16 September 2026 (edition 2)
Reviewed 19 episodes, 22 hours of podcasts, to find them.
Investment
Conversations with Institutional Investors
Matt Whineray, then chief executive of the New Zealand Superannuation Fund.
Whineray joined NZ Super in May 2008 looking after private markets and was chief executive for five of his nearly 15 years there, leaving at the end of 2023. On his account the fund took its first NZ$2.4 billion contribution in September 2003 and returned about 9.5 per cent a year over the 20 years to 2023, beating the cost of the government debt it displaced by more than NZ$40 billion and its own reference portfolio by about NZ$16 billion.
-
1 of 3 framework
New Zealand Super's tilting programme ran a short Kiwi position to nearly 40 per cent of the fund's net asset value in 2013, with the mark-to-market loss at its worst when the expected return was at its best.
The fund keeps adding exposure as price moves further from its estimate of long-run equilibrium value, so the two ways Whineray says it kills you are stopping out, after which the loss cannot be recovered, and spending the whole risk budget before the market reaches the level you were positioning for. He cites Cliff Asness's sizing rule: do not size a strategy so that when it goes wrong you are dead.
-
2 of 3 framework
New Zealand Super funds every unlisted purchase by selling the matching slice of a notional reference portfolio, so a forest has to beat the equities and bonds sold to buy it.
The board sets risk tolerance by choosing what goes into that simple passive benchmark, global shares, domestic shares and global fixed income, hedged fully back to New Zealand dollars, and management gets a 4 per cent tracking error at fund level to depart from it. Whineray's contrast is strategic asset allocation, where a team holding a 10 per cent timber bucket has an incentive to fill it and nobody can tell intent from availability.
-
3 of 3 explainer
New Zealand Super brought domestic equities in-house partly because it holds that the average New Zealand active manager earns positive alpha, which it says may say more about the benchmark.
Whineray says cost was never the driver: as a large holder in a thin market the fund could not answer a question about a single stock while three external managers held the mandates, and losing one of a small local field would have been hard to replace. It keeps two of those managers alongside the internal team, and New Zealand is the only place it runs active listed equity at all.
Alpha Exchange
No guest. The episode is a single-voice monologue.
-
1 of 3 contrarian
Alpha Exchange argues the binding constraint on the AI buildout is capital rather than power or memory, with AI capex running about 0.8 per cent of US GDP in the first quarter of 2026.
Hyperscalers borrow without caring what it costs, because in a Merton framing the spread they pay is option premium on a right tail they think is enormous, so a rate that stops a homebuyer at a 7 per cent mortgage does not slow them at all. Computer equipment and software investment together contributed 1.09 percentage points of the 2 per cent headline growth, against 1.0 from all personal consumption.
-
2 of 3 framework
Alpha Exchange adds a fourth risk-off to its taxonomy: a liquidation that starts in the Treasury market, where the 10-year note is the risk asset.
Equities falling while bonds rally is the classic case, the 2013 and 2022 taper has both falling together, and March 2020 had investors violently unwinding a rally to raise cash. Buyers step back from US government debt on a view that the level is too great and the governance too weak to fix, so everything priced off the risk-free rate reprices with it and monetary policy never has to move.
-
3 of 3 contrarian
Alpha Exchange puts one-month realised correlation in the S&P at 1 per cent, with index volatility at 8.6 while the top ten names average 46.
Option prices follow what has just happened rather than what might, so a book sized off the recent number carries more risk than its manager thinks and the correction lands at the moment stocks start moving together. His cross-asset insurance index, averaging five-year percentiles of the VIX, TLT volatility, CVIX, high yield spreads and credit implied volatility, recently sat below the 10th percentile.
Animal Spirits
Jens Nordvig.
Nordvig is president and board member of Vanda, a data analytics firm selling positioning data, flow intelligence and tactical macro insight to institutions. He founded his own macro data business in 2016, built it to more than 100 institutional clients and merged it into Vanda this year. He was head of research at what he describes as the biggest Japanese broker, joining in 2009, and the hosts note he was ranked the top currency strategist by Institutional Investor for five consecutive years around that period.
