Some bugs are just funny as hell #wowforever

Looking forward to later!

Gave GPT-6 Astra about 5 hours of end-to-end work yesterday. I used only 20% of my usage (flipping between xhigh for the plan and medium for work) and built a full feature for a platform we're shipping. It doesn't handle ambiguity quite as well as Fable, but it is great value.

If your new fancy model can't work for 8+ hours in one go on a long-running task... I'm not interested. Use Case: I like to wake up knowing it's worked while I sleep, not just stopped randomly.

You just know the current @github outage is bad when they clearly hand-wrote that latest status update about Vitess. /cc @QuinnyPig

Today, in 'stupid s**t' a financial company does. Last year I transferred out of an ISA with @Vanguard_UK ... okay, cool... and now I'd like to re-open a new ISA with them. BZZT Nope. 'Open a new account with another email address' - you what!? Guess not so 'cyber safe' there.

QOTD from Codex (GPT-5.6-Sol) - "I'll stop maligning Claude's perfectly normal tea-break review time and keep waiting" - I almost spat out my coffee - loved the humour. /cc @thsottiaux

Yo @bcherny can we please do something about Claude Reviewer not understanding perfectly valid Python 3.14 Exception Formatting - pretty please? Burning tokens for something that is 100% allowed and valid Python is unfun. See PEP 758.

You know, I'm sure @AnthropicAI love to keep telling people how their models are great for long-running tasks... but their end points... not so much cc @bcherny / @ClaudeDevs

I genuinely don't miss Windows for gaming.

I guess we just don't bother to check the copy in trailers now? cc @PlayStation

Curious question - but who at @incident_io thought this was inclusive/accessible (hint: it's not, and jilted text is even harder for people with dyslexia to read).

Would be great if @claudeai had as many integrations with @gitlab as it does with @github. Like really great. Wouldn’t have to get Claude to make quite so many yaml files to come close then.