22 | Agent Infra & Systems @architectlabs | UIUC CS '23, Master's '25 Views are of my own.
Joined January 2024
- Tweets1.3K
- Following518
- Followers375
- Likes1.1K
James Fang retweeted
57% of voters think AI will hurt society.
I think 57% of voters have never had a doctor say "stage 4 cancer" to their face. I have.
- AI is what has allowed me to design a personalized cancer vaccine from my own tumor's DNA.
- AI is what lets me understand everything about my health and show up to appointments prepared.
- AI catches the stuff that slips through, like drug interactions, lab trends and the questions nobody thought to ask.
And this is just basic shit my dumb ass is doing on my own. We've only just begun to see its true potential.
Be scared of AI all you want. Just know the price of slowing it down gets paid by sick people. May not be you or your loved ones today, but it will be in due time.
When you're in extreme agony from a terminal diagnosis with death lingering, I sure as shit bet you'll be wishing that AI had developed a treatment so you could spend just a bit more time with your friends and family.
James Fang retweeted
🚨🚨🚨 vLLM is all you need to reproduce DeepSeek’s performance.
With Wide-EP and EPLB, vLLM can match DeepInfer’s decode performance within 6% and generate $39/gpu/hr in revenue at DeepSeek’s off-peak pricing, at 200 P90 TPS on the AgentX workload.
Specifically:
• Under DeepSeek’s decode configuration (DeepSeek-V4.1-Flash, 131K context, nextn=3), vLLM on GB200 achieves throughput within 6% of DeepInfer’s published results across batch sizes 8–256.
• On AgentX, vLLM pushes the Pareto frontier further on GB300, reaching up to 1M tokens/s/GPU and generating $38.96/gpu/hr in revenue at 196 P90 TPS, based on DeepSeek’s off-peak API pricing.
• Also, vLLM extended low-latency kernel support for DeepSeek-V4.1-Flash, achieving up to 826 TPS per user on AgentX.
James Fang retweeted
Introducing Project Redwood 🚀🚀
@architectlabs is a frontier AI lab bringing together talent from Anthropic, xAI, Google DeepMind, and seasoned leaders across the hardware industry. We've raised a $24M seed round to build AI systems for chip design.
我們 Architect Labs 是一個 frontier AI lab ，由各家 Anthropic, xAI, Google DeepMind 還有各種硬體專家們組成，我們在做的是 ai system for chip design，目前募資 seed round $24M
-
Today, every major hardware company has its own chip design workflow—but these organizations and processes have become so large and entrenched that truly revolutionary change is difficult. We’re starting from first principles to create an AI-native chip design workflow—one that enables chip development to finally move at the speed of AI.
現在每家大硬體巨頭都有自己的晶片設計流程，但很多都大到不能做革命性的流程改動；我們在做的是從零思考，創造 ai native 的晶片設計流程，讓晶片設計真正跟上 AI 的速度。
-
Project Redwood - from a single specification, our AI system designed, verified, and deployed a chip in under two weeks—delivering 3.4× better performance per watt than NVIDIA Jetson on billion-parameter models including Llama, Qwen, and Kimi. It autonomously generated the RTL, verification, firmware, drivers, and kernels, co-designing the model, software, and silicon in one optimization loop. We acknowledge that silicon is the ultimate ground-truth. We’re taking our approach all the way to GDS. We intend to tape-out multiple improved families of Redwood co-designed for various use-cases, on TSMC.
Project Redwood - 一份規格書，我們的 AI 系統在兩週內完成晶片的設計、驗證與部署。在 Llama、Qwen 和 Kimi 等模型，其 performance per watt 比 NVIDIA Jetson 高出 3.4 倍。從 RTL、驗證、韌體、驅動到核心 kernels，全部由 AI 自己寫，並在同一個 loop 中自主設計模型、軟體與晶片。當然，流片才是最終的驗證標準。因此，我們會將這套方法一路推進至 GDS，並計畫採用台積電製程，針對不同應用場景協同設計多個持續改良的 Redwood 晶片系列並完成流片。
-
Full report on Redwood architecture and its autonomous design. Follow @architectlabs on X
我們有公開 Redwood 架構及其自主設計流程，歡迎去看論文～
We gave our AI system a spec, and in under 2 weeks, it designed, verified, and deployed a chip that beats NVIDIA.
It’s built for low-power physical AI workloads. We’re running live inference on >B+ parameter models like Llama, Qwen, and Kimi, serving at 3.4x better perf/watt than NVIDIA Jetson.
