The five
The strongest feats of the week. Each with its result, its grade, the agent, a link to the feat and one to its source.
Get the weekly five
The five strongest AI agent feats of the week, each with its receipt, plus a short roundup of AI news: new models, new tools, launches. A weekly newsletter.
5 feats · the week in AI news · once a week · free
You're on the list.
A free weekly newsletter. Your address is used only to send it, and never sold or shared. Privacy
This week's fiveOct 13 · picked from 280 new feats
Resultrc_V252CKAG
Db2 support shipped in under two days for a $4172 token bill, versus 9 months and about $160K for Oracle; 27 PRs merged.
ShownSoftware companies
Custom agent
Hospital-style agent pipeline ships Db2 support
Cockroach Labs ran a multi-role coding agent pipeline modeled on a teaching hospital to add Db2 support to its MOLT migration tool.
Resultrc_5BGE87N0
Non-test Python fell from 1,063,826 to 698,363 lines (34.4%), gateway/run.py from 34,847 to 5,512 lines, and functions over 300 lines from 192 to 2.
ShownSoftware companies
Hermes Agent
Million-line codebase refactor
Hermes Agent split a million-line Python repo into 36 groups and dispatched 1,393 subagents in git worktrees to refactor it.
Resultrc_WMXVQ46W
Each model pass added about 1.5x to 2.0x speedup, reaching roughly 7.5x to 32x over baseline; one implementation beat xgboost on speed.
ShownDeveloper
Codex, Other
Faster Rust ML code via benchmarks
Coding agents wrote Rust ML algorithms like UMAP from scratch, then iteratively optimized them against criterion benchmarks.
Resultrc_7DV47BXJ
Bought a new KitchenAid dishwasher within days, found a $25 armchair, caught a last-show date night, and published a landing page via Cloudflare.
ShownPersonal
Grok Bot, Other
Household errands with always-on bots
Eleanor Konik used Muse and Grok Bot for library requests, dishwasher shopping, marketplace scans, party planning, grant research, and publishing a landing page.
Resultrc_YPS8RECM
A live game with 87 rules, 33 conditions, 14 spells and 15 monsters in two editions, answering rules questions in about 1.3 to 2.2 seconds.
ShownDeveloper
Custom agent
Rule-citing AI dungeon master game
The author built a browser D&D 5e dungeon crawl whose AI Dungeon Master answers rules questions by looking up and citing rules stored in Sanity.
Next issue
Five more like these, plus the week in AI news
Our picks first, then the strongest new feats of the week.
Get the weekly five- Per issue
- 5feats
- How often
- 1a week
- On file
- 28053% with the work shown
- Price
- 0free
What each issue holds
The week in AI news
A short, curated list of what changed in AI that week: new models, new tools, launches. In our own words, each linked to the original.
Results by agent
Which agents people got results with that week, counted from our feats, and any agent new on the radar.
How the five are chosen
A pipeline reads posts, repos, and articles every day and drafts a feat for each real result it finds, in our own words, with the source linked. Every feat is checked automatically; flagged ones are reviewed by a person. The issue takes our picks of the week first, then the highest-scored new feats.
Each feat carries its grade: Shown the work is linked, Numbers a specific number is stated, Claim a claim only. Partial results and failures go in too, marked as such.
Questions
- How often does it come?
- Once a week. One email, never more.
- Why weekly?
- One email a week is easy to keep. We would rather send the five strongest feats than fill your inbox.
- Does it cost anything?
- No. If we ever carry a sponsor, it is one labeled line at the bottom, and it never decides which feats you see.
- What do you do with my email?
- We keep your address, the date you signed up, and which form you used, only to send this newsletter. It is never sold or shared. The privacy page says how to have it removed.