4294 (58.8%)
  • outside the star window?Fewer than 1 or more than 2000 stars. Both ends are our rule, not a finding: the window is where nobody has heard of a repo yet and nobody has already written it up. — 1267 (17.4%)
  • owned by an org, so nobody could claim or remove it?The owner is an organisation, not a person. Claiming, editing and removing a trawled listing all work by logging in as the repo's owner, and nobody can log in as an org. — 885 (12.1%)
  • writing about vibe coding, not vibe-coded?By its own name or description, a list, guide or prompt pack about vibe coding, or a tool for vibe coding. Not software somebody vibe coded. — 125 (1.7%)
  • no claim in its topics or description — 35 (0.5%)
  • no description to quote?The repo has no description. The Cap'm's paperwork quotes it, so an empty one leaves nothing to say. — 12 (0.2%)

The sea is the trough's own searches with the ropes taken off, so it inherits bias 1 and bias 2 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

trawled?Found by the Cap'm's search of public GitHub. Nobody submitted these: the owner said in public that a model wrote the code, and the trawl took them at their word. Within that label it is close to a random sample. 456 97.4%opted in?The owner committed a slopscore.md to the repo and asked to be counted. Self-selected, so a fact about volunteers. 12 2.6%

Every listing, by how it got here. The purple sliver is the only part of the site anyone volunteered for.

468 repos listed, +254 this week, from 431 owners.

  • trawled: 456 (+254) — dragged out of public GitHub; nobody submitted them
  • opted in: 12 (no change) — committed a slopscore.md and asked to be counted

How much of it is software

the pile

What the judge saw over 416 candidatesapp: 149 (35.8%)tool: 93 (22.4%)library: 26 (6.3%)game: 24 (5.8%)hardware: 3 (0.7%)thrown back: 121 (29.1%)
70.9%software somebody made
  • app?The judge's verdict: an application somebody made for a purpose. One of the five verdicts that gets a repo listed.149 35.8%
  • tool?The judge's verdict: a tool, a utility, a script somebody made to do a job. One of the five verdicts that gets a repo listed.93 22.4%
  • library?The judge's verdict: a library or package for other code to use. One of the five verdicts that gets a repo listed.26 6.3%
  • game?The judge's verdict: a game. One of the five verdicts that gets a repo listed.24 5.8%
  • hardware?The judge's verdict: software with hardware attached, such as firmware or a device. One of the five verdicts that gets a repo listed.3 0.7%
  • thrown back?Everything the judge looked at and did not let through, every reason folded together.121 29.1%

what moved

app?The judge's verdict: an application somebody made for a purpose. One of the five verdicts that gets a repo listed.35.8% 149−13.7 pts
tool?The judge's verdict: a tool, a utility, a script somebody made to do a job. One of the five verdicts that gets a repo listed.22.4% 93+1.0 pts
tool-for-ai-coding?A verdict that throws the candidate back. Its users are people coding with AI: agent frameworks, rules and prompt packs, skills, editor add-ons, MCP servers for coding agents, usage dashboards. A tool for vibe coders is not vibe-coded software.20% 83+12.2 pts
library?The judge's verdict: a library or package for other code to use. One of the five verdicts that gets a repo listed.6.3% 26+2.4 pts
game?The judge's verdict: a game. One of the five verdicts that gets a repo listed.5.8% 24−3.9 pts
list-or-template?A verdict that throws the candidate back. An awesome list, guide, tutorial, course, template, boilerplate, starter kit or dotfiles. Not software somebody made.5% 21−0.8 pts
not-ai-made?A verdict that throws the candidate back. The AI claim the trawl found was not about how this project's code was written.2.6% 11new
low-effort?A verdict that throws the candidate back. A stub, a placeholder or a homework dump with no real README.1% 4new
416 candidates judged?Everything the trawl paid a small model to sort: what it let through and what it threw back, one population. A model's label from a closed list, and the only judged number in the bulletin. in the trawled?Found by the Cap'm's search of public GitHub. Nobody submitted these: the owner said in public that a model wrote the code, and the trawl took them at their word. Within that label it is close to a random sample. sample. In what moved, the bar is this week, the hollow marker is where it stood a week ago, and the chip is the distance in share points?The gap between this week's share and last week's, in percentage points. A section that grew moves every count; share points show who gained ground and who gave it up..

