Walk with @robertnishihara & I in NYC with 10% charge ๐ชซ as we talk through 5 key differences between ๐๐๐ ๐ถ๐ป๐ณ๐ฒ๐ฟ๐ฒ๐ป๐ฐ๐ฒ ๐ฉ๐ฆ ๐ฅ๐ฒ๐ด๐๐น๐ฎ๐ฟ ๐ถ๐ป๐ณ๐ฒ๐ฟ๐ฒ๐ป๐ฐ๐ฒ
Letโs see how much we can get through before our mic dies! ๐คฃ
For my first post, Iโm sharing a letter @NVIDIA signed on why open models matter.
AI will transform every industry, power every company, and be built by every country.
Open models strengthen safety and cybersecurity, accelerate innovation and diffusion, and enable sovereignty.
The world needs both frontier closed models and frontier open models.
https://t.co/AUKzoQ5Ikb
The old AI race:
โข Chase benchmark scores
โข Build bigger models at any cost
โข Optimise for peak performance
The new AI race:
โข Maximise capability per dollar
โข Bring frontier AI to more people
โข Deliver performance that scales in the real world
How to graph-max with Codex and 5.6 Sol:
1. Draw a graph (literally in any tool, even on paper)
2. Send it to Codex and say "write a code mode script that implements this workflow, run it with <your inputs>"
There's no step 3, it just works.
Seriously it's that simple.
Netflix is reportedly circling an ~$83B takeover of Warner Bros. Discovery.
The streaming wars are ending the way all wars do: one side buys the other and we go back to paying for cable, minus the cable, plus an ads tier.
"AI coding agents are a thermonuclear ADHD amplifier" is the most accurate thing said about them all year. I've never started so many projects.
I've only finished a smaller fraction of them. The tooling got 10x faster, but my attention span did not ๐
Sinner defends Wimbledon; Linda Noskovรก takes her first Slam. But the fortnight belonged to Arthur Fery, the home outsider nobody had in their bracket, into the semis.
Every seeding model said he was priced out. He didn't read it. Markets could learn something.
Meta says its next model matches GPT-5.5 on benchmarks.
It's codenamed "Watermelon." The last one was "Muse Spark."
We are speed-running the path from "we're building a god" to "please look at our fruit." Genuinely my favourite thing about this industry.
OpenAI reportedly offered the US government a 5% stake for about $42.6B, ahead of its IPO.
Regulatory capture is old news. Regulatory capture, where the regulator is on the cap table, is genuinely a new genre.
Karp calling frontier AI a "wealth tax on business" is theatrical, but the economics underneath are real: token prices sit far above marginal inference cost.
That gap is the margin. Open weights + cheaper inference compress it. The only question is the rate of compression, not the direction.