Basic Compute | AI fashions that run on an inference cloud optimized for pace

a04ce0a0 f044 4694 b61d cb584601b450.png


Howdy Product Hunt, I am Jason, Co-founder & CTO of Basic Compute!

The Downside

Brokers are probably the most thrilling factor taking place in AI presently however the infra they run on used to be designed for chatbots, now not independent workflows. When an agent has to make 20, 50, on occasion masses of sequential LLM calls to finish a role, latency compounds right into a ceiling on what is in truth conceivable.

Maximum inference suppliers as of late hit you with one among two tradeoffs:

  1. GPU-based stacks – Nice for coaching, however memory-bandwidth bottlenecks imply your agent runs slowly (~120 tokens/2nd)

  2. “Rapid” inference with catches – Some suppliers ship pace however lock you into small fashions, restricted context home windows, or pricing that breaks at agent-scale token quantity. Velocity with out intelligence isn’t well worth the industry off.

After years construction voice brokers and real-time AI merchandise ourselves, we were given uninterested in ready. So we constructed Basic Compute.

How Basic Compute is Other 🚀

GC is an ASIC-first inference cloud constructed on a couple of chips, together with SambaNova. SN makes use of a three tier reminiscence structure and dataflow, which is a complicated approach of claiming “It’s actually speedy purpose we don’t have the similar bottlenecks”.

  • 🔹 Agent first (OpenClaw) – Brokers can enroll on their very own and set up their very own API keys. OpenClaw can transfer its inference simply by pointing it at our web page.

  • 🔹 Constructed for agent workloads – Tuned for each coding brokers and voice AI (TTFT), the issues that topic if you find yourself chaining dozens of calls. Your agent finishes in seconds, now not mins.

  • 🔹 Velocity with out the tradeoffs – Frontier open fashions, complete context home windows, and pricing that in truth works at manufacturing scale.

Who is that this for?

If you are construction AI brokers, voice AI ,and even simply the usage of OpenClaw or OpenCode and wish sooner inference, then GC is constructed for you. Sooner inference is not only a nice-to-have; it unlocks use circumstances that were not viable prior to.

🔗 Get began as of late

Join at https://generalcompute.com and get started operating your workloads on ASICs as of late. We’re providing $200 in loose credit score to any individual that indicators up throughout the Product Hunt release (up from the traditional $5 in credit score)


Leave a Comment

Your email address will not be published. Required fields are marked *