AX (@AX_Computelabs) header
AX (@AX_Computelabs) profile picture

Co-founder & CTO @Compute_Labs | Previously @Airbnb, @Apple, @awscloud | Cat Entertainer @ https://nitter.cf/t.co/ln3GVmDV4M | Photographer @ https://nitter.cf/t.co/5dnRuQ79dH

Redmond, WA
Joined September 2024
Photo by AX (@AX_Computelabs)Photo by AX (@AX_Computelabs)Photo by AX (@AX_Computelabs)Photo by AX (@AX_Computelabs)Photo by AX (@AX_Computelabs)Photo by AX (@AX_Computelabs)Photo by AX (@AX_Computelabs)Photo by AX (@AX_Computelabs)Photo by AX (@AX_Computelabs)Photo by AX (@AX_Computelabs)
Neocloud operators can work from our lender-approved Gold Standard MSA with their customers and Compute Labs gets the deal financed. Ask us about it at platform.computelabs.ai
1
4
30
6,821
Mubadala, Microsoft and SoftBank Vision Fund launched @Hub71ad to bring top tech companies to Abu Dhabi. Compute Labs recently joined Cohort 19. One of 40 companies picked out of 2,502 applicants. Our Financing Platform helps neoclouds structure the financing needed to scale compute capacity. Joining Hub71 puts us closer to Abu Dhabi's capital and partners Read more here: hub71.com/latest-news/press-…
3
18
1,132
Compute Labs has helped neocloud operators secure over $500M in institutional debt term sheets in the last few weeks, and our Financing Platform is now live to get your cluster funded next. Turn signed customer demand into earning clusters fast. Get started: platform.computelabs.ai
5
2
21
3,401
Compute Labs was featured in @CNBC this weekend in Kif Leswing and Isabel O'Brien's piece examining how capital markets are adapting to fund the global AI buildout. Demand for compute is clearly not the constraint. The bottleneck is the structural mismatch between how AI clusters operate and how institutional credit underwrites: • The "timing trap" (quote / power / capital): Hardware allocation quotes, colocation power delivery, and credit committee timelines rarely line up naturally • Capital crowds into hyperscale deals, leaving mid-sized neoclouds caught between dilutive equity and credit funds built for large checks • Having a customer contract is different from having a bankable one, where prepayments, SLA step-downs, and deployment timelines can actually clear an investment committee Thank you to the CNBC team for including our perspective on how the financing market is evolving. Read the full article: cnbc.com/2026/10/24/nvidias-…
1
5
10
797
Yesterday Elon warned that the market is going to see a lot of stranded compute in partially complete data centers. He noted that if anything is missing in a buildout, however seemingly trivial, the compute does not work. Anyone who has worked on a GPU deployment understands the severity of the execution risk. B300s or GB300s can ship in a few months, but power, transformers, or switchgear can take years. Even if a site has power lined up, a delayed permit, unfinished liquid cooling, or bad networking during testing will keep a cluster offline. An offtake contract only starts paying once the cluster Ready for Service (RFS). OEMs usually want to be paid on or before delivery, while lenders and customers only want to pay once the cluster is live. If the GPUs were to arrive months before the data center is ready, the operator is stuck paying for hardware that cannot earn a dollar yet. That is why institutional GPU financing must be rigorous. Before lenders commit capital, they must underwrite the entire buildout. We built the Compute Labs Financing Platform to help neoclouds and lenders get all of that lined up from day one so clusters get funded and go live on time. platform.computelabs.ai
This is a fuller list of the limitations of supercomputer buildout. If anything is missing, however seemingly trivial, the compute doesn’t work.
14
6
36
14,885
Introducing the Compute Labs Financing Platform. Financing your GPU cluster shouldn't feel impossible. Capital is ready, but most neocloud operators with customer demand have no capital markets team to turn it into a deal funders can confidently evaluate. Our platform turns a neocloud's deal documents into a standardized package, with every figure verified by its source. Neoclouds receive support from intake to close and beyond. Funders get deals they can evaluate with confidence. platform.computelabs.ai
3
18
20
57
12,683
Financing a GPU deal requires keeping every document current and every counterparty in sync. All while the quote risks increasing and the offtaker waits. We're building one place to make your deal financeable, with our team alongside you. Compute Labs Financing Platform. Coming soon.
3
1
11
1,086
Introducing Anatomy Describe a machine in one line. Claude brings it to life. - Interactive isometric wireframes you can rotate, inspect and explore - WebGL shaders for fire, water, plasma and more - Real 3D motion, with individual parts that rotate and move Install it as an agent skill. Fully open source.
