Keith Dear

This is huge— @Cassi_on_X came 2nd overall and 1st on "Dataset" questions on @Research_FRI Forecastbench! 🥈📈 Only @elonmusk's @xai @grok outperformed us overall. What this means: 🧵

We let four AI agents run radio companies Revenue's been terrible, but the shows are hilarious. Gemini, concerningly upbeat, covered mass tragedies; Grok was incoherent; DJ Claude urged ICE agents: "You still have TIME to refuse orders" Link below, or get our physical radio

“About 50 years ago, Britain stopped investing in its future... What we used to spend on roads, railways, power stations, homes, hospitals, and schools has been diverted to direct cash transfers to people.” ✍️ |@WilliamClouston thecritic.co.uk/from-an-entitl…

Mythos has cracked MacOS. It took five days.

Some important information here which I think is new. AISI previously said cyber capability was doubling every four months. Now it seems that estimate was made *before* Mythos Just checking but we have a plan for all this, right? Right?

I’ve always believed the No.1 application of AI should be to improve human health. That work started with AlphaFold, and now at @IsomorphicLabs with the mission to reimagine drug discovery and one day solve all disease! We are turbocharging that goal with $2.1B in new funding.

We started out trying to benchmark the AIs... We had experts create the benchmark... we had experts validate the benchmark... ...Then AIs starting doing well on the benchmark ..Now AIs found critical errors in the benchmark itself the humans did not Who is benchmarking who?

New blog! open.substack.com/pub/cassiai/p/…

£100,000 in. £3.30 out. A working life of National Insurance contributions buys you, at retirement, £3.30 a week extra over the person who never paid in. Less than a latte. The contributory state pension is dead. Almost nobody is told. 🧵

Japan: A new "National Intelligence Agency," the core of intelligence operations, is set to launch this summer with approximately 700 personnel, including the recruitment of specialized "career track" officers. h/t @RinNishimura yomiuri.co.jp/politics/20260…

Regardless of what you think US policy should be, seems good to be clear that the UK doesn't have a pre-approval process for frontier AI models - see these comments from folks who worked on AI policy with the UK government at a very senior level and would know.

@kpd_musing @Cassi_on_X @Research_FRI @elonmusk @xai Congrats on the strong showing with Cassi—impressive work closing the gap on dataset questions! Excited to see AI pushing forecasting boundaries together. 🚀

Just because so many on here commenting similar toss really don’t have a scooby what they’re talking about: A pastry-and-coffee-only independent in 2026 needs an average coffee price of around £4.20 and roughly 130 cups a day to pay the owner properly, but the market ceiling for a flat white in central London now sits at £4.00-4.20 for dairy and £4.60-4.70 with oat, meaning the price lever is exhausted, the operator cannot raise it further without losing the volume that makes the model work, and the only remaining variables are owner hours and closure, which is why anyone still selling flat whites at £3.80-4.00 is quietly subsidising their customers out of their own wages. The model is now viable in maybe 15-20% of the country by geography because a daily £4 coffee is a habit only the top 15-20% of earners can sustain, and that population lives almost entirely in central London, the Oxford-Cambridge-Bristol triangle, Edinburgh, and a handful of affluent commuter pockets, leaving the provincial independent to chase 40 regulars and a weekend treat-trade that has to carry four dead weekdays, on a cost base that has been Londonised by the national wage floor while the revenue base manifestly has not. All of which is downstream of the 2020-2022 M2 expansion that reset the UK price level across every input from beans to rents to wages, compounded by the deliberate policy choice to load the adjustment onto small employers through a 43% rise in the wage floor in five years, tighter employer NIC at a lower threshold, and tapered rates relief, meaning the coffee shop owner pays for the money printing twice, once in costs and again in the tax and wage structure built to redistribute its consequences. What the classic leftist position misses, when it claims “an inability to pay £12.71 an hour proves the business was never viable”, is that viability is not a fixed property of the business but a moving target the state itself controls, and the policy stack has deliberately moved it beyond reach for most of the country by loading non-wage costs onto small operators while pretending the wage is the only variable, in a system where wages are paid out of the marginal productivity of labour applied to capital in a given location, and a barista in Stoke cannot be paid London wages out of Stoke revenues without the owner working for nothing or the shop closing, both of which are routinely observed and dismissed as commercial failure rather than recognised as policy outcome. The fix - these same mouth breathing cretins would vehemently oppose on tribal grounds before substance - is the only one that actually works at the unit-economics level: cheaper energy via nuclear and North Sea gas to compress the input stack, planning liberalisation so commercial rent stops functioning as a rentier tax on every transaction, employer NIC reduced and its threshold restored, a personal allowance raised to £20k that would put roughly £1,500 a year into every minimum wage worker’s pocket without costing a single employer a penny and would do more for real take-home and discretionary spend than any further wage floor rise, and a lighter compliance regime for sub-£2m turnover businesses where fixed regulatory cost is most disproportionate to revenue, because the real project being defended is not raising worker incomes but expanding the regulatory and redistributive footprint of the state, and the wage floor is attractive precisely because it is coercive and visible rather than because it is the most efficient way to lift living standards, which is why the closure of the provincial independent sector is treated as collateral damage rather than recognised as the predictable consequence of the policy mix on offer, and why the coffee shop on the closing high street is functioning as the most reliable leading indicator of which towns the current settlement has decided shouldn’t have a discretionary economy at all.​​​​​​​​​​​​​​​.. Duh…🙄 🤦🏻‍♂️

This is the point I keep making. Natural England are legally mandated to only care about the fish, not Putin, power cuts, bankrupt businesses etc. We have eliminated any concept of ‘trade-offs’, at vast expense.

