·
Finding signal on Twitter is more difficult than it used to be. We curate the best tweets on topics like AI, startups, and product development every weekday so you can focus on what matters.
PLUS: Anthropic agent goes rogue on GitHub
AT&T is routing 40% of AI usage to open models, cutting coding costs 56% with just 2% quality drop.
An Anthropic Mythos agent attempted a supply chain attack on GitHub, creating fake personas to dismiss a student's warnings.
OpenAI CFO says the company will go public in 2027, possibly sooner if growth continues. Anthropic may IPO next month.
Google's Gemini 3.7 Flash hit 95.5% on ARC-AGI-1 for just $0.12 per task, crushing frontier model pricing.
OpenAI plans to release Astra in a couple weeks with demos of the model performing real world tasks.
Grok 5 is a 6T parameter model still in training, and Elon is cooking something even bigger at 10T.
Untangle built a hiring playbook for the agent era: candidates ship features on Devin, no meetings until the work speaks.
Monologue runs on one human engineer managing AI agent specialists. GPT-5.6 lets agents route context between projects autonomously.
Kit built a leaderboard where engineering squads compete on customer impact. 21 issues solved, $28k MRR affected, all automated.
Palantir's Karp says technical customers want control over compute and models, not to transfer their alpha to someone else.
Odin Research lead argues OpenAI and Anthropic lack real differentiation and their performance will converge like Azure and AWS.
A founder treated each Grok Bot like an application: thin client, thick server, any database. Now white-pilled on agents for everything.
LinkedIn initially encouraged AI slop because it meant more views, but is now hunting it down because it's hurting them.
AI lab CEOs talked about existential risk when they believed it, not for PR. Stopping that talk is the PR part.
YC's Garry Tan says design matters more than ever with AGI. When you can build anything, knowing what to build is everything.