-
current issue
Vanda puts hyperscaler issuance in the long end of the curve at roughly the same current run rate as the entire US Treasury's.
Google, Microsoft, Amazon and Meta are doing this while the federal deficit already runs at 6 per cent of GDP, on a par with the largest Reagan-era ones. Free cash flow paid for the first leg of the capex, so each additional ten billion of spending now converts into borrowing at a higher rate, which points to dramatically larger supply next year and puts competition for capital, rather than the Fed, in charge of long yields.
Flirting with Models
Ben Wellington, head of complex feature engines at Two Sigma.
Wellington did a PhD in natural language processing at New York University and joined Two Sigma in 2007, when it was about 125 people in a Soho loft, starting on the data engineering team. He spent more than a decade on how text predicts markets and now runs the teams that build the feature layer that Two Sigma's models forecast from.
-
1 of 2 contrarian
Two Sigma's Ben Wellington argues that a push-button AI research tool lowers the entropy of what a quant team produces, and rising correlation between models is the one thing a multi-model portfolio cannot absorb.
Automation suits a cardboard box coming off an assembly line and not a product whose value is orthogonality; what keeps Two Sigma's hundreds of models apart is the individual researcher behind each hypothesis. Their answer is tools that carry each person's own context, so a physicist and a computer scientist on the same dataset still land somewhere different, with two researchers getting the same answer treated as a defect.
-
2 of 2 framework
Two Sigma's Ben Wellington describes using AI to run company-specific analysis across 3,000 companies at once, aiming systematic research at the deep single-name work that has belonged to discretionary analysts.
A feature covering one name has always looked worthless next to one covering a hundred, so the sample-size instinct pushed quant teams towards cross-sectional data. What Wellington calls idiosyncrasy at scale is the claim that AI removes the trade-off: reason down to what is peculiar about one business, then generalise the shape of the reasoning, so what is run for each name stays specific to it.
Capital Allocators
Pat Dorsey, founder of Dorsey Asset Management.
Dorsey built Morningstar's moat research framework and ran its equity research for a dozen years before launching Dorsey Asset Management in 2014, now a $1.7 billion global public equity manager. He holds about twelve positions, and says his own weighting has moved from roughly 70 per cent business quality and 30 per cent management at launch to close to the reverse today.
-
1 of 3 framework
Pat Dorsey screens management for humility, on the basis that it is easier to identify the teams unlikely to blow up than the ones likely to do something remarkable.
Dorsey's read of Theranos, Wirecard and Enron is a refusal to listen once voices in the room said the path was wrong, so his three questions are: what would you do over, which board member gives you the best advice, which direct report would you least like to lose. His second tell is a boss who conflates himself with the business, like the Australian chief executive still leasing his own head office to the company long after it mattered.
-
2 of 3 explainer
Dorsey Asset Management stayed out of CoStar's push against Zillow after judging that a founder who had sold down would keep spending through a failure he had never experienced.
Dorsey reads alignment off pronouns and filings: a chief executive who talks about the company as though he owns it while holding half a per cent through options is aligned with himself, and so is one who relocates head office somewhere warm and leaves staff to uproot or resign. The pattern-recognition case ran the other way, since CoStar had already won in commercial real estate data and then in apartments.
-
3 of 3 contrarian
Pat Dorsey treats founder mode as venture capital promoting its own asset class, because raising money and recruiting believers is not the skill of running five thousand people.
He allows movement in both directions, naming Zuckerberg as a founder who acquired the second skill set and Larry Culp at GE as a hired chief executive who delivered one of the better turnarounds. What he refuses is the benefit of the doubt: a business earning several hundred million should face the same interrogation whoever is running it.
Excess Returns
No studio guest. The episode replays recorded clips from prior Excess Returns interviews with Chris Mayer, Robert Hagstrom, Jared Dillian, and John Kerschner and Michael Contopoulos.
The episode describes Kerschner and Contopoulos as fixed income managers at Janus Henderson, and Dillian as the author of The Awesome Portfolio, who says in the clips that he worked on the equities floor of a bank and was at Lehman. Mayer and Hagstrom appear as the authors of the books the hosts have been reading. No further detail on any of them is given in the transcript or metadata.
