United Healthcare’s plan in a nutshell
- Buy Optum
- Make it so bad everyone goes somewhere else (it was already there)
- Make reimbursements and billing so difficult other doctors stop taking UHC
- Hold on to the profits
I like asking Google AI silly questions. But I was not prepared for this. What does it mean that 6G will have built-in AI? Is it using string theory to resonate GPUs into existence?
I may have to see if I can patch this. It’s one of the biggest things that looks like it’s getting in my way of using sessions as long running placeholders for tasks.
@claudeai if I do find a fix, or at least a cause, what’s the best way to contribute it back?
In February the Claude Code engineers described their screen renderer like a game engine. Is that the reason Claude Code sessions never seem to be able to be swapped out?
Is there a way to stop this rendering loop when the session is idle? @claudeai
How much of the heap is getting touched on each render loop? How much intelligence is there to avoid duplicate rendering work?
It’s not about the CPU. It’s all about memory.
The heap I see is about 300 MB. If that 300 MB is touched every 16 ms it will never be evicted @claudeai
Apple error messages are the best. Here it says “an error”. In the Watch app it says nothing. In the watches software update screen it says nothing. Everything lets me try to update and fail with no error message. So good.
Github merge queues are so broken. Very often I set a PR to merge when it is ready and it never does. I then have to disable automerge on that PR and re-enable it and it instantly goes in the queue.
I've got a Github action that has been running for 23 minutes. It runs a 7 minute job that has finished. The UI says it's running. The API says it's running. But I can't cancel it because the UI also says it's not running.
Do I get billed for heisenstates?
.@AnthropicAI can you make Claude handle HUP signals? It should essentially just do a `/restart` on the session but only if it is idle and not asking a question.
Me: Change the word A to B
Claude: ...
Claude: ...
Minutes later
Claude: I changed A to B. It's fully mutation tested. I did red/green TDD. I'll be honest, one round of TDD got put into one commit. If you'd like we can do an interactive rebase on another branch to fix it.
Do the same health insurance company employees that dole out endless claim rejections get their claims rejected too? Do they just know that’s the game and they understand how to work the system better?
Has anyone else noticed that Claude now randomly includes Chinese characters in English language conversations? And they're not random. They're in place of the English word you'd expect to be there. So far I've seen it do it for "response" and "village".