2025 Thoughts
✦ November
We Are Not Losing Our Voice to AI
People are scared that AI is making the internet flat and boring. They worry it's creating a sea
2025 Thoughts
✦ November
Hot Takes from Ilya's Interview
Ilya Sutskever, on Dwarkesh Podcast:
But when people do RL training, they do need to think. They say, “Okay, we
2025 Thoughts
✦ October
The Art of the Productive Failure
I'm starting to believe the most valuable runs are the ones that fail catastrophically. The runs where the
2025 Thoughts
✦ October
The Frontend Verification Gap for AI Agents
Agentic coding loops are a marvel for backend work. An LLM can write some code, run a test to verify
2025 Thoughts
✦ September
Learnings from Cursor Tab
Cursor's "Tab RL" blog:
To use policy gradient methods to improve Tab, we defined a reward
2025 Thoughts
✦ September
Confidence != Wisdom
The single most dangerous design flaw in today's language models isn't that they hallucinate; it'
2025 Thoughts
✦ September
LLM's Data
We obsess over model architecture and FLOPS, but the most important component of any new language model is the one
2025 Thoughts
✦ August
The "AI Button"
The new trend of every app getting an "AI button" feels like a solution in search of a
2025 Thoughts
✦ August
The Browser Window
There seems to be a strange desperation in the air with AI companies trying to buy web browsers. It’s
2025 Thoughts
✦ July
Benchmark Matters
A recent clarity I’ve had is that the most consequential design document in modern AI research isn’t the
2025 Thoughts
✦ June
Type 1 / 2 Hard Problems
An underdeveloped meta-skill in AI research is distinguishing between "Type 1 Hard" and "Type 2 Hard"
2025 Thoughts
✦ May
Keynote / Experiments
It's that time of year again—keynote season. We're about to see a firehose of new