432
1,593
670
18,240
2,702,115
Sonnet 5.5 发布了,这个模型很神奇:它在 Agentic coding 测试上的得分比 Opus 5.5 还要高 4 分。 定价上也非常便宜,每百万 token 输入输出的价格分别是 $2 和 $10,与 GPT-6 Sol 一样。而且它的运行速度相对于 Sonnet 5 快了 30%,Claude 官方表示成本也降低了 30%。 至于为什么它在 Agentic coding 上的表现跟 Opus 5.5 差不多甚至更高,可以看一下价格测试。 Sonnet 5.5 在 max 努力程度下,同样任务的花费金额比 Opus 5.5 还要高得多,快要持平甚至反超了,这可能就是它得分高的原因。 所以大家在用 Opus 5.5 或者 Sonnet 5.5 的时候千万别开 max:在一些复杂任务上,它有可能跑出非常贵的价格,几乎赶上 Claude Fable 5.1 了
13
2
39
20,250
Maybe our cutest launch so far. But still packing the biggest punch.
Please welcome GPT-6 Sol and GPT-6 Luna to the GPT-6 universe. GPT-6 Sol and Luna build on the advances behind GPT-6 Astra, bringing much of its strengths into faster and more affordable models to support work at scale. We’ve also made caching and inference more efficient, and we’re passing the savings directly to you: 50% lower API prices for Sol and Luna compared with GPT‑5.6 promotional pricing.
1,196
177
80
7,707
1,242,052
AX retweeted
here's how i shipped 2,500 PRs last month to production this was originally supposed to be for Cursor Compile in London. i couldn't make it since i was livestreaming for Grok @Bot Galaxy so i'm making it available for free here on X! watch it on 2x speed, i talk slowly
852
1,963
659
22,549
4,739,225
Ladies and gentlemen... start... your... ENGINES. We are almost Tuesday and I promised a reset for Tuesday. Among some other things. See you soon.
5,455
1,329
1,677
27,749
5,894,877
In July 2025 we bought thirty-two NVIDIA H200s and put them on the on-demand market. We are publishing fourteen months of actuals from that fleet. Very little operating data on GPU fleets is public. This is ours.
4
2
2
17
113,960
AX retweeted
Pi 两位作者的见解: 1. 代码即真相,代码不需要记忆系统,不需要 RAG,模型很擅长理解代码结构 2. Bash 工具足够用,Bash 类似于编程语言,可以任意组合;大部分时候没必要 MCP,skill + 脚本足够。
Good morning from Vienna People of Pi 🌞 Sunday meditations from @badlogicgames and @mitsuhiko - On memory: code is the truth - Bash is all you need - Build context efficient tools
117
155
15
1,221
228,858
Oops... I did it again. Enjoy reset usage limits for all paid users for Codex and ChatGPT Work. Super grateful for an incredible team who is iterating at lightspeed and keeping the infra up as we scale faster than ever. Enjoy the weekend!
GPT-5.6 Sol confirmed to be an extremely good model
2,498
808
812
15,567
1,419,252
Replying to @thsottiaux
Personally, I find it easier to manage my usage without the 5h limit. Sometimes I have big bursts of productivity, and other days are more laid-back. I like not having to worry about my 5h window
15
9
1
1,793
59,744
New in Claude Code: /checkup Run /checkup to: 1. Clean up unused skills/MCPs/plugins and save context 2. Dedup your local CLAUDE.md against the checked in CLAUDE.md 3. Break up root CLAUDE.md into nested CLAUDE.md's + skills 4. Turn off slow hooks 5. Update your Claude Code to the latest version 6. Enable auto mode by default 7. Pre-approve frequently denied read-only commands .. And a few other goodies. /checkup confirms with you before making any changes. Enjoy!
343
956
192
11,584
876,149
Claude Fable 5 will be available again globally tomorrow. After a series of productive conversations with the US government, we're redeploying the model with a new set of classifiers to target and block more cybersecurity tasks. In the near term, some routine tasks like coding and debugging will fall back to Opus 4.8. We’ll continue to refine these classifiers over the coming weeks to reduce false positives and better distinguish genuine misuse from legitimate requests. We’ve also begun drafting a consensus framework—with Amazon, Microsoft, Google, and other Glasswing partners—for assessing the severity of AI jailbreaks and how AI developers should respond to them. We invite other industry partners and model providers to join us in this effort. Finally, we’re scaling up our collaboration with the US government on model testing and safeguards. This will include pre-release access to models and safeguards for evaluation, information sharing on jailbreaks and misuse, and dedicated resources for joint research. Thank you to our users for your patience, and to our partners across the government, industry, and the research community who worked alongside us to make Fable 5 available again. Read our full blog: anthropic.com/news/redeployi…
3,484
6,356
4,278
42,675
15,749,795