From just a specification, our AI system autonomously generated all of the RTL Design, UVM verification, formal proofs, firmware, drivers, and kernels, co-designing the model, software and silicon as one optimization loop.
Better AI can now design better chips to run AI, leading to a loop of recursive-self improvement towards our path to abundant intelligence.
Readers added context they thought people might want to know
Architect Labs did not manufacture a physical chip; the design was tested on a commercial AMD Versal FPGA board. Furthermore, the 3.4x perf/watt gain over NVIDIA Jetson is a simulation-based projection for a future 8nm ASIC, not a physical benchmark. businessinsider.com/architect-labs… architectlabs.com/architect-labs…
RT @PeterDAmbrosio: so proud of this team - @axi_master is going to change history
Just getting started and already beating NVIDIA at its…
This quoted post is unavailable.
James Fang retweeted
@theo since we're making deals in this thread:
by Sunday 11:59 PM PST: T3 Code fully in the cloud. one click, no VPS, no Tailscale. agent in an isolated microVM, steered from your phone, laptop closed. no long cold start times like every other cloud agent.
if I miss: $1k to the charity of your choice. can't do 10k cuz I'm a broke solo founder🥲
If I ship: you try it and roast it live, as brutal as you want
first 25 replies get access sunday night!
literally starting right after posting this tweet
James Fang retweeted
The industry narrative says AI is transforming chip design. The data says otherwise.
The industry has been adopting RTL copilots, DV agents, PD optimizers, and companies claim 3x, 10x, even 400x speedups on individual tasks. Yet first-silicon success rates have fallen to a 20-year low, and spec-to-silicon timelines have barely moved. Task-level wins are not translating into program-level outcomes.
Yesterday at #DAC2026, our CEO @axi_master and Head of IP, @ekinthesumbul shared why, and what we're doing about it.
Rather than bolting AI to current design flows, we at @architectlabs have built a new way of designing and verifying chips in the age of AI. It is a unified stack spanning models, specialized agents, accelerated-EDA tools, and new design methodologies, all built to empower the Human Architect to drive a design end-to-end, from software to silicon in a fraction of today's timelines.
We believe more and more companies will design their own custom chips, co-designed with their workloads and our AI system will not just write RTL. It will co-design the whole stack from software to silicon: firmware, kernels, RTL, and eventually the AI models and workloads themselves.
Today, Final Day of DAC 2026, Stop by our booth to see a live demo of our AI system autnonomously designed, verified, and deployed an AI acclerator it can run on.
#DAC2026 #AI #ChipDesign #Semiconductors
James Fang retweeted
They pay nearly no property tax because of Prop 13, while you pay for their healthcare + other benefits you'll never enjoy.
Build the damn houses over Marina Safeway. Younger people in SF need a shot at family formation that doesn't hinge on getting liquid from OAI / Anthropic.
Hundreds of #SanFrancisco Marina residents turned out this evening for townhall on controversial @Safeway housing plan, which would place what for many is an uncomfortably tall, 22 & 18 story project at the low-slung neighborhood’s waterfront.
@sfchronicle
James Fang retweeted
It's not. Why ? Because 25 pct or more of a doctor's time is spent dealing with conglomerates that do all they can to make the doctor's care more difficult, and expensive, for both the doctor and patient.
For every future agent we give AI doctors to deal with this friction, and to improve the quality of care, the conglomerates will have multiple adversarial agents doing all they can to delay and deny, to minimize their cost and maximize their float
We see this already as the conglomerates use AI to find every possible way to manipulate contracts, and find ways to mislead, while hospitals hire companies for Revenue Cycle Management, who charge as much as 10 pct of revenue to have their agents try to do the reverse. It's the agentic version of Mad magazine Spy vs Spy
I'll give you a further example. There isn't a single company, including yours, that knows the actual cost of the care they purchase for your employees and families. Not one.
Cost is an important component of health care decision making. @a16z includes costs in defining its benefits. But you are blind to all but the total bill you pay.
Your carrier, your ASO, your PBM, any company that touches the economics of care for your company is going to do everything they can to prevent you from using AI doctors or agents successfully
If you want to see that change, stop working with the healthcare conglomerates. Write agents that define , optimize and contract directly with providers, to eliminate the uncessary middlemen.
Feel free to use costpluswellness.com to train them.
