Founder @Venture4Canada. Curious person passionate about helping others achieve their full potential.
Toronto
Joined June 2010
- Tweets1.5K
- Following915
- Followers1.7K
- Likes772
Scott Stirrett retweeted
Detroit impressions:
• The downtown is full of beautiful buildings. All of them seem to have been built specifically in the 1920s. I guess that is after the city had accumulated enough auto wealth but before the twin hits of Modernism and the Depression. (I hadn't known that the GM Renaissance Center, built as a revitalization project, was at the time the largest private development in US history, and also at the time the world's tallest hotel. It may be large, but it is not pretty.) The downtown is surprisingly depopulated -- both the streets and the sidewalks feel empty. That said, it didn't feel at all unsafe. There are lots of great homes in the suburbs.
• The Henry Ford Museum of American Innovation is amazing, and it's worth visiting Detroit for it alone. Among many (many) other things, it contains the oldest known surviving steam engine in the world, the actual Montgomery bus on which Rosa Parks refused to give up her seat, a deconstructed Model T, a deconstructed Eames Chair, and many great cars, agricultural equipment, locomotives, industrial specimens, and more. (They have the Lincoln Continental that JFK was riding in when assassinated -- which, apparently, was returned to service and used by several subsequent presidents.)
• The museum made me wonder why American car design peaked in the mid-60s. (This fact is very evident at the museum.) The LLMs blame the 1966 National Traffic and Motor Vehicle Safety Act. (Not quite wtfhappenedin1971.com, but close.)
• Good food exists but it is hard to find.
• The Heidelberg Project also exists and is unique.
• We stayed at the Dearborn Inn, which is wonderful, and contains cottages modeled after the homes of significant American figures. Dearborn (and Hamtramck) are now predominantly Muslim, apparently for reasons that go back a century to Henry Ford's $5 wage. Dearborn felt noticeably prosperous (we stopped for coffee at a fancy Japanese cheesecake cafe); Hamtramck did not.
• Michigan.gov says that the Hispanic population of Michigan is just 6%. Coming from California, the absence is very striking.
• The Detroit Institute of Arts is remarkable, particularly the room with the American landscapes and the section with the Dutch masters (especially The Visitation). An obvious question is why there is nothing quite like it in the Bay Area given how much richer the latter is than Detroit ever was -- we techies are just so uncultured by comparison. The Diego Rivera murals are amazing (and quite strange; you can see why they were controversial).
• Detroit is full of historic plaques -- they are truly everywhere. This is presumably due in part to the fact that Detroit has a lot of history, but it still has many more than places with comparable historical depth. Some research suggests that it might be related to generous tax credits for historic preservation. Whether or not that is true, Detroit persuades me that other places should engage in more plaquemaxxing.
• I recommend a visit! You overall leave with some sense for how exciting America must have felt in the early 20th century.
Scott Stirrett retweeted
The inability of AI models to produce creative variation is a huge gap. The fact that they generate similar ideas limits their ability to do science & the same-y writing limits their usefulness in many other applications
This paper showed you can optimize models for creativity
Scott Stirrett retweeted
Excellent essay from @danbjork on the implications of AI for low and middle income countries—a critical topic that deserves way more attention.
dan.bjorkegren.com/blog/2026…
Scott Stirrett retweeted
🚨BREAKING: Anthropic just published a study mapping exactly which jobs its own AI is replacing right now.
The workers most at risk are not who anyone expected. They are older. They are more educated. They earn 47% more than average. And they are nearly four times more likely to hold a graduate degree than the workers AI is not touching.
The argument is straightforward. Anthropic built a new metric called "observed exposure." Not what AI could theoretically do. What it is actually doing right now in professional settings, measured against millions of real Claude conversations from enterprise users.
For computer and math workers, AI is theoretically capable of handling 94% of their tasks. It is currently handling 33% of them. For office and administrative roles, theoretical capability is 90%. Current observed usage is 40%. The gap between what AI can do and what it is already doing is enormous. The researchers are explicit about what comes next. As capabilities improve and adoption deepens, the red area grows to fill the blue.
