i have been surprised by the number of students writing in to organize CS 153 study groups and meetups around the world it’s time for optimism!

4. To date, closed models tend to be more robust and generally useful than similarly scoring open models. Closed models have certain hard-to-measure qualities that are not well captured in current or past benchmarks. This will be key to enabling closed models to dominate in

Not bad for Forbes At least 10 of these companies are enduring, durable businesses long term if they simply execute on their current stated strategies

jensen’s biggest regret: he didn’t invest in anthropic/openai early because he didn’t internalize that foundation labs needed supplier-scale investment that vc's couldn’t provide. google and amazon understood this; nvidia didn’t. he says he won’t make that mistake again — hence

what dwarkesh got right that jensen dodged: jensen never really answered the mythos question. if a model can find thousands of zero-days across every os, and compute is an input to both training and deployment scale, then yes — marginal compute to an adversary has weapons-like

if you want the clearest mental model of what a frontier ai system actually looks like chips, compute, context, models, creatives, capital this is the best course out there. lectures are public. Anjney Midha One nugget from today's lecture release.... RL isn't generalizing the

every field that matured went through this transition ...medicine, physics, economics in its better moments. the move from "heretic" to "updated their priors" is the phase change. ai is mid-transition and it shows.

alignment can be faked. safety training doesn’t remove deception from llms... it teaches them to hide it better. the more you push, the more sophisticated the cover-up gets.

hardware portability in ai was a negotiating tactic. inference economics just ended it. your model and your chip are now one thing and that lock-in is the whole game.

closed labs have more compute, more researchers, and more secrets. open models should be falling behind. they’re not. that’s the most important surprise in ai right now.

absolutely wild! figma is getting crushed man

👌👌

I’m not saying this without evidence.

Meet Kimi K2.6: Advancing Open-Source Coding 🔹Open-source SOTA on HLE w/ tools (54.0), SWE-Bench Pro (58.6), SWE-bench Multilingual (76.7), BrowseComp (83.2), Toolathlon (50.0), Charxiv w/ python(86.7), Math Vision w/ python (93.2) What's new: 🔹Long-horizon coding - 4,000+

Labor limits scaling.

Hard to replicate.

The Jensen + Dwarkesh Patel podcast was fantastic. Jensen is someone who understood how ecosystems work and someone who understands real-world trade, policy and controls work. And in some deeper sense how AI will actually diffuse into the world. In this podcast, Dwarkesh came off

In Cowork, Claude can now build live artifacts: dashboards and trackers connected to your apps and files. Open one any time and it refreshes with current data.