Newsletter
Every coding agent's scorecard, the day it lands.
We'll email you a confirmation link to finish subscribing. Unsubscribe any time.
What you get, and how often
A weekly issue, plus one the day a new model is graded. Grading runs start every Monday, so the weekly goes out once that week's runs are graded and published: the scorecards that moved, the models that joined or were retired, and what the week's numbers changed about which subscription is worth paying for.
When a provider ships a model we grade it that day and mail you the scorecard, rather than sitting on it until the following Monday. Those release issues are short and they say what is provisional: a day-one card is usually one suite deep and the quota figure needs a full window to settle, so a number can move by the weekly. We tell you when it does.
Expect four or five issues in a quiet month. Everything is on News andRSS the moment it publishes if you would rather read it there.
Recent posts
- GPT-6.1 Sol is good, just two weeks too late ·
- Claude Sonnet 5.5 is a Sonnet done right ·
- Claude pulls further ahead as GPT-6 agents climb - Week 40 ·
- Grok 4.7 looks like Grok 4.6 with the effort dial moved ·
- GPT-6 Sol is a step back from 5.6, Luna stands still ·
- Claude Opus 5.5 takes the top at a fraction of Fable's usage ·