Scale AI: Meta acquired 49% at a $30 billion valuation (roughly $15 billion paid) via a minority-stake structure explicitly designed to avoid regulatory scrutiny.

“Meta has done this interesting structure deal. They're buying 49% of the company. they're paying a $30 billion valuation. So, they're b paying effectively 15 billion. They're avoiding regulatory scrutiny.”

Meta instructed teams to stop using Codex and Claude Code on certain tasks out of concern that model outputs could contaminate its own training data, revealing an internal tension in AI-assisted development at frontier labs.

“One memo told teams to discontinue the use of Codex and Claude code on certain tasks for fear that model outputs could contaminate training data.”

Interval app's cost per trial start on Meta is approximately $12.

“The cost per trial start for us currently is about $12 on Meta.”

Microsoft, Meta, and Google are each on track to spend over 50% of their revenue on capex in 2026.

“Microsoft, Meta and Google are all on in line to spend over 50% of revenue on capex this year.”

Altria has outperformed Meta, Microsoft, and Amazon over the last five years despite US cigarette volumes declining 7 to 8% annually.

“Cigarette volume has been declining at 78% in US and guess for last 5 years Altria has outperformed meta I think now Microsoft and Amazon last 5 years.”

Meta Llama restricts free commercial use once a developer's user base exceeds 700 million users.

“Some of the players, most notably Meta with Llama, have a usage restriction against the free use of the model at 700 million and that's what you were referring to. And so at least in a tweet, Sam suggested they won't have that in theirs.”

Suppressing roleplaying and deception features in Llama 3.3 70B simultaneously improves TruthfulQA scores AND increases the model's self-reports of having subjective experience.

“The paper six months ago where they showed that when you suppress roleplaying and deception features in that was done on Llama 3.370B which is 2 years old already was 18 months old already when they did the When you do that suppression of those role playinging and deception features, the model becomes more truthful as measured by the truthful QA benchmark. And then it also becomes more likely to say that it has subjective experience.”

Marc Andreessen on the pace of revenue growth at the two leading AI model companies.

“Basically Anthropic and OpenAI are adding more revenue per month than Meta Google or Microsoft.”

Despite aggressive poaching campaigns by Meta and others, Anthropic lost only two technical employees while other labs lost dozens, suggesting mission-driven retention outperforms compensation competition.

“When you know Meta and others were out you know with these huge packages for some of the technical talent in across the large language labs I think we lost two people and other labs lost dozens.”

Meta Llama 4 is reported as a 400B parameter mixture-of-experts model with 50-70 experts and a 10 million token context window, targeting April 2025 launch.

“It'll be a 400 billion parameter model. It'll use like it'll be a mixture of expert model using you know 50 60 70 experts. It's going to have huge context window like 10 million context window you know and it's going to launch this month.”