NullifyNAN

LLM performance increases substantially when the exact software you’re writing is already in the training data.

I use coding agents to fill out dunkin donuts surveys for me

Yet it took Codex 8 months to get subagents after their introduction in claude code. This of course pales in comparison to having to symlink CLAUDE.MD to AGENTS.MD

Thank you for old vs new comparison

It seems like you’re so close but yet so far. It’s interesting that you concluded that Claude makes it easier to run tasks locally but did not conclude that claude makes it easier to use the cluster. Feels we’re in need of a part 2 where you use CC to automate all of the mundane cluster work.

Also just a hot tip, if you’re able to keep any always-only PC on a university subnet (your PI should have some form of a server) you can use tailscale as a jump proxy and bypass the whole VPN dance.

Yeah train wifi is generally terrible. Basically the only real solution is 100% 5G coverage along all train tracks or starlink. This is basically a global issue because line-of-sight for trains is a hard problem.

OpenAI has answered your prayers.

16 hours ago the readme for codex CLI was updated. Now codex cli supports openai login like claude does, no API credits.

From the readme:

After you run codex select Sign in with ChatGPT. You'll need a Plus, Pro, or Team ChatGPT account, and will get access to our latest models, including gpt-5, at no extra cost to your plan. (Enterprise is coming soon.)

Important: If you've used the Codex CLI before, you'll need to follow these steps to migrate from usage-based billing with your API key:

Update the CLI with codex update and ensure codex --version is greater than 0.13 Ensure that there is no OPENAI_API_KEY environment variable set. (Check that env | grep 'OPENAI_API_KEY' returns empty) Run codex login again

You’re not paying for a service, you’re bidding in an open market. They don’t tell you this but it’s the reality.

Drivers can tell if you don’t tip and all of the experienced ones will decline your order.

Though these apps have done a lot of work to conceal the amount the driver actually gets until delivery is completed.

Yeah but most fall apart at lower context than advertised. They do great at simple stuff like needle in a haystack tests but totally flop when you actually try and use that context for something productive.

100% true, it’s basically completely useless.

Yes because ultimately the goal of capacitor research is to make it more similar to a battery in terms of storage.

That's a completely different architecture than current lipid nanoparticle mRNA vaccines. It would be obvious to anyone with even a modicum of knowledge that the linked paper is not relevant.