The demographic finding is what makes the paper uncomfortable. The most AI-exposed workers earn 47% more on average than the least exposed group. They are more likely to be female. They are more likely to be college educated. This is not a story about warehouse workers or truck drivers. It is a story about lawyers, financial analysts, market researchers, and software developers. The exact group whose education was supposed to insulate them.
Computer programmers showed the highest observed AI exposure at 74.5%. Customer service representatives at 70.1%. Data entry keyers at 67.1%. Medical record specialists at 66.7%. Market research analysts and marketing specialists at 64.8%. These are not predictions. These are measurements of work that is already happening on AI platforms right now.
Then there is the pipeline finding nobody is talking about loudly enough.
Anthropic's researchers found a 14% decline in the job-finding rate for workers aged 22 to 25 in highly exposed occupations since ChatGPT launched. No comparable effect for workers over 25. Entry-level roles were never just jobs. They were the training ground where junior analysts became senior analysts, where junior lawyers learned how arguments hold together. If that layer disappears, nobody has answered the question of where the next generation of senior professionals comes from.
The detail buried in the paper that most coverage missed: 30% of American workers have zero AI exposure at all. Cooks. Mechanics. Bartenders. Dishwashers. The technology reshaping professional careers is completely irrelevant to roughly a third of the workforce. The divide is no longer between high skill and low skill. It is between presence and absence.
The company publishing this study is the same company selling the AI doing the replacing. Anthropic had every commercial incentive to soften these findings. They published them anyway.
If you spent four years and $200,000 on a degree to land a white collar career, the company that builds Claude just confirmed your job is more exposed than the bartender pouring drinks at your graduation party.
Source: Anthropic, "Labor market impacts of AI: A new measure and early evidence"
PDF: anthropic.com/research/labor…
Scott Stirrett retweeted
Very cool analysis of the submissions to a major management journal that shows how much the system of science, built for humans, is under strain as a result of AI.
AI can be used to do better science or it can be used to just do more stuff. The danger is that "more" is winning
Scott Stirrett retweeted
I have news to share:
I am writing a book, currently untitled, set to be published by Penguin Press next year.
I began my public writing career with the thesis that AI and related technologies will both necessitate and enable major changes to the institutional configuration of Western society. I believe it is possible for individual liberty, dignity, and property to survive the coming transformation, but that their survival will take serious collective effort.
This book is a contribution to that effort. It is an attempt to describe a positive vision for the future of the American republic in particular, and all ordered liberty worldwide. But it will simultaneously be a diagnosis of a crisis: open societies have drifted from the principles they once embodied, and those principles must be re-imagined for modern ears if we are to embody them in the future. This will therefore be a work of political theory just as much it is an "AI book."
There is not much more I can say about the project (yet), other than that it is, by far, the most ambitious work I have ever undertaken.
My research and writing will not replace regular writing on Hyperdimensional, but the cadence of publications is likely to remain lower than it once was. I anticipate writing essays every 2-3 weeks for the foreseeable future. I expect my output to shift a bit more toward mini-essays posted on X, which is a format I have been using more anyway.
I am immensely grateful for my readers' support over these last two years. Thank you to everyone who has taken the time to read my work, and to the countless people who have lent me a hand along the way.
More to come soon.
Scott Stirrett retweeted
We are learning whale using science
New paper out in Proceedings of the Royal Society B: we apply linguistic tools to sperm whale vowels.
The result: sperm whale vowels do not just look like human vowels. They also behave like them.
We found several parallels. Like in Latin, whales have short and long vowels. Like in Slovenian, some vowels prefer particular tones. Like in human language, there’s a lot of coarticulation (a process when you say “tense” but the word sounds like “tents”).
Observing vowels in whales is a matter of timing. Our vowels are fast, whale vowels are slow. Beats become pitch if they’re fast enough. If you slow down human vowels, they start sounding like whale clicks.
Applying linguistic tools to whales shows us that we’re much more similar to these wonderful ocean creatures than we previously believed and that their language is much more complex and structured.
@projectCETI @UCBerkeley
Scott Stirrett retweeted
Really great news I would say:
"Social media is populist and polarising. AI may be the opposite."