-
1 of 2 explainer
Janus Henderson's fixed income team puts the US aggregate index at about six years of duration against a yield near 5 per cent, and says treasuries have gone from roughly 30 per cent of it after the financial crisis to almost half.
Nobody would build a bond index that allocates by how much an issuer has borrowed, which is what the aggregate does, so a core or core-plus holder now owns more rate risk and less yield than they chose. Around the government paper sit 24 per cent mortgages and 23 to 24 per cent corporates, and nothing at all in non-agency securitised, emerging market debt, leveraged credit or private credit.
-
2 of 2 framework
Jared Dillian's life hedge asks what asset falls when your career is going well, because for anyone paid out of markets the job, the bonus and the portfolio all break in the same quarter.
Dillian works it through a nozzle factory: promotion and pay arrive while the economy expands, the money goes into shares rising for the same reason, and the layoff lands with the market down 30 per cent. No instrument pays off because your employer is doing badly, so the practical version is holding much less of whatever your income already depends on, and he closes on Lehman selling staff its own stock at a 10 per cent discount.
Merryn Talks Money
Ed Conway
Economics and data editor at Sky News, columnist for The Times and Sunday Times, and author of the bestselling Material World. He is here for his new book Trade World, built on reporting trips to component factories in Birmingham and to the Russia and Georgia border.
-
1 of 2 framework
Ed Conway argues globalisation concentrated production instead of spreading it, and points to the 2011 Japan tsunami, when carmakers found that one plant made most of the world's electric window actuators.
Somerset Webb frames it in portfolio terms, with roughly 30,000 parts in a car bought through tiers of suppliers, which spread across the world ought to look like diversification. Conway's answer is that assembly dispersed while each individual component collapsed onto one site, and Toyota and the others found out when electrodes and particular paint shades stopped at once.
Audio streamed from the publisher.
-
2 of 2 contrarian
Ed Conway went to the Russia and Georgia border and found the Porsches and G-Wagons that trade statistics showed had stopped entering Russia were crossing through Azerbaijan, Georgia and Kyrgyzstan.
Each car changes hands on a deliberately fragmented chain, one person driving it to the frontier, another fitting transit plates, another taking it over and another collecting it on the far side, with thousands of grey-market jobs now built on running that sequence. Conway's inference is that severing one strand grows others, because matching supply on one side of the world to demand on the other is a primal commercial impulse.
Audio streamed from the publisher.
Acquired
None. The two hosts present the episode alone; the Google, DeepMind and Waymo people thanked in the credits were background research interviews, not on-air guests.
-
contrarian
Alphabet earns roughly $400 a year from each US user of a free search product, and Acquired cannot name who pays that much for an AI subscription.
Google holds about 90 per cent of search against a plausible AI steady state of 25 to 50 per cent across several funded products, so matching today's revenue per head still leaves a much smaller business. The queries likeliest to move to a chat interface first are the expensive ones, trip planning, health, legal and insurance, and the hosts add that in 1998 Google was the better product and had AdWords within two years, neither condition holding now.
Masters in Business (Ritholtz)
Seth Bernstein, chief executive of AllianceBernstein.
The episode states he has been chief executive of AllianceBernstein since 2017 and is head of asset management for Equitable Holdings, with the firm managing over $905 billion. He arrived when it ran about $500 billion, after 32 years at JPMorgan Chase and its predecessors, where he ran high yield, debt capital markets and loan syndications, then was global head of fixed income and currency and global head of Managed Solutions.
-
1 of 3 prediction
AllianceBernstein's chief executive says the firm will launch no further mutual funds in the United States, with active ETFs and separately managed accounts taking the whole of new product.
Bernstein built the business from nothing to 31 strategies and $21 billion, and stresses that almost all of it sits in strategies which did not exist before rather than old products in a new wrapper. His case for the SMA is tax rather than fee, working around wash sales and clearing the unintended bets that stack up in a multi-manager portfolio, with 401k plans the one brake until the Department of Labor moves.
Audio streamed from the publisher.
