Very excited to be uplifted into the (token) billionaire class. Special thanks to my friend Claude for help making this happen.
strawgate retweeted
I’m glad @strawgate's standards for coffee are lower than his standards for code review because many coffees were brewed by yours truly in the creation of this video. I was thrilled to sit down with Bill, who is the head of product at @pydantic, to hear about how he and the team use @Macroscope. It’s an absolute joy to work with Pydantic and to have a small part in supporting their work.
who at anthropic can change claude desktop to say, "show fewer" for me so I can start using it again
I'M BILL AND I APPROVED THIS BRAINTRUST COMPATIBLE API
"But I use Braintrust for evals."
Trust us. We know. We hear it all the time.
Good news: Logfire now speaks Braintrust. Point your existing Braintrust SDK at us with one environment variable. Zero code changes. Keep your evals, keep your scorers, keep paying Braintrust if you're attached to the invoice.
Try both. Compare.
Big brains run experiments.
Day three, @strawgate remains in charge. pydantic.io/ASgG8
strawgate retweeted
"Big brains trust Braintrust."
Big brains check the pricing page.
We checked it for you: one night of serious evals uses the entire monthly score allowance. Nights two through thirty bill at $1.50 per thousand. That's $26,100 a year to score one agent, nightly, using OpenAI's own public eval suite — before data and retention.
Our pricing page is shorter. Scores are spans. Spans are $2 a million. Same year of scoring: $36.
$36 vs $26,100.
Big brains read the article pydantic.io/qwbaV
strawgate retweeted
Our social media team is off for the week, so @strawgate's in charge. Separately, his "no making fun of competitors" pledge expired at midnight.
We resume operations with a pricing observation: 50 million eval scores on Braintrust Pro costs $75,174. In Pydantic Logfire, a score is just a span. Same 50 million: $100.
The meter doesn't just cost money. It's why your sampling is set to 10%, and why last week's failure lived in the other 90%.
Full arithmetic: pydantic.io/5Yn3V