– @jburnmurdoch in the FT
Scott Stirrett retweeted
Some other points worth making:
1. A lot of people, including people in positions of authority, told us recently that models of Mythos capabilities wouldn’t be a thing—that models with obvious “national security” implications would not be forthcoming. Those people were wrong. There’s nothing to “do” about this. But you should remember it.
2. Mythos is the first model where the theft of the weights by an adversarial actor feels like it would be a major deal. You better believe they will try, and if they don’t succeed with Mythos, they will eventually.
3. We are thoroughly in the era of “the labs’ best models may well not be public in the way we are used to.” This will be because of a combination of compute constraints, economic reality, competitive advantage, and safety concerns.
4. (3) means that the most relevant models may be decreasingly legible to the general public over time. We might all have less of a sense of what is going on. This is where transparency, and eventually auditing, can come in handy. Internal deployments were one of the major motivations for my work on “entity-based” regulation (we have to regulate the company, not the model, because the details of the model itself are fuzzy if it isn’t released), and my work on private governance: independent bodies that would evaluate frontier AI company safety practices, including internal deployments.
5. Depending on the extent and duration of the coming compute squeeze, we could enter a market dynamic where the best models are only available to the highest bidder—in other words, where compute is a sellers market rather than a buyers market. Imagine competing firms in the economy bidding against one another for access to the best and most tokens, and the frontier labs as, in essence, kingmakers. The governance regime I have described above in (4) is not designed to stop this dynamic, but for obvious reasons policymakers and the broader public may find it unpalatable. On the other hand, if the compute squeeze is *that* bad (not a guarantee, and I rate this unlikely to be clear), it probably means malicious actors struggle to find tokens too. That is good for “safety”!
Some brief thoughts on Mythos
We’ve known this was coming for a long time. At least, we *should* have. Extremely effective software vulnerability discovery was clearly coming to anybody paying attention.
It has also been clear that all AI policy so far has been made and executed with training wheels. It was always clear that, sometime soon, the training wheels would come off.
The training wheels aren’t fully off just yet—this model is being kept under lock and key, and Anthropic does not seem inclined to release Mythos preview to the public anytime soon, if ever. The training wheels will be off when these capabilities are fully diffused in ways centralized actors cannot control. It is inevitable that this will happen. The point is not to argue about whether we should “ban open source” or similarly unrealistic notions. The point is to harden the world for this new reality.
I applaud Anthropic—and I especially applaud @logangraham—for doing so. But their efforts alone are not close to enough. Project Glasswing—a partnership with Anthropic and other companies—seems nice, but unsurprisingly it lacks uniform frontier lab participation.
It would probably be ideal, for our national cyberdefense, if the federal government were not trying to destroy Anthropic and eliminate their models from government systems. If anything, the government should be trying to work more closely with Anthropic. As a side note, I hope Anthropic is working with state and local government entities on cyber vulnerability discovery, since many of our adversaries know that state and local is America’s soft underbelly in so many ways.
In any event, the Mythos news should lay bare how stupid and counter-productive the Department of War’s feud with Anthropic really is. As someone who suspected all this was coming (not from inside knowledge but from it being ~obvious), that probably explains why I have had such a strong reaction to that feud. It’s this senseless distraction just at the time that the training wheels are coming off. I hope the two parties can resolve their differences now, for the sake of the country, but I am not hopeful.
I do want to call out, however, the numerous political and career civil servants in the Trump Admin who do get these issues, know how stupid the Ant-DoW stuff is, and want to work with the frontier labs like adults. I wish you all utmost success.
I find myself inclined to end on some positive notes. Mythos appears to be—according to Anthropic at least—“the most aligned” model Anthropic has ever trained. We are approaching superhuman capabilities in some domains, and yet alignment is getting better rather than worse. That’s not nothing. I know some of you think the model is faking its alignment, or aware when its alignment is being tested. I don’t have a good answer.
Finally, there is this: Mythos was made by an American company, and like most successful American companies, it has a vested interest in maintaining order and peace, and it is investing substantial resources in mitigating the risks of its technological progress, as I expect most of the American labs would. This is cause for optimism: The incentives of capitalism are working.
