David Xia retweeted
Back from @modal eng hackathon in Mohonk. My team built a Portal-themed AI escape room, all running on Modal. The twist: your AI helper, GLaDOS, Is secretly trying to phish you for more permissions.
60+ eng played and we won “For the Love of the Game” category. Check it out:
Runtime is happening in two days.
Join us to hear speakers from @AnthropicAI, @typesafeai, @nvidia, @tryramp, and more.
David Xia retweeted
Qwen 3.8-Max is live on Modal, with the full 1M context window and a custom DFlash speculator under the hood.
Love seeing our launch partner Modal go all in from Day 0!⚡️
2.4T parameters. 1M context. Somehow, still just a Modal call away. @modal
Qwen3.8-2.4T-A95B by @Alibaba_Qwen and @alibaba_cloud is now available on Modal.
Served with a custom DFlash speculator trained on tool-call-heavy data. Full 1M context window.
David Xia retweeted
Added a fun lil widget to the LLM Engineer's Almanac -- a "Token Timing Simulator" so you can get a visceral feel for what a benchmark perf number means.
Here's @_dcw02's latest work with @zhijianliu's DFlash technique in @sgl_project -- ~1k TPS!
modal.com/llm-almanac/token-…
David Xia retweeted
** TRAVEL BAN FOR NYC **
All bridges, highways, and streets will be closed to traffic from 9pm Sunday thru noon Monday. No cars, trucks, scooters, or e-bikes. Limited exemptions for essential and emergency movement.
This is projected to be NYC’s worst blizzard in a decade, with extremely dangerous conditions.
Please take this seriously. Stay home if at all possible.
David Xia retweeted
The best part about working on @modal is watching our customers literally build the future. Today, they’re using @modal for everything from generating AI music to powering real-world robots to curing cancer.
And yet, this is just the beginning. We have so much more to build.
If you like making computers do impossible things to solve user problems, please reach out. Especially if you are interested in low-level or large-scale systems, applied AI, GPU performance, delightful user experiences, security, or any combination of these.
It's true – @modal has raised a $87M Series B at a $1.1B valuation to advance the future of AI infrastructure.
Thank you to @Lux_Capital, @Redpoint, @AmplifyPartners, and others.
Now more than ever, AI demands a complete reinvention of traditional compute infrastructure
David Xia retweeted
It's true – @modal has raised a $87M Series B at a $1.1B valuation to advance the future of AI infrastructure.
Thank you to @Lux_Capital, @Redpoint, @AmplifyPartners, and others.
Now more than ever, AI demands a complete reinvention of traditional compute infrastructure
David Xia retweeted
I’m joining @modal to lead Product Marketing, and I couldn't be more excited.
I’ve learned to go where the developer love is. Find that, and the rest will follow.
That’s exactly why I was drawn to Modal. @bernhardsson and @akshat_b are building the tools that data and AI teams need, with a level of care and understanding that shines through in the product.
I'm so pumped to tell this story 🚀
David Xia retweeted
At @modal we've built every layer of the AI infra stack from scratch — from filesystems and networking to our own async queues and multi-cloud GPU orchestration.
I sat down with @narayanarjun from @amplifypartners to go into depth on all of this, including the fun ways the Linux kernel has tried to stop us along the way:
David Xia retweeted
The uv build backend is now stable, and considered ready for production use.
An alternative to setuptools, hatchling, etc. for pure Python projects, with a focus on good defaults, user-friendly error messages, and performance.
When used with uv, it's 10-35x faster.
David Xia retweeted
KubeRay v1.3.0 is here with major upgrades for running @raydistributed on @kubernetesio. 🚀
Key highlights:
⚡️Enhanced observability: The new RayService Conditions API and RayCluster Conditions API have reached Beta, along with the Kubernetes Events API for better tracking and monitoring.
⚡️Stronger reliability: Improved failure recovery, better resource management for more stable Ray clusters, and a doubling of end-to-end tests from 23 to 46.
⚡️Improved usability: Ray Kubectl plugin (Beta) simplifies job and cluster management, along with a new unified GCS fault tolerance API.
