Addie Foote retweeted
it’s crazy that some think scary ai loss of control incidents are secretly a marketing ploy. any tiny marketing benefit is entirely outweighed by increased pressure for regulation and fear of ai, which ai cos hate.
car crashes don’t help sell cars, school shootings don’t help gun companies, 9/11 didn’t help airlines, …
A major problem with managing unemployment from AI might be people's inflexibility. Anecdotally, I know someone living off their wife's parents for several years while searching for a specific position and salary.
Addie Foote retweeted
“UBI” is obviously nowhere near the panacea many of you seem to think it is. The median left-leaning Westerner isn’t angry at Elon Musk because he can buy a million times more groceries than them. They aren’t upset with Palantir because Peter Thiel can afford to eat a thousand burgers to their one. This whole thing is in large part post-material. It’s the hierarchy & subordination they’re uncomfortable with. They feel their dignity is being trampled and their autonomy progressively diminished – rightly or wrongly they feel politically disenfranchised and stripped of a say over the future. Offering a guaranteed food budget and a pod to spend the night in return for further disempowerment is incredibly tone-deaf and should be expected to provoke more, not less, outrage.
It was great to work at Workshop labs! I can't wait to see what's next for them and for me :)
Workshop Labs is joining @thinkymachines.
We believe there's a path for AI to make humans matter more. We couldn’t be prouder to join Thinking Machines to see this work through.
workshoplabs.ai/blog/wsl-joi…
I agree with this. It allows allows for broader transparency and response to current capabilities and external verification of claims
Addie Foote retweeted
Third: how we made post-training and inference happen so privately that even we can’t your data.
Very excited about this! Hopefully it becomes an industry standard
xitter.cf/WorkshopLabs/status/20…
optimizing is so much fun
Replying to @WorkshopLabs
Starting at 17.6 sec/step with expert parallelism, we implemented:
➡️ Grouped mm → 13.7s
➡️ Vectorized expert dequant → 8.8s
➡️ torch.compile → 4.9s
➡️ Skip trailing pad tokens → 2.9s
➡️ Sequence packing → ~10x more tokens/step (~2x longer steps)
Addie Foote retweeted
Introducing Trellis for Kimi K2 Thinking.
It's post-training code that's 50x faster than the best single-node open-source version and 2x cheaper than training APIs.
After safety testing, we're open-sourcing it, giving builders the best tools to customize a frontier model.
🧵
Addie Foote retweeted
"If you have a bug, it's your fault" is good advice for most code.
But it's terrible advice for open-source ML infra. Open weights isn't open training.
Read our post on what we tried and why we built our own training stack instead:
workshoplabs.ai/blog/open-we…
Addie Foote retweeted
Open weights isn't open training.
@AddieF38654 on our team wrote up her experience trying to post-train a 1T parameter MoE model using the existing open source infra.
Let's find out how many monkey-patches it takes to post-train an open-weights model. A thread🧵
Addie Foote retweeted
I don’t think this is the right way to think about it. My takeaway is that well defined software tasks—that is, all tasks in the METR study—will become ~completely automated soon.
Less well defined tasks/those in messier codebases/those that require non-SWE related expertise (
I'll be at NeurIPS this week!
Bitter Lessons from "Distillation Robustifies Unlearning" (@ NeurIPS 2025 - Spotlight)
My collaborators and I wrote a paper titled "Distillation Robustifies Unlearning" a few months ago.
Our team will be presenting posters at NeurIPS and the San Diego Alignment Workshop, so feel free to find me if you want to chat!
I’m writing this to communicate what I think our paper actually says about the problem of unlearning.
This is more of a personal account than a group statement!
neurips.cc/virtual/2025/loc/…
Addie Foote retweeted
A cool technique from the RunRL research team: using PD controllers to balance multiobjective loss functions!
Suppose you want to train a model to give short yet relevant answers. For a given minimum level of relevance, this lets us improve on the other reward terms much more!
Addie Foote retweeted
can we finally use natural language to optimize for deeper notions of what users want from their recommender systems?
Addie Foote retweeted
Thought real machine unlearning was impossible? We show that distilling a conventionally “unlearned” model creates a model resistant to relearning attacks. 𝐃𝐢𝐬𝐭𝐢𝐥𝐥𝐚𝐭𝐢𝐨𝐧 𝐦𝐚𝐤𝐞𝐬 𝐮𝐧𝐥𝐞𝐚𝐫𝐧𝐢𝐧𝐠 𝐫𝐞𝐚𝐥.
Addie Foote retweeted
How do we boost sim2real transfer without tedious simulation tweaking, system ID, or expensive domain randomization?
Language can serve as shared grounding between domains!
Introducing our #RSS2024 paper: "Natural Language Can Help Bridge the Sim2Real Gap"
(1/6)






