The training wheels are coming off, but at least we are the ones removing them, as opposed to our enemies. Perhaps we can be the first to learn to bike for real. The first step would be to get beyond all the low-fidelity, under-specified, pimply little fights of AI policy’s prepubescent era. That goes for me too.
“What hath God wrought,” wrote the first telegram. What, indeed. In this case, the answer is still up to us.
Scott Stirrett retweeted
Excellent opinion piece in the Globe & Mail about the need - and desirability - of much closer ties between Canada and Australia. Both Commonwealth countries with strong democracies,
tons of vital mineral deposits, and more. theglobeandmail.com/business… #Canada #Australia
Scott Stirrett retweeted
When Canada and Australia team up, nobody can stop us theglobeandmail.com/business…
Scott Stirrett retweeted
US productivity growth is likely to come in at about 2.7% for 2025.
That is nearly double the average of the previous 10 years.
There are many factors at work, but part of the story is that businesses are finally beginning to reap some of AI's benefits.
I discuss the latest evidence in my column in the @FT this morning.
See ft.com/content/4b51d0b4-bbfe…
Scott Stirrett retweeted
The FT just quantified what the “vibe code is a toy” crowd has been ignoring for 12 months.
New websites up 30%+ YoY. iOS app submissions up nearly 60%. GitHub pushes in the US and UK spiking to levels never seen in the data. All four charts show the same inflection point: late 2024, right when Claude, Cursor, and Replit Agent hit mainstream adoption.
The scale of what’s happening is hard to overstate. 25% of Y Combinator’s Winter 2025 batch reported codebases that were 95% AI-generated. Replit says 75% of its users now have zero coding background. A solo founder built Base44, sold it to Wix for $80M cash after six months. Lovable raised at a $6.6B valuation and is targeting $1B in annual revenue by late 2026.
This is real shipping. Real apps in the App Store. Real websites with real domains. The skeptics kept framing vibe coding as “demo code that never sees production,” and meanwhile the production numbers were quietly going vertical.
The part most people skip: this data probably understates the shift. Tons of internal tools, prototypes, and MVPs built with AI never touch GitHub or the App Store. They live inside companies, replacing processes that used to require a dev team and a 6-week sprint.
Collins Dictionary made “vibe coding” their Word of the Year in 2025. When the dictionary notices before the industry analysts do, you know the discourse is lagging the data by about 18 months.
Scott Stirrett retweeted
Underrated factor in why English-speaking countries have especially bad housing crises is their common law systems (adversarial and litigious) vs judge-led civil law systems elsewhere.
Makes Anglo planning/permitting systems especially vulnerable to NIMBYs and other objections.
Scott Stirrett retweeted
“Roughly 35 percent of Canadians aged 20 to 34 lived with at least one parent in 2021. Meanwhile, the share of 20 to 34 year olds living with a spouse, partner and/or child without their parents has declined from almost 50 percent in 2001 to under 40 percent in 2021.”
.@Sean_Speer and @TaylorJJacks: The fragile state of the Canadian family in 5 charts
thehub.ca/2026/02/16/diverse…
Scott Stirrett retweeted
Fascinating insights from senior engineers on how AI is changing their jobs. Interesting how automation also creates all sorts of new tasks and bottlenecks. thoughtworks.com/content/dam…
Scott Stirrett retweeted
The transformation of Toronto, Canada 🇨🇦
ALT All rights of the first image belong to Bill Thomson.
Scott Stirrett retweeted
I think this was a really bad year for American politics, a mediocre year for the American economy, and an exceptional year for America.
In 12 months, you’ve got
- the largest decline in murder rate ever recorded
- huge declines in traffic fatalities, drug overdoses, and suicide
- first ever personalized gene editing treatment and breakthroughs in HIV and cancer therapy
- continued advances in GLP1 technology that seems to reduce weight and inflammation and a bunch of other stuff
- declines in teen anxiety and despair
- surge in self-driving car technology*
- all this happened in a period when both the SP500 and inflation adjusted median wages hit record highs
Scott Stirrett retweeted
Most divides aren't due to differing values. They're driven by focusing on different things.
After a 10-minute talk, people with opposing views are less polarized a week later. They find more common ground than expected.
Prejudice festers in silence. Civil dialogue opens minds.
