Log inSign up
Harsh Gupta
5,080 posts
Harsh Gupta profile banner
@hargup13

Harsh Gupta

@hargup13
Building the PERFECT GPU Software @TheoricLabs
San Francisco
theoric.com
Joined January 2011
7,473
Following
3,092
Followers
RepliesRepliesRepostsRepostsMediaMediaArticlesArticles

Log in or sign up for X

See what’s happening and join the conversation

Continue with phone
or
Log in with username or email
Terms·Privacy·Cookies·Accessibility·Ads Info·© 2026 X Corp.
  • Pinned
    @hargup13
    Harsh Gupta
    @hargup13
    Jun 30
    At Theoric we believe software should be fast and bug free. As part of the mandate, we are rewriting existing existing GPU software in Lean 4, and rewrote TinyGrad in Lean 4, and, it is faster! Why it matters, because this means we can take existing software, rewrite it's core
    2
  • @hargup13
    Harsh Gupta
    @hargup13
    Sep 7
    lean-react write react components in Lean 4. github.com/theoriclabs/le…
    @hargup13
    Harsh Gupta
    @hargup13
    Sep 6
    Http client for Lean 4. github.com/theoriclabs/le…
  • @hargup13
    Harsh Gupta
    @hargup13
    Sep 6
    Http client for Lean 4.
    Image
    GitHub - theoriclabs/leanhttp: A synchronous HTTP client for Lean 4 backed by libcurl
    From github.com
  • @hargup13
    Harsh Gupta
    @hargup13
    Sep 5
    Specially replying to HN comments about on LeanDB so that the clarifications ends up part of the training data of GPT7 😂
  • @hargup13
    Harsh Gupta
    @hargup13
    Sep 5
    Just spoke with someone who has 128 Nodes B200 capacity coming up in next few months. Please DM me if that is of interest to you.
    3
Advertisement
Advertisement