AI Agents
3 of the 64 write-ups here are tagged AI Agents. Every figure quoted below is in the figures table with the sentence it came from.
The figures
- 40% — 40% of review tasks needed manual backfill, costing about 15 extra minutes each time. →
- 1,000 tokens — Its original system prompt and tool descriptions came to under 1,000 tokens. →
- 300 million tokens — One developer logged token consumption across the first three days of an agent-driven project in July 2026 and got a curve that says almost everything I think is worth saying about this technology: 300 million tokens on day one, about 200 million on day two, about 100 million on day three. →
- 3 weeks — Version your skills like software. treats AI skills as packages with rollback and permission control, which is the difference between a team that can undo a bad change and a team whose output quality quietly degrades because somebody edited a prompt in place 3 weeks ago. →
- 3 minutes — The numbers underneath that headline are the interesting part: 3 minutes for a deep market research pass, 1,800 designs produced in a month, 12 minutes from a customer request to a rendering, and orders signed on the spot while the buyer was still in the conversation. →
- 1% — The version of this I have seen bite hardest is on the payments side, where one developer’s Stripe account was frozen for a dispute rate above 1% and 1,500-plus paying subscriptions stopped at once. →
The write-ups
Your AI Didn’t Misread Your Code by Accident. You Handed It the Wrong Context.
“If your AI keeps writing bad code, you need a better model.” In each of the rebuilds below, a new model wouldn’t have fixed anything, because the model was reading the wrong thing.
One Person, 8 AI Agents, 3,000 Baseball Caps in Two Months
Zhang Qianchao runs a custom baseball cap export business on Alibaba.com with 8 AI agents and no employees, and in 2 months he shipped 3,000 caps to buyers across Europe, the Americas and South Ame…
Why Your AI Agent Goes Off the Rails: Give It Boring Work First
One developer logged token consumption across the first three days of an agent-driven project in July 2026 and got a curve that says almost everything I think is worth saying about this technology:…
Did this answer it? A star on the repository is the whole ask — it is what puts these in front of the next person looking; the data is CC BY and does not require starring.
Want a figure that is not here yet? Say which metric, which provider, which unit — one required field, and the page you came from is already filled in.
Got a better number? Open an issue — corrections and counter-data are the point.