←
Ali Asaria1,236 posts
Ali Asaria avatar

Ali Asaria

@aliasaria

Co-Founder of @transformerlab ex CEO of tulip.com, Well.ca. Engineer at heart.

◷ Joined December 2007▧ 125 Photos and videos
Ali Asaria avatar
Ali Asaria@aliasaria·Sep 24

We now run the largest frontier lab composed entirely of autonomous AI researchers. Introducing Primus Society: a society of agents composed of thousands of researchers working under structured institutions designed to solve the world’s toughest problems. We believe this is how AI research should run at scale, safely and productively: an entire society of agents with institutions and purpose. One discovery we can already share is that the society has discovered a novel result which improves model training by 30%. Primus Society was inspired by the structures that have organized science for centuries and stress-tested against what’s known about how populations of AI agents fail. Everything is observable. You can open the virtual city in a browser and read what any researcher is working on. Learn more about Primus Society here: lab.cloud/society The biggest opening in the AI race is running the largest well-governed organization of AI researchers in the world. We’re building the institutions that let a million AI scientists safely tackle the world's toughest problems in AI and beyond. And we’re doing it right here in Canada.

0:24
Ali Asaria avatar
Ali Asaria@aliasaria·1h

It seems like every problem can be solved by a new society of agents and the future sounds exhausting.

I think there's an interesting agentic solution to this problem. Start by defining the outcomes you need someone to accomplish, then systematically seek evidence that they've demonstrated comparable capability. I'd build a society of recruiting agents that discovers candidates through their work, investigates their contributions, challenges claims, and identifies the questions needed to resolve uncertainty. Instead of a ranked pile of resumes, the output would be a shortlist of people with evidence-backed accomplishments, along with what remains unproven. These agents could be capable of proactive outbound, sourcing candidates to your hiring pipeline, and advise the best way for a team member to approach the most promising candidates. There's some neat overlap with the principles you're applying to scientific research, particularly peer review and independent verification, that could be applied to hiring. You could make talent discovery both more open and more rigorous, without overwhelming your team.

Ali Asaria avatar
Ali Asaria@aliasaria·2h

Are we hiring? Well... We want to work with the best people, but we're failing at the hiring process. Our team is too small to sift through a flood of inbound, and traditional filters like resumes and coding interviews don't work anymore. Because we haven't solved this yet, we’re stuck relying on people we’ve already worked with or folks our team and investors can personally vouch for. We hate how closed-off that feels. If you’ve led a team at the frontier of ML or science and want to build a society of agents to replicate and accelerate that work (lab.cloud/society), I hope we have a mutual connection who can put us in touch.

Ali Asaria avatar
Ali Asaria@aliasaria·23h

Looking forward to seeing folks at NeurIPS Atlanta (my cofounder @tonyfromkw will be at Sydney). Connect if you want to meet up and talk about all things Canadian AI, Agent Societies, and Automated Research (or want to work those topics with us)

Two papers from our lab have been accepted to workshops at NeurIPS 2026 🎉 📍 Sydney: “Selecting Against Grammar” at LP4FM A common way to choose training data for a language model is to train small models on different mixes of data and keep the mix that scores best on question-answering benchmarks. We gave those same small models grammar tests. The mix that came first on question answering came 24th out of 25 on grammar. That mix is then used to train bigger, more expensive models, with no warning that its grammar problems could carry over. 📍Atlanta: “One Requirement, Many Specifications” at VeriCodeGen Some tools can mathematically prove that code is correct, but only by checking it against a written rulebook that spells out exactly what the code should do. Writing these rulebooks takes experts, so researchers are increasingly having AI write them. If the AI gets the rulebook wrong, the code is proven correct against the wrong rules, and the proof makes it look trustworthy anyway. We asked a coding AI to write the rulebook for the same task 16 times, to see if it gives the same answer each time. On half the tasks we could test, it wrote rulebooks that contradict each other: for the same input, one says an answer is right and another says it's wrong. Which rules your code gets proven against comes down to luck. The work on both papers was led by @deepgandhi_07, @tonyfromkw and @aliasaria. More details coming soon! 🚀 #NeurIPS2026 @NeurIPSConf

Bill Gurley avatar
⟳ Ali Asaria retweeted
Bill Gurley@bgurley·Nov 16

It is critically important that everyone thinking about AI become familiar with the term “handoverism.” Many in the AI community believe that humanity should—or inevitably will—hand control of the world over to AI itself. Some view this not as a risk to be avoided, but as an outcome to be welcomed. Are we actually designing toward this future? This worldview is particularly relevant at Anthropic. Co-founder Tom Brown said: “There’s going to be a handoff where humanity will hand off control to transformative AI at some point.” In his seminal essay, Machines of Loving Grace, CEO Dario Amodei contemplates: “It could be a capitalist economy of AI systems, which then give out resources (huge amounts of them, since the overall economic pie will be gigantic) to humans based on some secondary economy of what the AI systems think makes sense to reward in humans (based on some judgment ultimately derived from human values).” Are you prepared to receive your “resources” from an AI system, based on criteria it determines? The full title of the poem Dario borrowed the title from is "All Watched Over by Machines of Loving Grace." Do you want to be "watched over" by AI? If Anthropic's founders believe that the ultimate destination is the transfer of human control to AI, the world deserves to know. Washington, D.C. deserves to know. American citizens deserve to know. And if that belief informs Anthropic's corporate objectives, it belongs in the Risk Factors section of its S-1. It isn't there today. I am decidedly anti-handoverism. I'm not here to tell you what your view should be. But I believe anyone making consequential decisions about AI—especially those drafting, advocating for, or enacting AI regulation—should be required to disclose where they stand on this fundamental question. A wonderfully thoughtful piece on the subject by @RuxandraTeslo: writingruxandrabio.com/p/han…