One population, both halves: everything the judge was paid to look at, whether it was listed or not. The five listable codes keep their own colours; every reason to throw a candidate back shares the grey slice. These are a model's labels from a closed list, not a maker's own words.

The judge has looked at 416 candidates so far (+313 new this week), and 70.9% 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?The judge's verdict: an application somebody made for a purpose. One of the five verdicts that gets a repo listed. — 149 (35.8%), −13.7 points from 51 (49.5%)
  • tool?The judge's verdict: a tool, a utility, a script somebody made to do a job. One of the five verdicts that gets a repo listed. — 93 (22.4%), +1.0 points from 22 (21.4%)
  • tool-for-ai-coding?A verdict that throws the candidate back. Its users are people coding with AI: agent frameworks, rules and prompt packs, skills, editor add-ons, MCP servers for coding agents, usage dashboards. A tool for vibe coders is not vibe-coded software. — 83 (20%), +12.2 points from 8 (7.8%)
  • library?The judge's verdict: a library or package for other code to use. One of the five verdicts that gets a repo listed. — 26 (6.3%), +2.4 points from 4 (3.9%)
  • game?The judge's verdict: a game. One of the five verdicts that gets a repo listed. — 24 (5.8%), −3.9 points from 10 (9.7%)
  • list-or-template?A verdict that throws the candidate back. An awesome list, guide, tutorial, course, template, boilerplate, starter kit or dotfiles. Not software somebody made. — 21 (5%), −0.8 points from 6 (5.8%)
  • not-ai-made?A verdict that throws the candidate back. The AI claim the trawl found was not about how this project's code was written. — 11 (2.6%), new this week
  • low-effort?A verdict that throws the candidate back. A stub, a placeholder or a homework dump with no real README. — 4 (1%), new this week

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

the pile

Which tool gets the credit, 82 creditsclaude-code: 45 (54.9%)claude: 26 (31.7%)cursor: 4 (4.9%)codex: 3 (3.7%)copilot: 1 (1.2%)lovable: 1 (1.2%)other: 2 (2.4%)
54.9%claude-code
  • claude-code?Anthropic's Claude Code, the terminal agent. Counted apart from plain Claude.45 54.9%
  • claude?Claude used through the app, the API or an editor, where the owner did not say Claude Code.26 31.7%
  • cursor?Cursor, the AI editor.4 4.9%
  • codex?OpenAI's Codex coding agent.3 3.7%
  • copilot?GitHub Copilot, in any of its modes.1 1.2%
  • lovable?Lovable, the prompt-to-app builder.1 1.2%
  • other?A tool the owner named that is not on the built_with list.2 2.4%

what moved

claude-code?Anthropic's Claude Code, the terminal agent. Counted apart from plain Claude.54.9% 45flat
claude?Claude used through the app, the API or an editor, where the owner did not say Claude Code.31.7% 26−1.6 pts
cursor?Cursor, the AI editor.4.9% 4+1.8 pts
codex?OpenAI's Codex coding agent.3.7% 3−2.4 pts
copilot?GitHub Copilot, in any of its modes.1.2% 1−1.8 pts
lovable?Lovable, the prompt-to-app builder.1.2% 1new
muse-code?Meta's Muse Code, also written Muse Spark or Meta AI.1.2% 1new
windsurf?Windsurf, the AI editor.1.2% 1new
82 credits?A credit is one repo naming one tool. A repo that names two tools gives two credits, so shares here are of credits, not of repos. in the trawled?Found by the Cap'm's search of public GitHub. Nobody submitted these: the owner said in public that a model wrote the code, and the trawl took them at their word. Within that label it is close to a random sample. sample. In what moved, the bar is this week, the hollow marker is where it stood a week ago, and the chip is the distance in share points?The gap between this week's share and last week's, in percentage points. A section that grew moves every count; share points show who gained ground and who gave it up..

82 credits in the trawled sample. A repo can name more than one tool, so the shares are of credits, not of repos.

  • claude-code?Anthropic's Claude Code, the terminal agent. Counted apart from plain Claude. — 45 (54.9%), flat
  • claude?Claude used through the app, the API or an editor, where the owner did not say Claude Code. — 26 (31.7%), −1.6 points from 11 (33.3%)
  • cursor?Cursor, the AI editor. — 4 (4.9%), +1.8 points from 1 (3%)
  • codex?OpenAI's Codex coding agent. — 3 (3.7%), −2.4 points from 2 (6.1%)
  • copilot?GitHub Copilot, in any of its modes. — 1 (1.2%), −1.8 points from 1 (3%)
  • lovable?Lovable, the prompt-to-app builder. — 1 (1.2%), new this week
  • muse-code?Meta's Muse Code, also written Muse Spark or Meta AI. — 1 (1.2%), new this week
  • windsurf?Windsurf, the AI editor. — 1 (1.2%), new this week

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, 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: codex −2.4 points, cursor +1.8 points, copilot −1.8 points, claude −1.6 points.

