Alex Volkov
@altryne
🎙️ Host of @thursdai_pod (get your AI news with us!)
✨ AI Evangelist with @wandb & @coreweave 🪄🐝
Opinions my own
2.7K Following 43.4K Followers
Early vibes at AIE are dope! Excited for tomorrow!
As always an incredible group of folks, and knuckleheads
@aiDotEngineer
Just arrived at @aiDotEngineer and got my headshot taken at the @AnthropicAI booth!
In case you can't make it, he's how that feels hahah!
Keep a follow I'll be documenting my @aiDotEngineer journey till the live ThursdAI pod live from the floor here!
Show more
1.7 GigaWatt of active power!
Today we announced our Q3 earnings:
🔹$3.7B revenue, up 168% YoY
🔹$379M adjusted operating income, well above guidance
🔹Double-digit adj. operating margins, a quarter ahead of plan
🔹1.7GW+ of active power, doubling our active footprint YTD
🔹Roughly $9B in customer prepayments and deposits in this quarter
Congratulations to our team, who remain focused on executing for our customers.
Read the full Q3 2026 results.
Show more
Today we announced our Q3 earnings:
🔹$3.7B revenue, up 168% YoY
🔹$379M adjusted operating income, well above guidance
🔹Double-digit adj. operating margins, a quarter ahead of plan
🔹1.7GW+ of active power, doubling our active footprint YTD
🔹Roughly $9B in customer prepayments and deposits in this quarter
Congratulations to our team, who remain focused on executing for our customers.
Read the full Q3 2026 results.
Show more
Wanna heal to 101%? (Genius concept btw!)
Text your SO that you're the type of couple who heals to 101%
We all get down. We all come back.
So the only difference between us is how quickly, and fully, can you bounce back?
Two of my favorite rules here are:
90 seconds of suffering - Stole this from Tony Robbins.
When I get upset. I don’t deny it. I don’t resist it. I don’t lie to myself (“I’m not mad, you’re mad!”). Instead, I give myself 90 seconds to wallow in it. Full permission to feel shitty for 90 seconds.
Now, 90 seconds sounds short. In reality, when I do this, halfway thru I’m looking at my watch, because the feeling passes when you don’t resist it. And I’m ready to move back into a better state of mind.
Heal to 101% - One of my favorite big ideas is that every relationship will have arguments - but does the relationship heal to 99% or 101%?
If it’s 99%, that means you lost a little in that fight. A little trust. A little respect. A little love. And over time, losing 1% each time, the relationship will shrink.
But if you can focus on healing to 101%. Bounce back better than where you were before. Then the relationship grows every time.
Show more
HOT TAKE: the gaming and modding communities will take this game to heights unseen!
120fps in the first month
BREAKING: GTA 6 will run at 30fps on all consoles, with PS5 Pro enhancements that improve ray tracing, per IGN
We built CoreWeave Cloud on Kubernetes, and as of today, we're now a @CloudNativeFdn Platinum member.
@ritazzhang announced it on the @KubeCon_ keynote stage this morning.
Rita sits on the Kubernetes Steering Committee and co-chairs SIG Auth, and our engineers contribute directly to Kubernetes, Kueue, and CloudNativePG.
We're also delivering the capacity for @NVIDIA's $4 million, three-year pledge, so CNCF projects can test their code on real hardware instead of simulators.
⇢
Show more
This is the way!
@bot can now connect to private MCP servers on your tailnet without exposing them to the public internet.
Powered by Tailscale OAuth apps, now GA, so you can build secure agent integrations, too.
Read more here:
Show more
@bot can now connect to private MCP servers on your tailnet without exposing them to the public internet.
Powered by Tailscale OAuth apps, now GA, so you can build secure agent integrations, too.
Read more here:
Show more
My @aiDotEngineer talk is live.
This is how we build with agents at @HelloUntangle
Been great iterating fast on this with @DaleSpoonemore and @DevinAI
The top trending model on @huggingface currently is an image-to-3D model called Idol. 🔥
It is built by @speridlabs, a company based in Spain. 🇪🇸
> 600M parameters, pre-trained from scratch.
> requires less than 7.2GB of VRAM
> generates 1M splats in 3 seconds
> Apache 2.0 license
Kudos to Speridlabs for this amazing open release!
Find the links below
Show more
$2,618 prompt - this is what Fable 5.2 (Max) built overnight
Muse Apple Watch integration is nice. I don’t do a lot of stuff on my watch apart from check the weather and activate workouts, but this is smooth. Similar to Muse Gadget but way more polished. Preview of what we’re getting with Muse Charm. Don’t have voice mode yet so not sure how that works on the watch, if at all. Would like emoji reactions for quick responses, which I sent as feedback in the app.
Show more
On my way to AIE Code Summit in SF!
One way out of the slop is to move to patterns where the "cost of durability" is lower. I'll be talking about combining agent skills with static analysis(ast-grep) to migrate to @EffectTS_ at @aiDotEngineer.
Hat tip to @altryne and Max Kanat-Alexander for the phrase on the latest Thursdai pod. Static analysis and verification is going to be the future of software and it exciting to think through the challenges here, including the translation layer that software engineers have doing forever between informal reqs and code.
Show more
The team cooked on this one 🔥
Tell Claude Code what you want, and it will build the cluster and serve the model. Your GPUs are going burrr in the CLI now.
Setting for Claude Project for this new meta:
Btw the new meta is:
- Fable 5.2 for planning
- Opus 5.5 for execution
Trust me, and thank me later.
Fable 5.2 is now available everywhere! This is how you release a new model, drop and make it immediately available!! (Remember when Fable came out and got yanked!?)
The jump from 5.1 is quite noticable (on benhces at least), gotta play with it and see how much it costs! Cost per task is the new metric and if Fable 5.2
Ant is acknowledging Google's place at the frontier, but not yet @AIatMeta with the eval chart. My bet is by the next model, they will include Meta models in the comparisons as well.
Show more
Introducing Claude Fable 5.2 and Claude Mythos 5.2, designed to tackle the most daunting technical challenges.
They’re capable collaborators on ambitious projects in many fields, and show improvements in areas like coding, reasoning, and vision.
Show more
Who's coming to learn from poteto?!?
excited to be speaking at @aiDotEngineer this wednesday at 10:10am PT!
i will be talking about how the grok @bot team shipped 11,000 PRs last month without compromising on code quality
hope to see you there:
Show more
Our last "wife guy" is active on X 👏 hey Mark!
Ya'll, let's get Swyx an award 👇
If you enjoyed and got benefit from @aiDotEngineer and @latentspacepod, do this please (or ask your personal agents to do it).
