Agentic AI in the wild. Construction, credit, and any shop still held together with CSVs. NYC.
New York, USA
Joined June 2012
- Tweets5K
- Following789
- Followers615
- Likes20.2K
Paul Garcia retweeted
Honestly, shame on the CIA for not getting Jensen a patriotic state issued baddie.
musk says that in a world of AI abundance there's no need to save money.
maybe. nobody knows the timing, so i'm not betting rent on it.
but he also said the real question would be meaning. that's the part worth sitting with.
if you never had to work again, what would you do on a tuesday?
the quotes, in his words: "no need to save money" (dec 2025) and "the question will really be one of meaning" (may 2024). coverage: foxbusiness.com/economy/musk…
Paul Garcia retweeted
Jane Street engineer:
"95% of Jane Street Quants are running swarms of AI agents
We build JaneGPT in 3 days - our own AI we use for trading and research to make $39.6B/year"
In 1-hour Jane Street team explained how they uses AI for their fund - and how they build their own AI that generates profit
5:14 → how AI lives inside Jane Street's
20:22 → AI agents working on Jane Street's big data research
45:26 → why trading firm builds its own data-centers
1:10:05 → AI that watches live metrics and helps when things break
Learn how Jane Streaet use AI - no matter what, this is the best AI + trading conversation out there:
Support, finance, legal, ops: each department rebuilds the same harness in a different room. The case studies are mostly vendor numbers, so trust the pattern, verify the percentages. Ep 8 of Harness, from zero is live. #AI #EnterpriseAI #AIAgents #LLM
Support, finance, legal, ops: each department rebuilds the same harness in a different room. The case studies are mostly vendor numbers, so trust the pattern, verify the percentages. Ep 8 of Harness, from zero is live. #AI #EnterpriseAI #AIAgents #LLM
Paul Garcia retweeted
Starting today, eligible ChatGPT Pro and Business subscribers can claim up to $500 per month in API credits to build apps and agents.
Context rot: across 18 frontier models, performance sagged roughly 113,000 tokens into million-token windows, while about 300 tokens of focused context stayed reliable. Ep 5 of Harness, from zero is live. #AI #LLM #AIAgents #ContextWindow
Context rot: across 18 frontier models, performance sagged roughly 113,000 tokens into million-token windows, while about 300 tokens of focused context stayed reliable. Ep 5 of Harness, from zero is live. #AI #LLM #AIAgents #ContextWindow
Sometimes it's not the model. It's an expired key.
Keys, quotas, balances: alert before they expire, budget them, health-check each provider, and give the whole layer one named owner.