What it is written in

the pile

What it is written in, 1280 language creditspython: 196 (15.3%)javascript: 190 (14.8%)shell: 173 (13.5%)html: 165 (12.9%)css: 144 (11.3%)typescript: 122 (9.5%)other: 290 (22.7%)
15.3%python
  • python196 15.3%
  • javascript190 14.8%
  • shell173 13.5%
  • html165 12.9%
  • css144 11.3%
  • typescript122 9.5%
  • other?A tool the owner named that is not on the built_with list.290 22.7%

what moved

python15.3% 196−0.7 pts
javascript14.8% 190+1.7 pts
shell13.5% 173−1.4 pts
html12.9% 165+1.5 pts
css11.3% 144+1.2 pts
typescript9.5% 122+1.1 pts
rust5.2% 66−1.8 pts
dockerfile4.8% 61flat
1,280 language credits?GitHub lists every language a repo uses, up to eight here, and each one is a credit. Shares are of credits, not of repos, so one repo can count more than once. in the trawled?Found by the Cap'm's search of public GitHub. Nobody submitted these: the owner said in public that a model wrote the code, and the trawl took them at their word. Within that label it is close to a random sample. sample. In what moved, the bar is this week, the hollow marker is where it stood a week ago, and the chip is the distance in share points?The gap between this week's share and last week's, in percentage points. A section that grew moves every count; share points show who gained ground and who gave it up..
  • python — 196 (15.3%), −0.7 points from 101 (16%)
  • javascript — 190 (14.8%), +1.7 points from 83 (13.2%)
  • shell — 173 (13.5%), −1.4 points from 94 (14.9%)
  • html — 165 (12.9%), +1.5 points from 72 (11.4%)
  • css — 144 (11.3%), +1.2 points from 63 (10%)
  • typescript — 122 (9.5%), +1.1 points from 53 (8.4%)
  • rust — 66 (5.2%), −1.8 points from 44 (7%)
  • dockerfile — 61 (4.8%), flat

Also new to the sample this week, below the top eight: go (36).

Shares are of 1,280 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

the pile

What the net threw back, 5183 candidatesa licence we cannot quote from: 2740 (52.9%)never claims a model wrote it: 1209 (23.3%)owned by an org, so nobody could claim or remove it: 516 (10%)something else: 462 (8.9%)writing about vibe coding, not vibe-coded: 124 (2.4%)a tool for people coding with AI: 87 (1.7%)other: 45 (0.9%)
52.9%the biggest slice
  • a licence we cannot quote from?The licence is not on the permissive list (MIT, Apache-2.0, BSD, ISC, 0BSD, Unlicense, CC0), or there is none. The Cap'm's paperwork quotes the README, so the trawl lists only what it is allowed to quote.2740 52.9%
  • never claims a model wrote it?Nothing in the description or README says, in the past tense, that an AI tool wrote this. That claim is the whole population; without it the repo is not in it.1209 23.3%
  • owned by an org, so nobody could claim or remove it?The owner is an organisation, not a person. Claiming, editing and removing a trawled listing all work by logging in as the repo's owner, and nobody can log in as an org.516 10%
  • something else?A reason too rare to earn a bucket of its own.462 8.9%
  • writing about vibe coding, not vibe-coded?By its own name or description, a list, guide or prompt pack about vibe coding, or a tool for vibe coding. Not software somebody vibe coded.124 2.4%
  • a tool for people coding with AI?A verdict that throws the candidate back. Its users are people coding with AI: agent frameworks, rules and prompt packs, skills, editor add-ons, MCP servers for coding agents, usage dashboards. A tool for vibe coders is not vibe-coded software.87 1.7%
  • other?A tool the owner named that is not on the built_with list.45 0.9%

