@alpha_convert

Aspiring Abyss Domain Expert

 On The Computer
Joined April 2014
  • Tweets16.2K
  • Following2.1K
  • Followers4.3K
1.3KMedia
At the PyTorch conference today and tomorrow, come and say hi! We have lots of swag!
11
10
411
16,864
sent myself a text and then got excited when my phone vibrated. 😭😪
105
3,020
191
18,836
After 5 months, 385 citations, and 50 pages I’m excited to finally share the preprint of the survey paper my collaborators and I have been working on: 

 sainati.pl/gpu-survey-prepri… We wanted to put it on arXiv but tragically they don't take survey papers anymore.
2
7
1
32
2,233
Joe retweeted
joomy@joomySep 23
> i ask Claude if the program is verified > he says yes > i ask if he verified the program or a smaller, better-behaved program he imagined > he laughs and says “it’s verified sir” > i use the program > it has a bug
Boris Cherny
@bchernySep 22
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)?
16
24
584
26,284
100 problems is nothing, in programming languages we routinely solve exactly the next 700 problems
1
4
56
2,486
humans, a bipedal primate notable for early contributions to matrix multiplication
75
920
39
14,465
311,934
Joe retweeted
roon
@tszzlAug 30
there are some number of bad abstractions in anthropomorphizing ai intents but there are at this point more dangers from avoiding anthropomorphism at all costs. if you have a mental picture of guys living in computers, it’ll likely prepare you for the future better than otherwise
271
250
144
3,524
384,006
They call it amtrak because the train am on the track
150
4,746
96
56,542
2,264,178
Up first at NYC Systems, Manya Bansal on Modular GPU Programming with Typed Perspectives
1
6
207
8,097
Dishonored is one of the best video games of all time. This is because it had a graphics option called “Rat Shadows on/off”
1
9
765
First batch of invites out for August 20th for NYC Systems! Featuring: ​Manya Bansal of MIT on "Modular GPU Programming" and Sugu Sougoumarane of Supabase on "Multigres: Vitess for Postgres" Get on the list! luma.com/6mnm99ne

NYC Systems August 2026 · Luma

Manya Bansal of MIT on "Modular GPU Programming with Typed Perspectives" Sugu Sougoumarane of Supabase on "Multigres: Vitess for Postgres" Plus a lightning…

luma.com
1
3
13
1,693
Why don’t they make the whole database out of indexes
88
26
23
1,207
586,034
Caught myself briefly thinking that the goose had no value (0?) hah
7
3
143
15,303
This bool could've been an enum
79
69
24
2,050
137,165
At PLDI currently! So excited to share that our paper on modular gpu programming with @sainati_ @alpha_convert got a distinguished paper award (dl.acm.org/doi/10.1145/38082…)! come say hi if you're around :)

Modular GPU Programming with Typed Perspectives | Proceedings of the ACM on Programming Languages

To achieve peak performance on modern GPUs, one must balance two frames of mind: issuing instructions to individual threads to control their behavior, while simultaneously tracking the convergence of...

dl.acm.org
2
5
11
1,640
I've been telling people for 25 years that Jane Street is not interested in formal methods. No more! And we're actively hiring to form a new formal methods team!
30
75
14
1,068
127,072
Joe retweeted
roon
@tszzlJun 2
the frontier labs don’t have “comms problems”. reality right now has a comms problem. what is happening is a little scary and there’s no nice words anyone could say, especially not those profiting from it, that’ll make it feel that much better
193
240
84
4,030
405,019
Replying to @lion_tender
What’s HPMOR?
42
19
14
307
62,778
New blog post: On the Unreasonable Effectiveness of Property-Based Testing for Validating Formal Specifications. proofsandintuitions.net/2026… The gist: randomised testing can validate formal specs. It's very cheap and powerful: we found bugs in specs of VERINA and CLEVER benchmarks.

On the Unreasonable Effectiveness of Property-Based Testing for Validating Formal Specifications

In this post, we show that property-based testing (PBT) is surprisingly effective for validating LLM-synthesised specifications of Lean programs: it is a cheap alternative to symbolic proofs, which...

proofsandintuitions.net
2
24
97
7,144
We gave @dwarkesh_sp a tour of one of our new GPU-filled data-centers. Much fun! youtube.com/watch?v=8J-GUnfS…

Dwarkesh Goes Inside Jane Street's Latest AI Data Center

Twenty years ago, our "cluster" was just 6 Dells stacked on the flo...

youtube.com
24
83
26
1,014
617,429
Load more