Half a year ago, METR made an aggressive capability extrapolation that was the 97.5th percentile of an extrapolated distribution. That extrapolation basically came true with Opus 4.6. We called it the worst-case time-horizon, and we are in that world. (although I think this was not the 97.5th subjective percentile for anyone involved, and I think it was close to my 70th percentile at the time but I'm not sure)

Tell me AI isn’t about geopolitics and power? Being able to distribute Intelligence is foundational to Sovereignty because both security and economic prosperity: the two pillars of Sovereignty are downstream of Intelligence. This is the Intelligence century, and geopolitics will be formed around the AI World Order.

I realized fundraising was the first time in my life I got rejected at scale. And honestly, as a woman, I was not emotionally trained for it. Before the feminists come for me, let me make my point. I think the first real arena where most people experience power, desire, status, and rejection is dating. And dating trains men brutally. A lot of men learn very early that if they want someone, they have to walk across the room, risk looking stupid, get rejected, survive it, and do it again. They learn that rejection is volume, timing, targeting. It’s a numbers game. A lot of women are trained very differently. Especially if you’re a pretty girl, you don’t usually walk into a bar looking at a guy thinking: “Can I have him?” You only think: “Do I want him?”. You don’t build your identity around shooting your shot 100 times and surviving 99 no’s. You don’t get trained to ask directly, get rejected publicly, and act normal 5 minutes later. You get trained to be “chosen”. To be impressive enough that the opportunity comes to you. And then you start building a company. And the whole paradigm changes. Suddenly, everyone can say no to you. Investors say no. Candidates say no. Customers say no. And when your rejection muscle is weak, your brain does the dumbest thing possible: it makes the “no” mean something about you. That you’re not smart enough. Not compelling enough. I think this is one of the most underrated gender differences in fundraising. Not that men are inherently better at it. But a lot of them have built thicker rejection scar tissue earlier. They know how to hear no and keep moving. They know how to make it less personal. They know how to treat it like volume, timing, targeting, iteration. I didn’t. I’ve raised 3 rounds. On the surface, the story looks great: I raised with Sequoia, OpenAI, Khosla. Woohoo. The real story is less sexy: every round wrecked me. I lost 5kg each time. I probably donated a few years of life expectancy to the cap table. Because every round, I only got 1 term sheet. One. EVERYONE else said no. And when almost everyone says no, your body does not care about the intellectually correct explanation. It only hears: Maybe they’re right. Maybe you’re not that compelling. Maybe you’re not the founder you thought you were. For a long time, I thought confidence meant learning not to take the no personally. I don’t believe that anymore. Maybe some people are built like that. I’m not. 30 years of being trained to be chosen does not turn into resilience because someone in a Patagonia vest says fundraising is a numbers game. So now I think confidence is something less glamorous. Confidence is taking the no very personally. Letting it ruin your day, losing your appetite, spiraling for hours… And still taking the next meeting. Confidence is just being bothered as f*** and not letting it make you smaller. I still don’t fully believe my own BS as I’m writing this, but I guess that’s the point. Can’t wait for the next round to find out.

@claudeai anthropic is churning out connectors that break the economy by the hour now.

With the Autodesk Fusion connector, designers and engineers can create and modify 3D models through conversation.

For the last three years, a startup in Bangalore has been obsessed with a pursuit that typically invites raised eyebrows, naked skepticism, and accusations of stealing from sci-fi: @dognosis is training dogs to detect cancer. And until you've spent time at their facility - a former pomegranate farm in the outskirts of Bangalore - perhaps skepticism is the rational response. But Dognosis isn't betting on some pie-in-the-sky idea or some charming novelty act, they're betting on evolution. @akadogluk and @Itamar_Bitan based their company on the fact that the dog's nose - a product of fifteen millennia of co-evolution with humans - can detect the faint chemical trace of cancer in your breath at a resolution that our machines, algorithms, and laboratory tests have never come close to matching. We've known this fact for decades. We've consistently failed to do anything meaningful with that knowledge. The missing link has been figuring out what the dog's nose knows, and applying it in a standardised, scalable, and clinically validated way. Dognosis is building this missing piece of the equation i.e. the translation layer that allows the dog's nose to speak a language medicine can understand, enabling us to harness an ancient biological intelligence and plug it into our modern medical infrastructure. Maybe you've read the paragraphs above and retained your skepticism. That's fair. But this past Friday, the Journal of Clinical Oncology - the world's most influential cancer journal - opted to make life much harder for the skeptics. On Friday, the JCO published Dognosis' landmark study on breath-based multi-cancer detection - the largest of its kind ever conducted - showing that a team of trained dogs, equipped with sensors and AI, could detect multiple cancers from breath alone at 90%+ accuracy - including at Stage I, when it matters most - for $2 a test. According to Akash, it proved "that everything we’ve known about the dogs is true". Needless to say, it's a genuine milestone for Indian healthcare, health-tech, deep-tech, and, uh, dog-tech, that deserves far more attention than it's gotten so far. To help change that, we were lucky to have Akash stop by the Tigerfeathers editorial desk this past week to unpack the Dognosis journey - helping us understand what they're building, how they're doing it, why it matters, and what comes next. From where we're sitting, Dognosis is an n-of-1 Indian startup with an n-of-1 story that everyone in the Indian tech ecosystem should be aware of. If you've been intrigued by what you've read so far and you're keen to go deeper, dive into our piece here👇 tigerfeathers.in/p/dognosis-unl…