KubeRay v1.3.0 improves observability, reliability, and usability making Ray on Kubernetes easier to monitor, recover, and manage. 🙌
Read full blog article here: anyscale.com/blog/kuberay-v1…
David Xia retweeted
Introducing the Ray Kubectl Plugin (Beta)! 🙌
Now, managing @raydistributed clusters on @kubernetesio is faster and more efficient.
Key enhancements:
🔹Streamlined workflows - Run your Ray scripts from your laptop on a remote K8s cluster with a single command.
🔹Create Ray clusters & worker groups with a single command, no YAML needed.
Read more about streamlining your Ray cluster management on Kubernetes here: anyscale.com/blog/introducin…
David Xia retweeted
“Why aren’t you cheering Trump & DOGE? I thought you wanted spending and deficit cuts!"
Because I’ve been doing this for 25 years and can’t be tricked by gimmicky nonsense. Trump’s first term added $8 trillion in enacted spending hikes and tax cuts to the deficit - half of which was unrelated to the pandemic. This time around, Trump has proposed roughly $8 trillion more in tax cuts and spending hikes over the decade. And right now, a GOP Congress is preparing to abandon most reconciliation cuts and instead add $325 billion this year in new spending. We’re headed towards $4 trillion deficits within a decade.
So, no, I don’t get excited when DOGE cancels $1 billion in govt contracts. Or saves $3 billion in federal workforce reductions out of a $7,000 billion budget. Not when Trump and Congress are also preparing to add $800 billion more annually in proposed new tax cuts and spending.
And no, the huge savings are not coming. Even (unrealistically) eliminating 20% of the federal workforce would save $60 billion, and overhauling federal systems to sharply reduce payment errors may save perhaps $80 billion (and is probably unlikely too). For all of DOGE’s bluster, administrative and executive reforms would at best save 1-2% of federal spending and offset only a small fraction of Trump’s red ink agenda.
That leaves trying to unilaterally impound spending such as USAID—which is wildly illegal—or actually going to Congress to pare back spending the constitutional way. But Trump has already taken Social Security, Medicare, defense, veterans, border (and interest) off the table, which is 2/3 of all spending and is driving deficits. And the GOP Congress seems ready to give up on cutting the remaining one-third of spending. Want to cut spending and the deficit? How about they stop passing budget-busting bills. Don’t brag about your coupon-clipping frugality at the same time you are buying a $250,000 Ferrari. I’m not going to cheer Trump and DOGE for adding “only” $750 billion to deficits instead of $800 billion. We’re still going backwards.
I’ve spent decades studying the federal budget. I know that $7 trillion(!) behemoth inside and out – where the money really goes, and where the savings opportunities lie. So I can also detect bullshitters who talk tough about trillion-dollar spending cuts without doing their homework. It’s the ones who claim most spending goes to undefined “waste,” federal salaries, immigrants, foreigners, Ukraine, or non-working welfare recipients. It’s the ones who claim we can easily balance the budget or cut $1 trillion without specifying exactly what line-items to cut. Or that we can return to 2019 spending levels for each program, which means a 20% inflationary cut, defaulting on the federal debt, and kicking off every senior who has since retired into Social Security and Medicare. It’s all hot air and empty bluster. Tough talk without following through on anything substantive. Just wait until you see the final deficit numbers in October.
And this is why GOP movements to cut spending always fail. They make absurdly ambitious promises without doing their homework, understanding where the money goes, and specifying real plans to fix it. You can’t significantly cut the deficit just by cutting waste, firing bureaucrats, and defunding immigrants and foreigners. There are no easy short cuts. You have to stop cutting taxes and then address Social Security, Medicare, defense, and a lot of other popular programs. Wake me when the GOP goes there.
So, no, I will not get excited about a couple billion in DOGE savings on one hand while Trump pushes Congress to add $8 trillion over the decade in tax cuts and spending with the other hand. I’m not that gullible.





















