simon

Founding MTS @Axiommathai, 🚗 @drivingbench, prev PyTorch/MTIA @meta, @uchicago

[simon @nautsimon_]

living through what is undoubtably humankind's second renaissance

[simon @nautsimon_]

@bogu_gireesh @a_ramabadran @tobiges @comma_ai Lmfao

[simon @nautsimon_]

Last week @a_ramabadran, @tobiges, and I hooked up Codex (Astra) to a rented Toyota Corolla + a @comma_ai and had it drive us around a parking lot. We tried a few others (Fable, Grok) but Astra was the clear winner, and the only one to complete our course. Harness is open sourced for anyone else that wants to have codex drive them around ;)

[simon @nautsimon_]

Verification is the future

[Boris Cherny @bcherny]

I used Opus 5.5 to formally verify the Claude Agent SDK using Lean. A couple short prompts = 16 PRs fixing various bugs and race conditions. Video attached.

TLA+ also works well. I sometimes combine Lean and TLA+ to look for issues around data flow, concurrency, and state mgmt.

I don't know either language well, but Claude is excellent at both. This approach is super useful for formally modeling your code and finding bugs that a human probably wouldn't have spotted.

Is formal verification the future of coding (or at least, bug finding)?

[simon @nautsimon_]

@tobiges @a_ramabadran yum

[simon @nautsimon_]

@biosemiote @a_ramabadran @tobiges @comma_ai we tried haha
https://t.co/zAJIfAQbtS

[Aditya Ramabadran @a_ramabadran]

Trying to evaluate Jev! But it’s not easy since Jev doesn’t natively take vision inputs (so the demos you’ve seen of it playing games etc required giving whole state… which we can’t do.) Trying to build out a pipeline that utilizes it and its speed well, pairing it with Luna for perception…

[simon @nautsimon_]

@a_ramabadran @tobiges @comma_ai To see all the traces and results, check out: https://t.co/zHKf2gT84b

code: https://t.co/vWEwavUi9l

[simon @nautsimon_]

Car

[DrivingBench @DrivingBench]

We gave ChatGPT, Claude, and Grok control of a real Toyota Corolla 🚗, steering/gas/brakes, no human driving (just a foot over the brake).

Only one model was able to complete our entire driving course. Introducing DrivingBench. 🔥⌛️🏁

[simon @nautsimon_]

@deancureton

[simon @nautsimon_]

@AIslop_ @CryptoMaxPlanck hol up???

[simon @nautsimon_]

Been thinking about the great sloppening and impending entropy blow up of scientific publishing. Feels like the only way to combat the scaling of sci-slop is by scaling automated reproducibility, centralizing it, and making it more of a first class aspect of the reviewing process... of course, someone's gonna have to pay for the tokens, but maybe theres data one could farm to subsidize the cost. who's building this?

[simon @nautsimon_]

@gpusteve hold on

[simon @nautsimon_]

@gpusteve @CarinaLHong @potterlab We are building

[simon @nautsimon_]

@wafer_ai

[simon @nautsimon_]

@yacineMTB should i ropemax

[simon @nautsimon_]

you either die a coper or live long enough to see yourself become a larper

[Axiom @axiommathai]

We reward collaborative agency and whole-problem ambition. There is no fluff, and it is not chill, as we are busy bringing the future forward.

If this sounds like you, join us.

One title, full stack. No manager. No recruiter.

New website; apply today: https://t.co/Kh1trztcEL

[simon @nautsimon_]

how much impact do you think a "tibo reset ™️" has on the gdp

[simon @nautsimon_]

@CarinaLHong

[simon @nautsimon_]

@identity_matrix

[simon @nautsimon_]

hard problems create cracked folks
cracked folks create high aura
high aura creates larpers
larpers create hard problems

[simon @nautsimon_]

@MqwantyMqwanty investing in MqwantyMqwanty