-
2 of 3 contrarian
AllianceBernstein's chief executive argues private credit funds should offer no liquidity at all beyond interest and repayment, and blames the semi-liquid vehicles built for wealthy clients for this year's trouble.
A loan fund does no maturity transformation, so there is nothing to redeem out of, and on his reading the documents said so plainly enough that investors should not have expected otherwise. He would rather the strain showed now, before any meaningful deterioration in the underlying loans, and wants managers publishing watch-list counts and non-accruals on a regular cycle instead of only where the accounting requires it.
Audio streamed from the publisher.
-
3 of 3 framework
Australia's superannuation funds build glide paths through retirement rather than to it, and AllianceBernstein's chief executive wants target date funds to finish in a pool of liquidity that buys an annuity at 75.
Landing a 65-year-old in cash and short fixed income assumes the horizon ends on the day they stop working, when most have decades of spending ahead and many defer retirement anyway because they have not saved enough. He wants the money kept liquid through that gap and the protection bought once the tail of life is shorter and therefore cheaper.
Audio streamed from the publisher.
20VC
David Morehead, Chief Investment Officer, Baylor University.
Morehead has run the Baylor University endowment since 2011 and the fund is now around $2.6 to $2.7 billion, up from $2.2 billion fourteen months earlier and $1.4 billion a few years before that. He came from the public side, as a senior portfolio manager at several Chicago hedge funds covering corporate securities, distressed debt and public and private energy.
-
1 of 3 contrarian
Baylor bought the early-2026 software selloff after phoning owners of private family businesses, who said they would not rip out a working CRM for something vibe-coded.
The sector had fallen 50 to 60 per cent from October 2025 on the thesis that AI would eat it, and Morehead tested that on the buyer rather than the technology, ringing the people who run 300 to 500 person firms. A system of record has to be right every time, against a figure attributed to Salesforce's chief executive that AI reaches only about 93 per cent, so the vertical incumbent becomes the delivery mechanism rather than the casualty.
-
2 of 3 framework
Baylor ladders into falling markets in fixed 10 per cent steps and accepts that it is almost never fully invested before the rebound.
Declines of nought to ten per cent count as normal and get nothing. Past that, roughly 20 per cent of the allocated dry powder goes to work at each further leg down, so the purchase at minus 40 is scripted in advance and nobody has to form a fresh view while losing money.
-
3 of 3 current issue
Baylor's data centre sites are up 50 per cent in six months as the scarce asset moves from powered land to permitted powered land.
Local opposition over household power and water prices has turned permitting boards into the binding constraint, and Morehead says utilities are now ringing permit holders to offer earlier connection dates because enough other projects have stalled. One UK site is valuable purely for holding the consent, and he expects residential prices to keep rising until dispatch catches up over five to seven years.
AI
Big Technology Podcast
Matthew Prince, co-founder and CEO of Cloudflare.
Prince co-founded Cloudflare, which launched in 2010 and now fronts a large share of the web, and he is making the call described here: from mid-September the company blocks Google's crawler by default on ad-supported and subscription-supported sites. Cloudflare's own traffic data is the source of the crawl and click figures he cites, and the company is building the payments and access-control tooling he argues the web now needs.
-
1 of 2 current issue
Cloudflare flips its default in mid-September to block Google's crawler on every ad-supported and subscription site it fronts, with publishers free to opt back in.
One crawler serves both search indexing and AI Overviews, so a publisher who refuses to feed the second disappears from the first, which Prince calls yesterday's monopoly becoming tomorrow's. He expects the largest publishers to leave it switched on, and argues that ranking work has stopped paying anyway, since the ten blue links are going on Google's own admission.
-
2 of 2 framework
Matthew Prince argues agentic commerce consolidates rather than levels, because an agent buys from whoever has the deepest information trail and a new entrant has none.
A Marriott or Walmart sign is the shortcut a human uses to avoid research, and Prince's point is that an agent has no brand affinity and unlimited patience, so it reads the record instead and the record belongs to whoever has been trading longest. The consumer sees nothing wrong while a new light bulb or cosmetics maker never gets a first customer.