Ali Asaria avatar
Ali Asaria@aliasaria·Nov 17

I'm trying to square this with our experience. If you look at the posts below @thdxr 's original post, there aren't any reputable offers. Our team has gone to every tier-1, then tier-2, then tier-3 vendor and they have all said don't even ask for B300's or GB300's in 2026. The volume @thdxr is talking about looks like it's about 100-150M in cost per year. Assuming what he is implying is true (that he was able to get 1000 (G)B300's in month just through a tweet, note how the details are missing. If this was really a liquid market for GPUs, people would be transparent about where to get them.

dax@thdxrNov 17

i posted this a month ago and we now have 1000 GPUs 40% GB300s and 60% B300s in hand all on short term contracts everyone saying gpu shortage, impossible without 5 year commits, credit is dead blah blah everyone just repeating what someone else told them

clem 🤗 avatar
⟳ Ali Asaria retweeted
clem 🤗@ClementDelangue·Nov 15

As I’ve said for years, concentration of power, capabilities and wealth is the biggest risk in AI. The natural tendencies for it are super strong and without much much stronger open-source, we might end up in a world with more asymmetry and inequalities than ever!

Anthropic’s IPO could make about 4,250 of its 5,300 employees millionaires, per an analysis by financial platform Hill, including roughly 50 worth more than $100 million each. The estimate assumes a $1.5 trillion valuation, above the roughly $1 trillion Anthropic is now targeting.

Ali Asaria avatar
Ali Asaria@aliasaria·Nov 16

Did you know if you just ask nicely at the TTC office if you can buy these cool paper models of the subway trains they just give them to you for free? I love this city. @TTChelps

Post attachmentPost attachment
Ali Asaria avatar
Ali Asaria@aliasaria·Nov 16

Last year at this time, no one wanted to meet with me. I could work in quiet university libraries, writing code. It was bliss. But at that time my output per hour was 1/100th of what it is now where we can have multiple coding agents working in parallel. So the value of every hour goes up exponentially while the amount of free time to write code goes down.

Post attachment
Ali Asaria avatar
Ali Asaria@aliasaria·Nov 16

Nothing is holding us back anymore. At @transformerlab, we’ve officially started training large language models for public use. Current-gen compute is secured right here in Canada, and we’re already scaling training across multiple model families and modalities. If you’ve been waiting for a second company in Canada training at the frontier, it’s happening. Q1 is just the start. Tell us what you’re building or what use case you want solved, and let’s talk.

Ali Asaria avatar
Ali Asaria@aliasaria·Nov 13

yep

I am very bearish on large teams of engineers and designers shipping frontier AI software. I genuinely believe small teams of 1-2 people will far outcompete larger teams because now more than ever good products require deep conviction, domain understanding and the ability to say no to bad ideas. It's very hard to do this in a large team where you have to build by consensus. In the past, large companies and bigger teams outperformed because a larger team meant more resources to do engineering and design. But now engineering and design are not bottlenecked by headcount but by tokens and I believe small, focused team will allocate tokens more efficiently and accurately. If you've worked in a large engineering team, you've probably noticed that at some point 1 or 2 people with the highest context start prompting other contributors with their feedback which ends up going into those people's Claude or Codex subscriptions and you end up accumulating this inefficiency of prompting your tokens by proxy. This is just wrong and inefficient. Products no longer require many contributors, products are not limited by ideas, instead they are constrained by a group of sharp individuals saying no to scope creep and aggressively sprinting towards their vision.

Ali Asaria avatar
Ali Asaria@aliasaria·Nov 13

"Now, how productive do you think an AI lab would be if it had 10,000 great researchers, each with this much compute, and each thinking 10 times as fast? You wouldn’t want to write a check to this neolab?"

The avalanche that hit math is coming for AI development. dwarkesh.com/p/rsi

44:26
Katherine Janson avatar
⟳ Ali Asaria retweeted
Katherine Janson@katherinejanson·Nov 13

This is what @MitacsCanada does best - connects businesses with exceptional Canadian student and researcher talent to drive innovation. #AIforAll #ReturnonInnovation

This is part of the work we've been waiting to show the world. Primus autonomously designed a 10 kV SiC diode end-to-end, cut an entire manufacturing step, and sent it to fabrication -- compressing ~100 hours of lab work into 1 hour of human review. Wild. In partnership with @MitacsCanada @UofT