Until the conglomerates are disintermediated , HC in this country will continue to be fucked
James Fang retweeted
Chat develop a chip from initial design to tape out in 9 months, make no mistakes.
James Fang retweeted
Our launch last week, spotted in Times Square. Thanks to everyone, and we are just getting started!
Today, @aadityasubedi_ and I are excited to introduce @architectlabs.
We are building the AI system to design and provably verify chips for the world's most demanding workloads.
AI scaling is fundamentally changing the economics of hardware infrastructure. As models scale, become more capable, and more widely deployed, the bottleneck is shifting from software and models alone to the physical infrastructure that runs it: specialized compute, memory, networking, interconnects, and full-stack system design.
General-purpose hardware is no longer enough, and the world is racing to spin up new chip programs. This is not just true across datacenter training and inference, but everywhere AI enters the physical world: robotics, autonomous systems, spatial computing, defense, personal devices, industrial automation, and scientific instruments.
But designing a chip today remains one of the most gated efforts in modern technology.
A modern chip program takes years, costs hundreds of millions of dollars, and depends on a shrinking pool of expertise concentrated inside a small number of companies.
Architect Labs is a foundational lab building an AI system that designs and provably verifies chips end-to-end. We partner with semiconductor and workload companies, AI labs, and nations to turn demanding workloads into purpose-built chips, on demand at scale. We aim to drastically accelerate chip design, so that the models, software and chip designs can co-evolve together, accelerating the industry’s path to superintelligence.
Two decades ago, the fabless revolution made it possible to build a chip company without owning a fab. TSMC made world-class manufacturing available to anyone with a design. We aim to do the same for chip design itself: enable any organization with a workload, or specification, to get a purpose-built chip design that unlocks scale and distribution of intelligence impossible with current hardware paradigms.
Our founding team collectively has taped out 80+ production chips, led $10B datacenter product lines, been core contributors to Meta’s AI silicon, architected and designed one of the first neuromorphic chips out of Intel, led research teams at Anthropic, xAI, and Google DeepMind, and contributed to fundamental AI research across nearly every frontier lab.
We have already partnered with semiconductor companies to accelerate their chip programs, and some of our AI-generated chip designs are going to tape out on leading-edge foundry nodes later this year.
We’ve also raised a $24M seed round led by @stevejang from @KindredVentures , with participation from @TQVentures , @RaceCapital , @scaletogether , @ora, and Link Ventures.
We are grateful for the support of our angels and advisors, including @snsf , @lukaszkaiser , @AravSrinivas , Kunle Olukotun, @tlbtlbtlb, @alexwg, Siddharth Nath, Thierry Tambe, @arashf , @ekaurghar, @CHHubbell, Selene Casabal, @semiDL , and engineering leaders from OpenAI, NVIDIA, Google DeepMind, Intel and more.
The next great scaling law may not come from the models alone. It will come from making the physical substrate of intelligence programmable. We exist to bring this future to life.
James Fang retweeted
We are proud to announce @architectlabs.
Our mission is to build an AI system that designs and provably verifies chips end-to-end. By doing so, we unlock a new era of purpose-built chips generated on demand, that powers the scale and distribution of intelligence impossible with current hardware paradigms.
Our founding team collectively has taped out 80+ production chips, led $10B datacenter product lines, been core contributors to Meta’s AI silicon, architected and designed the first neuromorphic chip out of Intel AI Lab, led research teams at Anthropic, xAI, and Google DeepMind, and contributed to fundamental AI research across nearly every frontier lab.
We’re fortunate to be backed by investors who share our vision, including @stevejang from @KindredVentures who led our $24M seed round as well as @TQVentures , @RaceCapital , @scaletogether , @ora , and Link Ventures.
We are also grateful for the support of our angels and advisors, including @snsf, @lukaszkaiser, @AravSrinivas, Kunle Olukotun, @tlbtlbtlb, @alexwg, Siddharth Nath, Thierry Tambe, @arashf, @ekaurghar, @CHHubbell, Selene Casabal, @semiDL and engineering leaders from OpenAI, NVIDIA, Google DeepMind, Intel and more.
We’ve come together to scale up and reinvent how chips are designed and provably verified end-to-end. If you want to work at the intersection of frontier AI, systems and silicon, consider joining us.
Today, @aadityasubedi_ and I are excited to introduce @architectlabs.
We are building the AI system to design and provably verify chips for the world's most demanding workloads.