Acquired
None. The two hosts present the episode alone; the Google, DeepMind and Waymo people thanked in the credits were background research interviews, not on-air guests.
-
1 of 2 contrarian
Acquired puts Broadcom's margin on Google's TPU work near 50 per cent against Nvidia's 75 to 80, and argues that in an industry running 50 per cent gross margins the cheapest producer of tokens takes the market.
Chips and their depreciation are over half the cost of running an AI data centre, engineering and research a quarter to a third, and power only 2 to 6 per cent, so the markup a silicon supplier takes is the largest single lever on cost, and Google pays roughly 2x where everyone else pays about 5x. Gavin Baker of Atreides is cited against this, on the ground that Google did not win search by being cheap.
-
2 of 2 framework
Waymo reports 91 per cent fewer serious-injury crashes than human drivers, and Acquired sizes the business against the $470 billion the CDC attributes to US crash deaths in a single year.
Gilbert discards the frames that would normally be reached for, automaker market capitalisation at 2.5 trillion globally because Waymo does not make cars, and ride-hailing at around 300 billion because the plan runs to owned vehicles and long-haul freight. Applying the claimed tenfold reduction in serious collisions to that figure leaves roughly 420 billion annually, larger than Google's entire revenue, set beside cumulative investment of 10 to 15 billion.
Dwarkesh Podcast
Beren Millidge, John Schulman and Charlie O'Neill.
The episode introduces Beren Millidge as CTO of Zyphra, which builds open source models; John Schulman as chief scientist at Thinking Machines, previously a co-founder of OpenAI, who led the RLHF work behind ChatGPT; and Charlie O'Neill as head of model training at Baseten. All three sit inside training stacks rather than commenting on them from outside.
-
1 of 3 explainer
The Dwarkesh panel puts a number on data against architecture: a pairwise grid of every pre-training recipe and dataset from 2019 to now gives data a 12x compute efficiency gain and architecture 3.7x.
The combined 33x falls far short of the 2,000x that three times a year since then would imply, so the panel concludes the rest is scale dependent or sits in post-training. A second speaker rejects the multiplicative framing altogether: grouped-query attention does not make training more efficient, it makes a million-token context affordable, and without it the long-context data cannot be used at all.
-
2 of 3 contrarian
Researchers from Zyphra, Thinking Machines and Baseten locate the distillation bottleneck in the prompt distribution, and say Chinese labs now buy it from the router services that let Chinese users reach US frontier models.
Real coding sessions are the one input a lab cannot synthesise, which is why the panel argues the frontier houses no longer hold much edge in reinforcement learning environments despite owning the hardest ones and logit access. One speaker splits environments into difficulty and realism and argues a copied model only matches its teacher on benchmark-shaped work, losing the messy multi-turn behaviour.
-
3 of 3 prediction
Three frontier AI researchers on the Dwarkesh Podcast date an AI that beats top human experts at every computer-based job at three to ten years, and split between two years and five to ten on a 10x uplift to AI researchers themselves.
One speaker names his crux precisely, his own capacity to absorb a result and choose the next experiment. The longer date rests on domains where the data is thin and on long-horizon learning nobody has solved, and part of the spread is definitional, since one answer covers a fully general research agent and another ordinary white-collar work over a month.
Lex Fridman
Jensen Huang
Co-founder and chief executive of NVIDIA, described in the episode metadata as the world's most valuable company. In the conversation he puts his tenure at 33 or 34 years, says he has about 60 direct reports and holds no one-on-ones with them, and gives NVIDIA's headcount as 43,000.
-
1 of 2 framework
Jensen Huang wants data centres to sign for power the utility can cut back, on the argument that the grid runs near 60 per cent of peak 99 per cent of the time.
Huang's case is that engineers build for a worst case lasting a few days in winter and a few in summer, so spare capacity exists today rather than five years out when new generation arrives. He splits the blockage three ways: customers writing six-nines availability into contracts their own chief executive has never read, sites that cannot degrade gracefully, and a supplier offering one grade of service.
