SutroYaro
AGENTS.md16 starsChanged 6 months agoUnlicenseView on GitHub
These instructions apply to all Codex CLI sessions in this repo. Load current project state before running experiments, reviewing PRs, or writing findings. If telegram.db is missing or stale, run bin/tg-sync. Run at session start and before any push: Before pushing: 1. Update docs/changelog.md (bump version) 2. python3 -m mkdocs build to verify no broken links 3. Show the diff and wait for approval before git push Apply these to all prose (findings docs, DISCOVERIES.md updates, PR descriptions): Full guide:…
- Commits and pushes
Where files are headingNew AGENTS.md files have more rules about keeping secrets out of the code, and fewer about code style.
Of the AGENTS.md files we read, the 33,315 first seen in the last six months, against the older ones: the kind of rule that gained most share, and the kind that lost most. About files like this in general, not this one, and an estimate, since a kind is a pattern in the words.
In this repository9 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- anti-slop-guideSkill
- info-defragSkill
- prepare-meetingSkill
- run-experimentSkill
- sutro-contextSkill
- sutro-syncSkill
- weekly-catchupSkill
The file
We split this AGENTS.md into the rules it gives an agent
11 rules. About 0 are common, 9 are distinctive, and 2 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈9
Distinctive · estimate
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 9
Read these files in order:
- Distinctive: line 15
Check recent Telegram activity (if synced):
- Very specific: line 22
If telegram.db is missing or stale, run bin/tg-sync.
- Distinctive: line 24
Check GitHub for open work:
- Distinctive: line 31
Before writing code, check:
- Distinctive: line 49
Run at session start and before any push:
- Very specific: line 70
Update docs/changelog.md (bump version)
- Distinctive: line 82
Prefer plain verbs. "used" not "leveraged," "showed" not "showcased."
- Distinctive: line 83
Use simple copulatives. Write "X is Y" not "X serves as Y."
- Distinctive: line 85
Never triple. Two items in a list, not three.
- Distinctive: line 87
No AI vocabulary: delve, tapestry, landscape, pivotal, showcase, testament, underscore, foster, garner, interplay, intricate, vibrant, …
≈ “Common” is an estimate: at least 1 in 5 of the 42,283 AGENTS.md files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Codex Project Instructions
These instructions apply to all Codex CLI sessions in this repo.
Before Any Research Task
Load current project state before running experiments, reviewing PRs, or writing findings.
- Read these files in order:
- CODEX.md -- project context, current best methods, constraints
- DISCOVERIES.md -- what's proven, what failed, open questions (bottom of file)
- AGENT.md -- machine-executable experiment loop (if running autonomous)
- LAB.md -- experiment protocol, rules (especially rule #9: metric isolation)
- Check recent Telegram activity (if synced):
test -f telegram.db && sqlite3 telegram.db \
"SELECT date, sender, text FROM messages ORDER BY date DESC LIMIT 10;"If telegram.db is missing or stale, run bin/tg-sync.
- Check GitHub for open work:
gh pr list --repo cybertronai/SutroYaro --state open
gh issue list --repo cybertronai/SutroYaro --state open- Before writing code, check:
research/search_space.yamlfor allowed parameter rangesresearch/questions.yamlfor the dependency graph of open questions
Current State
| Fact | Value |
| Fastest historical method | GF(2) Gaussian elimination, 509us, ARD ~420 |
| Primary energy metric | ByteDMD, byte-granularity LRU stack cost |
| Legacy metrics | ARD and DMC values remain useful for historical comparisons |
| Experiments done | 37+ (see research/log.jsonl and docs/research/survey.md) |
| Open questions | Bottom of DISCOVERIES.md |
| Next milestone | ByteDMD-aware sparse parity and nanoGPT experiments |
| Meeting cadence | Mondays 18:00 at South Park Commons |
Sync Routine
Run at session start and before any push:
# Telegram (daily, incremental SQLite sync)
bin/tg-sync
# Query recent messages
sqlite3 telegram.db "SELECT date, sender, text FROM messages ORDER BY date DESC LIMIT 10"
# Post status updates when configured
bin/tg-post --topic agent-updates "Status update"
# Google Docs (weekly, after Monday meetings)
python3 src/sync_google_docs.py
# GitHub
gh pr list --repo cybertronai/SutroYaro --state open
gh issue list --repo cybertronai/SutroYaro --state openBefore pushing:
- Update
docs/changelog.md(bump version) python3 -m mkdocs buildto verify no broken links- Show the diff and wait for approval before
git push
Writing Rules (Anti-Slop)
Apply these to all prose (findings docs, DISCOVERIES.md updates, PR descriptions):
- Cut filler phrases. Say the thing directly.
- Break formulaic structures. No binary contrasts, no dramatic fragmentation.
- Vary rhythm. Mix sentence lengths. Two items beat three.
- Trust readers. State facts directly.
- Prefer plain verbs. "used" not "leveraged," "showed" not "showcased."
- Use simple copulatives. Write "X is Y" not "X serves as Y."
- Kill em dashes. Use commas or periods.
- Never triple. Two items in a list, not three.
- Be specific. Replace generic statements with concrete details.
- No AI vocabulary: delve, tapestry, landscape, pivotal, showcase, testament, underscore, foster, garner, interplay, intricate, vibrant, robust, seamless, paramount, multifaceted, nuanced, groundbreaking, cornerstone, transformative, synergy.
Full guide: .claude/skills/anti-slop-guide/SKILL.md (plain markdown, readable by any tool).
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