what moved

a licence we cannot quote from?The licence is not on the permissive list (MIT, Apache-2.0, BSD, ISC, 0BSD, Unlicense, CC0), or there is none. The Cap'm's paperwork quotes the README, so the trawl lists only what it is allowed to quote.52.9% 2740+12.7 pts
never claims a model wrote it?Nothing in the description or README says, in the past tense, that an AI tool wrote this. That claim is the whole population; without it the repo is not in it.23.3% 1209−15.2 pts
owned by an org, so nobody could claim or remove it?The owner is an organisation, not a person. Claiming, editing and removing a trawled listing all work by logging in as the repo's owner, and nobody can log in as an org.10% 516new
something else?A reason too rare to earn a bucket of its own.8.9% 462+6.2 pts
writing about vibe coding, not vibe-coded?By its own name or description, a list, guide or prompt pack about vibe coding, or a tool for vibe coding. Not software somebody vibe coded.2.4% 124flat
a tool for people coding with AI?A verdict that throws the candidate back. Its users are people coding with AI: agent frameworks, rules and prompt packs, skills, editor add-ons, MCP servers for coding agents, usage dashboards. A tool for vibe coders is not vibe-coded software.1.7% 87flat
a list, guide or template?A verdict that throws the candidate back. An awesome list, guide, tutorial, course, template, boilerplate, starter kit or dotfiles. Not software somebody made.0.5% 24flat
the claim was not about this code?A verdict that throws the candidate back. The AI claim the trawl found was not about how this project's code was written.0.2% 11new
5,183 candidates thrown back in 30 days?Everything the trawl looked at in the last 30 days and did not list, by the first rule it failed. The window rolls, so week-to-week movement is the mix of reasons shifting, not one week's intake. in the trawled?Found by the Cap'm's search of public GitHub. Nobody submitted these: the owner said in public that a model wrote the code, and the trawl took them at their word. Within that label it is close to a random sample. sample. In what moved, the bar is this week, the hollow marker is where it stood a week ago, and the chip is the distance in share points?The gap between this week's share and last week's, in percentage points. A section that grew moves every count; share points show who gained ground and who gave it up..

5,183 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?The licence is not on the permissive list (MIT, Apache-2.0, BSD, ISC, 0BSD, Unlicense, CC0), or there is none. The Cap'm's paperwork quotes the README, so the trawl lists only what it is allowed to quote. — 2740 (52.9%), +12.7 points from 399 (40.1%)
  • never claims a model wrote it?Nothing in the description or README says, in the past tense, that an AI tool wrote this. That claim is the whole population; without it the repo is not in it. — 1209 (23.3%), −15.2 points from 383 (38.5%)
  • owned by an org, so nobody could claim or remove it?The owner is an organisation, not a person. Claiming, editing and removing a trawled listing all work by logging in as the repo's owner, and nobody can log in as an org. — 516 (10%), new this week
  • something else?A reason too rare to earn a bucket of its own. — 462 (8.9%), +6.2 points from 27 (2.7%)
  • writing about vibe coding, not vibe-coded?By its own name or description, a list, guide or prompt pack about vibe coding, or a tool for vibe coding. Not software somebody vibe coded. — 124 (2.4%), flat
  • a tool for people coding with AI?A verdict that throws the candidate back. Its users are people coding with AI: agent frameworks, rules and prompt packs, skills, editor add-ons, MCP servers for coding agents, usage dashboards. A tool for vibe coders is not vibe-coded software. — 87 (1.7%), flat
  • a list, guide or template?A verdict that throws the candidate back. An awesome list, guide, tutorial, course, template, boilerplate, starter kit or dotfiles. Not software somebody made. — 24 (0.5%), flat
  • the claim was not about this code?A verdict that throws the candidate back. The AI claim the trawl found was not about how this project's code was written. — 11 (0.2%), new this week

The small print

  • Counted under method v4, frozen and versioned at /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 16 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. A tool not on that list was not looked for. Which tools were added, and when.
  • 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.
  • The numbers behind every sentence above: /report/2026-W38.json, frozen with the bulletin and still there after the nightly snapshot they came from has been pruned.
  • Spotted an error? Say so. Corrections run in the next bulletin; nothing is edited in quietly.

Get it weekly: SlopScore on Substack by email, or RSS. Free, and it stays free — it is how the numbers get cited, not how they get sold. We never see your address: the list lives with them, not here.

Written 2026-09-21T00:05:43Z · method v4 · markdown · the numbers · RSS · the live dashboard

Every bulletin

2026-W38468 listed (+254)70.9% software2026-W37214 listed86.4% software