LogoThread Easy
  • Explorer
  • Composer un thread
LogoThread Easy

Votre partenaire tout-en-un pour les threads Twitter

© 2025 Thread Easy All Rights Reserved.

Explorer

Newest first — browse tweet threads

Keep on to blur preview images; turn off to show them clearly

...

HVM4 now includes a general method to compile Interaction Calculus functions to zero-overhead machine code, including functions with superpositions.

Note that HVM2 (which Bend used originally) always relied on an interpreter. We promised we'd eventually compile it. HVM3 compiled Haskell-like functions only, getting a 100x+ boost in these cases, but still needed to fall back to the "interpreter" when superpositions were involved. Finally, for the first time ever, HVM4 is now capable of running Interaction Calculus functions, even these with superpositions, in full compiled mode, with no overhead.

It is weird how we're pushing Interaction Nets beyond anything that has ever been done before, to the point we don't even have anyone to communicate the progress. The field is all but nonexistent, there are no papers, no conventions, no community. My tweets are basically pushing the edge of what humanity knows about this paradigm. That feels really awkward, and I think this is mostly attributable to it being genuinely unintuitive.

In any case, below is a lazy commit message explaining the last case of this algorithm: how to deal with applying a static (compiled) superposition to a dynamic argument...

... HVM4 now includes a general method to compile Interaction Calculus functions to zero-overhead machine code, including functions with superpositions. Note that HVM2 (which Bend used originally) always relied on an interpreter. We promised we'd eventually compile it. HVM3 compiled Haskell-like functions only, getting a 100x+ boost in these cases, but still needed to fall back to the "interpreter" when superpositions were involved. Finally, for the first time ever, HVM4 is now capable of running Interaction Calculus functions, even these with superpositions, in full compiled mode, with no overhead. It is weird how we're pushing Interaction Nets beyond anything that has ever been done before, to the point we don't even have anyone to communicate the progress. The field is all but nonexistent, there are no papers, no conventions, no community. My tweets are basically pushing the edge of what humanity knows about this paradigm. That feels really awkward, and I think this is mostly attributable to it being genuinely unintuitive. In any case, below is a lazy commit message explaining the last case of this algorithm: how to deal with applying a static (compiled) superposition to a dynamic argument...

Kind / Bend / HVM / INets / λCalculus

avatar for Taelin
Taelin
Mon Nov 03 12:17:01
OpenAI could have done better here

was definitely expecting a cool mug

OpenAI could have done better here was definitely expecting a cool mug

Built Tweet Hunter, Taplio (sold $8m) Growing https://t.co/OyNJ8ZUyOh - https://t.co/jS9GQJ5Ps8 - https://t.co/EFUcKeBbpU - https://t.co/JkVOl1O0S1 - https://t.co/KG9PgxJabg Sharing weekly tips about growth: https://t.co/ereQodN3Ov

avatar for Tibo
Tibo
Mon Nov 03 12:16:48
dumb question, how do you cool a datacenter in a vacuum?

also H100s are already nearly obsolete oops gotta send a datacenter sysadmin astronaut out there to swap in B200s

dumb question, how do you cool a datacenter in a vacuum? also H100s are already nearly obsolete oops gotta send a datacenter sysadmin astronaut out there to swap in B200s

research @cornell // language models, information theory, science of AI

avatar for Jack Morris
Jack Morris
Mon Nov 03 12:12:26
dumb question, how do you cool a datacenter in a vacuum?

also H100s are already nearly obsolete oops gotta send a datacenter sysadmin astronaut out there to swap in B200s

dumb question, how do you cool a datacenter in a vacuum? also H100s are already nearly obsolete oops gotta send a datacenter sysadmin astronaut out there to swap in B200s

research @cornell // language models, information theory, science of AI

avatar for Jack Morris
Jack Morris
Mon Nov 03 12:12:26
dumb question, how do you cool a datacenter in a vacuum?

also H100s are already nearly obsolete oops gotta send a datacenter sysadmin astronaut out there to swap in B200s

dumb question, how do you cool a datacenter in a vacuum? also H100s are already nearly obsolete oops gotta send a datacenter sysadmin astronaut out there to swap in B200s

research @cornell // language models, information theory, science of AI

avatar for Jack Morris
Jack Morris
Mon Nov 03 12:12:26
Thread

Thread

We're sharing/showcasing best of @github projects/repos. Follow to stay in loop. Promoting Open-Source Contributions. UNOFFICIAL, but followed by github

avatar for GitHub Projects Community
GitHub Projects Community
Mon Nov 03 12:12:03
  • Previous
  • 1
  • More pages
  • 1219
  • 1220
  • 1221
  • More pages
  • 2127
  • Next