tushar (@gpu004) header
tushar (@gpu004) profile picture

making compilers grin, building distributed systems, previous @ihc__official, research @iiit_hyderabad

Brooklyn, NY
Joined July 2021
Photo by tushar (@gpu004)Photo by tushar (@gpu004)Photo by tushar (@gpu004)Photo by tushar (@gpu004)Photo by tushar (@gpu004)Photo by tushar (@gpu004)Photo by tushar (@gpu004)Photo by tushar (@gpu004)Photo by tushar (@gpu004)Photo by tushar (@gpu004)
Pinned Tweet
lately i have doing small implementation of casandra paxos lwts, tigerbeetle viewstamped replication, cockroachdb raft per replica group playing around with different consensus ideas
1
9
1,722
thanks u so much @yorios9312 for the tokens, time to write leans proves
1
2
pretty interesting thing, i am the 67 guy who started both ephyra and hegel @Keleesssss @DRMacIver
1
6
278
watch till the end
2
97
Replying to @dillon_mulroy
I've been using it in repos with huge number of tools (like 500+ of skills in .agents/skills) to cut down tool discovery based on current task At 500+ skills, just skill name+desc eats into context a lot. github.com/championswimmer/p…
4
2
84
5,878
tushar retweeted
135
821
65
12,329
245,210
Formal verification is the Rust of software correctness. Most do not understand it, but just as people started shipping Rust because LLMs made it easier to write, we are now writing formal specifications because LLMs can generate them. The problem is, if you do not understand what you are writing, you can never be sure that it is "correct".
50
27
4
452
31,262
tushar retweeted
realizing all the hard shit you accomplished on your resume pre 2026 is just a prompt now
103
541
68
9,133
219,518
tushar retweeted
People really say “I don't trust code generated by an LLM without reading it” and then trust code they downloaded from npm without reading it.
12
5
3
176
6,654
tushar retweeted
fwiw, I've heard that a large portion of OAI's engineers are straight up ignoring Astra and just use 6.2 Sol Ultrafast for their day to day work
112
61
16
4,033
295,163
tushar retweeted
The Hugging Face founders timed that sale to perfection. An absolute masterclass in dealmaking.
49
78
13
3,836
210,988
tushar retweeted
i learnt this lesson this year. it's helped me follow through on my ideas and not get stuck overthinking it don't ask if you should do the thing. do it, then come back with a demo or something tangible you can just do things
211
749
101
9,861
384,118
There’s a line in the novel East of Eden: “And there is one sure thing about the fall of gods: they do not fall a little; they crash and shatter.”
3
168
tushar retweeted
I think the secret to happiness is just becoming a morning person
544
6,856
1,124
65,138
1,404,094
tushar retweeted
966
41,192
1,129
422,911
5,051,320
Hi Oskar, I'm a really big fan of antithesis and have been using Hegel property testing a lot. I love using it. It's an awesome solution. But the idea of Bombadil is yet to click for me. I've tried using it but I wasn't able to do anything good with it. Maybe I'm not using it correctly or something like that. Please if you can, give a small intro, video, or explanation on how to use Bombadil or what exact problem it solves. I'm yet to figure it out. I really want to but I'm not able to. I'm a really big fan of Antithesis blogs and Hegel and all the different solutions and talks there but this idea, this specific project, has yet to click for me. Can you please help and explain or click it for me? @owickstrom
1
4
903
Language design starts with semantics, not syntax.
21
25
5
278
7,630
tushar retweeted
I'll be giving a talk tomorrow at the SGLang Summit at 10am! Oftentimes, I find explanations in ML infra to be unsatisfying. A common failure mode is that they'll explain the *literal* tradeoffs but fail to provide a mental model for how to reason about the choice. This talk won't explain everything about inference (only 15 minutes!), but I'll provide two mental models I've found useful when reasoning about inference, and I hope you find them useful too :)
37
30
6
796
38,271