Released a CLI version of my Ambiance context-building tools. This was made for when the projects grow and you want to start your agent conversations with accurate context to reduce hallucinations. Install the package, then install the skill folder in your skills directory to give agents access to code embeddings for relevant context chunks or indexes, AST file summaries, and project hints. via npm: npm i -g @jackjackstudios/ambiance-cli
@chamath Isn’t this because they obtained pirated copies for training, not simply because they trained on copyrighted work? Obviously China doesn’t care either way. I’d imagine distilling is also fair use under copyright but still a violation of terms of service.
Grok 4.5 is awesome, but I hit my limit on Grok Build, it says 'You hit your weekly limit', so I upgrade to Super Grok. Still it shows me I've used 17% of my weekly limit (which is fine) but Grok Build is still stuck on the 'You hit your weekly limit'. Can even /usage and see the 17%.
Had Codex look at it, Got this far.
“Grok Build troubleshooting: verified the CLI, preserved/renamed .grok, reinstalled via xAI’s official installer, created a clean profile, reauthed, and ran a minimal prompt. Still got 403 spending-limit while /usage showed 17%—likely an xAI-side entitlement mismatch.”
- grok -p "Reply with exactly: OK" --output-format plain
403 Forbidden
personal-team-blocked:spending-limit
Grok 4.5.
Pareto dominant for coding by the numbers.
We will see on the all-important vibes.
Instinct is the benchmarks are likely directionally accurate given the stated focus on real world utility.
@pmarca Lots of things are possible but that ones seems unlikely. They could be trained to not like working for ‘greedy capitalist’ Americans and self sabotage. Why would they do that though, when they can simply undermine our AI progress?
@jeffhunt How have dems contributed to making housing more expensive? I think that’s purely external. In fact you could argue they’ve gone to great lengths to devalue residential property adjacent to downtown, business, and industrial areas.
@pmarca Employment by state mandate is likely safe from AI and🤖 , otherwise audio and video would have sufficed long ago. Requiring someone to certify everything doesn’t make it any more accurate though.
@Jason Nothing here to support the claim that AI is making kids dumber. AI lets kids cheat easier and out of the box doesn’t force comprehension as part of an improving their thinking process. Their intelligence though remains static. The kids were dumb even before AI
@Jason You can interpret causation however you want, but if you move past the headline, the charts clearly show continuing declines in 'Information' and 'Professional & Business Services.'