dan

[dan @danvnx]

@ellie_huxtable @solid_js 2.0

[dan @danvnx]

@firefox hey guys can you pls fix the memory leaks? one of many examples: opening a PDF makes the GPU ram jump to 1GB+

[dan @danvnx]

@signulll Palm Springs is a good one!

[dan @danvnx]

@iykshani in the CLI you can already do it by customizing the statusline

[dan @danvnx]

@kvoulou I dump all into workflowy under a triage list, and tidy things up later- very minimal, just infinitely nested bullet points

[dan @danvnx]

@christianreber just let me opt out of the pension Ponzi scheme system

[dan @danvnx]

@flapprdotnet TIL miss Denmark is Mexican

[dan @danvnx]

@thsottiaux @stemonteduro @thsottiaux credits are for people bad at math - stacking another subscription is always cheaper

[dan @danvnx]

@stefan_meyer98 @schrep prospecting is the easy part, a decent (automated) GIS/EO pipeline can do that with a very low false positive rate

[dan @danvnx]

@JustJake fuzzy behavioral testing - simulate users trying to do whatever beyond the happy path

[dan @danvnx]

Claude Code CLI: can we pls get a setting to make it so that a /fork is quiet and the main agent is not automatically notified of all that happens there?

often a /btw reveals work that could be done in parallel with forks, and auto-reporting to main is annoying and self-defeating

[dan @danvnx]

@lydiahallie /low-priority, and wish it was active for all subscription types

[dan @danvnx]

@cremieuxrecueil curious if a level of analysis below countries there are (Western) happy bubbles in which fertility is okay

[dan @danvnx]

even assuming there’s a useful kernel in what he’s saying, I don’t like the Oppenheimer vibe

[Dario Amodei @DarioAmodei]

We Must Pace the Frontier: I’ve written a new essay on why the AI industry should slow down, with a three-part plan for doing so.

Anthropic is unilaterally committing to the first of these steps. We’ll provide third-party evaluators with permanent, employee-level access to our systems, so that they can verify adherence to our safety measures, report on incidents, and assess models’ alignment during training.

You can read the full post here: https://t.co/OGyPb7yaYt

[dan @danvnx]

@lydiahallie yes, that + deny hook so that subagents can't spawn other agents

Claude really likes levels-deep nested delegation, it's sooo wasteful

[dan @danvnx]

@CodeWithTamara please just get a Hetzner VPS with a Linux distro you tolerate + docker compose to wrap whatever your backend does + frontend served/cached from Cloudflare if you feel like it

AWS is just horrible

[dan @danvnx]

@thsottiaux even today Astra (high) is being very lazy (frequent hangs and neurotic checkpointing) and has sloppy execution - very different from early experiences - any thoughts as to why this might be the case?

[dan @danvnx]

@0xSero https://t.co/nUpwOIWa0q, done by hand in the pre-LLM era

still kinda spartan, will prob refresh it from the bottom up at some point

[dan @danvnx]

@Steve_Yegge I use Muxy and then use different terminals within each project folder - it's the most sane option I found