sensible: "makes the model speak less like a polite chatbot & more like a terse tool … Same substance, fewer words. In my evals, Caveman cut output tokens by roughly 65–75% versus default verbose output, & still beat a normal ‘be concise’ instruction" https://t.co/HCDF…
Companies Are Making Claude and Codex Talk Like Cavemen to Stop AI’s Soaring Costs 404media.co
AI Weekly's analysis
→
- A plugin called caveman, written by Julius Brussee in early April, strips verbose model output and cut tokens by roughly 65 to 75 percent in his tests.
- Shayne Sweeney, OpenAI's director of engineering, contributed code to caveman to support Codex, and developers at Nvidia and GitHub are reportedly using it.
- GitHub shifted to per-token billing in April, Uber blew through its entire AI budget in four months, and Legrand's internal memo points staff at caveman.
Read full analysis →
If rewarding AI mathematicians by the same metric that we have rewarded humans ones make math go badly, then we are probably rewarding the human ones badly. With good reward metrics, we shouldn't care who wins them. https://t.co/CO6Z9o3mcW
A Severe Misalignment of AI in Mathematics terrytao.wordpress.com
“Synthetic intimacy creates what economists call negative externalities — the hidden costs that fall not on the person making the choice but on the relationships that are changed, ever so slightly, around you.” https://t.co/l8gaIXNuiK
nytimes.com
Should you believe advocates for X when they announce: “breaking news: the vibe has shifted, everyone agrees with us on X now”? https://t.co/EjSq4PWQiN
nytimes.com
I can't use ChatGPT Astra if I log into my $20/mo. ChatGPT Plus account, but can use it via https://t.co/ppHN1buJ0I available free to all.
Ask-Chat.ai — the best AI models in one chat ask-chat.ai
Huh, a LOT of math problems have substantial chances of being solved soon: https://t.co/XVvWNVQLr3
Which math problems will AI solve in 2026? polymarket.com
@Nox7hhsjj https://t.co/h2SdzKTp9C
overcomingbias.com
@adamvsteele Global regulation looks even less workable. https://t.co/CQakS89V18
overcomingbias.com
@NathanpmYoung @mattparlmer Sure liability is a well-established and clearly limited way to deal with potential externalities. https://t.co/KpPhGvIRRg
overcomingbias.com
@computablesoul @robertskmiles https://t.co/y3pkfMWLH4
AIs Will Be Our Mind Children quillette.com
On 13% chance by 2027 of US Congress "Restrictions on AI models above a certain size or amount of compute". Higher chances of other kinds of regs: https://t.co/6ftAjDkrSx
What AI regulations will the US Congress pass into law before January 3, 2027? [ADD ANSWERS] manifold.markets
A solution to cultural drift: human level AI which people neither love nor hate too much. https://t.co/C0vnNjiIzB
overcomingbias.com