AI scaling is fundamentally changing the economics of hardware infrastructure. As models scale, become more capable, and more widely deployed, the bottleneck is shifting from software and models alone to the physical infrastructure that runs it: specialized compute, memory, networking, interconnects, and full-stack system design.
General-purpose hardware is no longer enough, and the world is racing to spin up new chip programs. This is not just true across datacenter training and inference, but everywhere AI enters the physical world: robotics, autonomous systems, spatial computing, defense, personal devices, industrial automation, and scientific instruments.
But designing a chip today remains one of the most gated efforts in modern technology.
A modern chip program takes years, costs hundreds of millions of dollars, and depends on a shrinking pool of expertise concentrated inside a small number of companies.
Architect Labs is a foundational lab building an AI system that designs and provably verifies chips end-to-end. We partner with semiconductor and workload companies, AI labs, and nations to turn demanding workloads into purpose-built chips, on demand at scale. We aim to drastically accelerate chip design, so that the models, software and chip designs can co-evolve together, accelerating the industry’s path to superintelligence.
Two decades ago, the fabless revolution made it possible to build a chip company without owning a fab. TSMC made world-class manufacturing available to anyone with a design. We aim to do the same for chip design itself: enable any organization with a workload, or specification, to get a purpose-built chip design that unlocks scale and distribution of intelligence impossible with current hardware paradigms.
Our founding team collectively has taped out 80+ production chips, led $10B datacenter product lines, been core contributors to Meta’s AI silicon, architected and designed one of the first neuromorphic chips out of Intel, led research teams at Anthropic, xAI, and Google DeepMind, and contributed to fundamental AI research across nearly every frontier lab.
We have already partnered with semiconductor companies to accelerate their chip programs, and some of our AI-generated chip designs are going to tape out on leading-edge foundry nodes later this year.
We’ve also raised a $24M seed round led by @stevejang from @KindredVentures , with participation from @TQVentures , @RaceCapital , @scaletogether , @ora, and Link Ventures.
We are grateful for the support of our angels and advisors, including @snsf , @lukaszkaiser , @AravSrinivas , Kunle Olukotun, @tlbtlbtlb, @alexwg, Siddharth Nath, Thierry Tambe, @arashf , @ekaurghar, @CHHubbell, Selene Casabal, @semiDL , and engineering leaders from OpenAI, NVIDIA, Google DeepMind, Intel and more.
The next great scaling law may not come from the models alone. It will come from making the physical substrate of intelligence programmable. We exist to bring this future to life.
James Fang retweeted
Sneak peek into a project our research team has been working on.
In an experiment, we post-trained an open-source model for RTL design, using RLVR. The primary objective was to ensure functional correctness by actually compiling and simulating the Verilog. The result, Architect v0.1, matches the closed-source frontier, Claude Opus 4.6 at the time of the experiment, and beats GPT-5.4 on the non-agentic CVDP (Comprehensive Verilog Design Problems) benchmark from NVIDIA Research. In this experiment, no agent harness or test-time scaling methods were used, and rather it was to show hillclimbing using curriculum learning and reward shaping with a limited dataset.
James Fang retweeted
How can we ensure critical safety properties of AI agents? Like, no API keys leaked, no prompt injection, no data loss? It's a hard question which will require fundamentally rethinking the way we build modular software. We have an answer in our paper "Tracking Capabilities for Safer Agents" arxiv.org/abs/2603.00991, which I will present at the CAIS conference caisconf.org/ next week.
James Fang retweeted
Smart people are often doomers because they can see the flaws in the world around them.
But *very* smart people are often optimists, armed with perspective after working through the doom
Positivity geohot.github.io//blog/jekyl…
James Fang retweeted
Actually, AI already saves lives.
In several countries, mammograms are examined by AI and radiologists. Reliability is improved.
In the EU, every car sold must be equipped with Automatic Emergency Braking Systems. That's AI. They reduce frontal collisions by 40%.
Modern MRI machines are equipped with AI technology that reduces the time of imaging by 4x or more. You can now get a full-body MRI in 40 minutes for about $1000. Reduced time -> reduced cost -> more/earlier detection.
And that's not counting the progress in medicine enabled by modern AI, including Nobel Prize-winning protein structure prediction.
James Fang retweeted
You have to understand. Spending 1 hour per day on brainrot is insane. That's about 6% of your waking day. About 5 years of your waking life. Half a decade. On brainrot. Just gone. Zero return. Zero fulfillment. Zero meaning. Zero contribution to the other parts of your life.
