CTO @ General Compute
San Francisco, CA
Joined April 2020
- Tweets951
- Following695
- Followers3.8K
- Likes9.1K
Pinned Tweet
We raised $15m to build the ASICs-first inference cloud.
We're betting big on alternatives to GPUs, and the result is that we are already 5-8x faster on most models.
Read more about General Compute on Tech Crunch!
@FPuklowski @fastinference
techcrunch.com/2026/05/28/ha…
This is the way
DeepSeek V4.1 Flash 1017 runs at 1,521+ tok/s on Fireworks Ultrafast.
Fast enough to ship Tetris in 11 seconds, 7x faster than Claude Opus 5.5.
Join the Ultrafast waitlist: bit.ly/3UA0wKq
Jason Goodison retweeted
DeepSeek V4.1 Flash 1017 runs at 1,521+ tok/s on Fireworks Ultrafast.
Fast enough to ship Tetris in 11 seconds, 7x faster than Claude Opus 5.5.
Join the Ultrafast waitlist: bit.ly/3UA0wKq
Jason Goodison retweeted
Our heterogeneous cloud just got better. We're announcing a cluster scale deployment of @positron_ai as our next partner.
We believe demand for fast inference will outstrip supply for the next decade. Our bet on Positron is another step towards getting fast inference for opensource models into builders' hands.
Hardware goes live Q1 2027.
Thanks to @mitesh711, @trsohmers and the whole Positron team! We're excited for this partnership.
LFG!
Excited to partner with @positron_ai. We go live in Q1 2027.
We've spent a lot of time with @mitesh711, @trsohmers and the whole Positron team. I am really excited about what they have built and continue to develop.
Decode is a memory problem. Positron builds it's chips with a memory-first architecture.
GPUs on prefill, Positron on decode in @general_compute
xitter.cf/general_compute/status…
Our heterogeneous cloud just got better. We're announcing a cluster scale deployment of @positron_ai as our next partner.
We believe demand for fast inference will outstrip supply for the next decade. Our bet on Positron is another step towards getting fast inference for opensource models into builders' hands.
Hardware goes live Q1 2027.
Thanks to @mitesh711, @trsohmers and the whole Positron team! We're excited for this partnership.
LFG!
Jason Goodison retweeted
Compute & Cocktails 🍸
How do you finance AI infra? @medshelton_ (@GeneralCompute) + Stanley Lee (Liquid Compute) on underwriting compute, getting past prepay and the right debt/equity mix.
📍 Dawn Club, SF · Wed, Oct 28
Request to join → luma.com/g9h2d9ai
1.5B! Insane @tenstorrent. We'll done.
I've been programming with TT-Metal and TT-NN for the last week in my spare time. Great software and awesome team
We raised $1.5B to deploy Galaxy globally.
Thanks to our lead investors Trace Capital and Innovation X for backing that growth.
More to come: tenstorrent.com/newsroom/ten…
Keep up the great work @opencode. Its a fantastic product, and incredible team, and you guys are handling this situation extremely professionally
We had a security breach last night related to a large scale distillation attack that impacted 1.4% of OpenCode Console users
We have notified them all this morning via email. If you did not get an email you were not impacted, it was a subset of users who had external providers linked
The incident is ongoing so we cannot share more details at this point but we are working with Anthropic, OpenAI, and Google to remediate
Center stage for @GoldenGateVC's panel in Singapore last weekend. Thanks for having me @vlauria. Great to be a part of F1 Founders Rally!
AI is the biggest capital deployment in history, and it’s still capital constrained. For alternative silicon, it’s even tighter.
Demand for fast tokens has gone exponential in 6 months. Capital is following it to alternative accelerators. @general_compute is in pole position to deliver this.
ALT image
Today we are announcing our biggest bet yet on our own silicon.
Starting next calendar year, we will scale our Maia family to multi-gigawatt deployments, making it the primary inference system for our own product and services.
And we’re introducing Maia 300, which delivers 40% more tokens per watt compared to leading GPU-based systems.
We are already deploying Maia 200 across the US, Europe, and Australia, delivering up to 1.7x more tokens per watt across experiences in Teams and Copilot.
Our goal is to build on the full-stack efficiency gains we’ve already seen, while continuing to bring the latest innovations from across the ecosystem into our fleet.
The hard part of splitting prefill and decode isn't the split. It's picking the ratio.
Get it wrong and half your fleet sits idle while the other half queues. And the right ratio moves through the day as traffic shifts between input-heavy and output-heavy work.
@sgl_project shipped a fix for this last week: a node can now switch between prefill and decode without a restart.
Big win and awesome work to their team
Jason Goodison retweeted
Big news: @dMatrix_AI is partnering with @nvidia.
Our next-gen Raptor XPUs with our path-breaking 3D DRAM stacked memory will plug directly into NVIDIA MGX rack-scale systems via NVIDIA NVLink Fusion — delivering ultra-low latency inference for premium token services.
Inference is an infinite opportunity, and it will demand a diversity of compute. Customers will be able to run a Raptor rack standalone, or alongside NVIDIA GPUs — matching the best compute to each phase of the workload.
Grateful to @JensenHuang and the NVIDIA team for building an open ecosystem that makes this possible. Let's go build the future!
🔗 d-matrix.ai/announcements/d-…
Jason Goodison retweeted
Woah, @sidsheth at @dMatrix_AI making the big moves
You know the future of compute is heterogeneous when @JensenHuang
starts putting down some chips
Now we know why they integrated into nvidia's nvlink a few weeks back
P.s you could have mentioned this at dinner last night buddy 😄
Instinct, Muse are changing the face of how users interact with the web.
A site built for humans makes agents wade through hero images and animations to find a base URL.
So we built a mode toggle into our site. Human is the normal site. Agent is just markdown: API base URL, protocol, and sign-up steps an agent can follow on its own.
This is now a standard for developer docs, it is becoming a standard for AEO, but it will also become a standard for web pages in general.
The way I see the transition:
UX was about getting a person to the value in fewer clicks. AX is about getting an agent there in fewer tokens.
I honestly think the future is AX-only - at least for most non-physical experiences.
But i'm long curl+markdown and short mcp
ALT image
A KV cache doesn't have to live in HBM.
Most serving stacks now tier it: HBM for the hot sessions, CPU memory for warm ones, SSD for the long tail.
Each step down is cheaper and bigger, and slower to pull back from. But pulling a cached prefix back from CPU memory is still usually faster than recomputing the whole prefill.
It's one of the ways coding agents hit 90%+ cache rates
Jason Goodison retweeted
Disaggregated inference is the future.
Great panel with @general_compute discussing their heterogenous compute cloud, the speciation of compute for disaggregated inference, and how thermo compute will fit in the picture
More coming soon. Stay tuned 😉
LFG @beffjezos 💪
