Congrats to Karpathy. I wonder whether this is the right time to join Anthropic. Looks like it from the outside. But - unpopular opinion - I believe Anthropic is one open-source model away (that can code well) from a massive revenue/stock crash. We're already seeing Claude's cost escalate to astronomical levels. Most coding work is medium difficulty in the grand scheme of things. So the future is an open source model…
those models are light years away from opus or sonnett right now though -- is the context problem really solvable?
FWIW while Karpathy was at Tesla he was basically working on the vision component. The actual driving component (using vision as an input) was originally all C++. They may have started migrating parts of the driving component from C++ to neural networks while Karpathy was there, but most of it happened after he left in 2022, with the big switch being FSD 12 in 2024. User reports from before/after FSD 12 are like nigh…
I was never convinced by the "vision only" approach - I don't see the point of throwing out or refusing to have additional data from other sensors. I suppose that with modern ML they can just toss it in the blender and reap the benefits ...
I'm not talking about vision (cameras) vs lidar etc, just the Tesla FSD architecture that separates the "vision" component (turning camera/sensor inputs into symbolic road/sign/vehicle/pedestrian/etc data), and the driving component which takes the vision data, plus current location and destination, and uses that to actually drive the car - switch lanes, make turns, avoid obstacles etc etc.
Congrats to Karpathy. I wonder whether this is the right time to join Anthropic. Looks like it from the outside. But - unpopular opinion - I believe Anthropic is one open-source model away (that can code well) from a massive revenue/stock crash. We're already seeing Claude's cost escalate to astronomical levels. Most coding work is medium difficulty in the grand scheme of things. So the future is an open source model…
GLM 5.1 is almost there. These guys should be scared. The valuations these companies have is insanity.
Not in extended sessions, I've noticed. It's good at targeted edits, but not "build a small tool that XYZ".
Congrats to Karpathy. I wonder whether this is the right time to join Anthropic. Looks like it from the outside. But - unpopular opinion - I believe Anthropic is one open-source model away (that can code well) from a massive revenue/stock crash. We're already seeing Claude's cost escalate to astronomical levels. Most coding work is medium difficulty in the grand scheme of things. So the future is an open source model…
Model diversity is really their weak point. OpenAI has embeddings, audio, image, video (RIP). Anthropic has ... Claude. It's a great model for a lot of things, but it's super risky to just have one thing you're good at (from a business standpoint).
1. Bun is right there on Github. You can download can use it right now. 2. Sure, that's one thing. 3. Coefficient Bio is not a thing. They don't have a product. Ever. It's just Anthropic hired 10 people for a ridiculous amount of premium bonus. (Time will prove it's a bad decision, btw) 4. (snorts)
I believe the Bun reference is the Rust "rewrite" that came up recently.
Congrats Andrej. Let me know if you are looking for someone to take over Eureka Labs. We were in the same WoW classic guild. domrdy on x. We can duel for it :). Also, I need $10M if any VCs are reading this.
I have been impressed by some of his work, especially on the vulgarisation and simplification. Excellent communicator and engineer. But I am a bit more skeptical about his taste and vision. Leaving OpenAI to work for Elon Musk was a poor move, and AFAIK his work on CV at Tesla did not bring anything groundbreaking, unfortunately probably the opposite (the bet on camera-only driven system did not pay off) and his talk…
> the bet on camera-only driven system did not pay off This is a pretty unsubstantiated claim. Tesla is now launching robotaxis at a fraction of the cost of Waymos, in part because they don't need all the Lidar.
Let's say it is an opinion.
But Tesla has been promising full self-driving "next year" for quite a long time now, and it seems they are stuck at the "95% there" stage basically forever.
Anyone want to comment on what it's like to work for Anthropic (as an ordinary software/AI engineer, not as Karpathy)? Compare and contrast with working at OpenAI, Google, etc.?
Congrats to Karpathy. I wonder whether this is the right time to join Anthropic. Looks like it from the outside. But - unpopular opinion - I believe Anthropic is one open-source model away (that can code well) from a massive revenue/stock crash. We're already seeing Claude's cost escalate to astronomical levels. Most coding work is medium difficulty in the grand scheme of things. So the future is an open source model…
I heavily weight the explosive revenue growth of Anthropic and OpenAI above speculation about what open source models may do in the future. I've heard for over 6 months that there's no moat but the revenue growth keeps proving it wrong. Opinions have to adjust to meet reality. There's some kind of moat, for now at least, that is not being appreciated in the conventional wisdom.
(If they were just burning Capex and nobody wanted to use their product or their gross margins were bad then I'd agree with you)
> just becomes a glorified marketer That implies Karpathy is either dumb or desperate and he is neither of those by a long shot.
He already stated his motivation a few months ago in an interview with Dwarkesh - basically saying that he might join one of the big labs, for a while, to keep in touch with frontier research. Andrej seems like a great guy, but him joining Anthropic feels a bit like a transactional relationship (rich old guy marries hot young chick). Anthropic get a "glorified marketer", and he gets a front row seat at SOTA LLM dev 2…
An employment relationship is transactional??? Like the employer pays money and the employee provides labor???