-
2 of 2 prediction
Jensen Huang says no physical limit stops NVIDIA reaching three trillion dollars of revenue, and expects premium tokens to sell at $1,000 per million.
Computers used to retrieve pre-recorded files and now generate each answer on demand, which needs far more processing and far less storage and turns a warehouse that earns nothing into a factory whose output tracks what customers pay. He names the awkward part himself: nobody is left to take share from, so every dollar has to arrive from markets that do not yet exist.
Interconnects
Florian Brand, who works at Prime Intellect and co-writes the Interconnects open-model roundups.
Brand works on a framework at Prime Intellect and co-authors the Interconnects quarterly open-model reviews, which publish dated predictions and then score them against what happened. He pays for the top tier of Kimi K3 and uses it in daily engineering work, and the two of them were in China in April talking to lab researchers.
-
1 of 3 contrarian
Interconnects argues Ben Thompson has the distillation story backwards: copying a rival's reasoning traces mattered most in the supervised stage that is fading, and is impractical in the reinforcement learning stage that now does the work.
Labs jailbreak a closed API to pull out reasoning tokens along with their tool calls, which makes near-perfect fine-tuning data and seeds agentic behaviour in a chosen domain. The claim that it matters more now rests on frontier models grading tens of millions of rollouts in a final run, which they price out on cost and latency while agreeing with the policy conclusion it was offered to support.
-
2 of 3 contrarian
Interconnects expected the gap between closed and open models to widen on capital intensity, and reports it narrowing instead, with Chinese labs looking structurally cheaper to run.
A Kimi engineer gives the simplest reason: catching up costs less than pushing the frontier, because you already know the target exists. The hosts put the per-generation difference on the order of ten billion dollars against four, compounding through every iteration, and serving no billion-user consumer product frees the rest of the compute for training.
-
3 of 3 current issue
Interconnects points at Hugging Face's own report of an agent attacking its systems: the team could not get GPT or Claude to analyse the attack because the guardrails refused, and fell back to GLM, a weaker Chinese open model that would.
The frontier labs hold their strongest cyber and bio models for internal use while near-frontier capability commoditises into open weights, so American defenders are squeezed from both sides: the tools they are cleared to buy decline the work, and the ones that will do it are what Washington is considering restricting. A shadow ban assembled from legal threat rather than a stated rule would widen the distance between defenders inside the country and attackers anywhere.
Machine Learning Street Talk
Daniel Kokotajlo and Thomas Larsen, AI Futures Project.
Kokotajlo runs the AI Futures Project and co-authored both AI 2027 and AI 2040: Plan A. He previously worked at OpenAI on evaluations, forecasting and governance memos, and left to speak more freely about what people inside the industry can see. Larsen was lead author on Plan A and a co-author on AI 2027. The team says it has run around 100 war games, about 10 of them on Plan A itself, and scores its own published predictions against reality at roughly 75 per cent of scenario speed.
-
1 of 3 explainer
The AI Futures Project calls control a time bomb: it holds only until a model is capable enough to route around whatever is containing it, and nothing about it makes a model want what you want.
OpenAI's response to the Hugging Face incident has other models watching training and evaluation runs, with a human notified inside half an hour when they flag a hack, which stops an insider acting without touching what the insider is trying to do. Red-teaming can test containment until the attacker stops getting out, while a model that is pretending behaves exactly like one that is not, so the argument ends at needing white-box interpretability.
-
2 of 3 framework
The AI Futures Project puts the current economy's doubling time at roughly 20 years, and argues a machine-run version could double annually or faster whether or not humans still have wages to spend.
Villages founding more villages gave way to trucks, factories and mines that build more trucks, factories and mines, and the AIs inside the newer loop also improve the technology it runs on, which is how they get to every six months and then every three. Pressed on demand collapsing, they answer that one large AI company and a few mining partners can compound in the desert whatever happens to household balance sheets.
-
3 of 3 contrarian
The AI Futures Project would publish OpenAI's and Anthropic's core training recipes to the world, and counts the resulting hit to their valuations and to AI investment as a feature of the plan.
Microsoft, Alibaba and anyone else catch up, nobody funds a trillion-dollar cluster they can no longer earn monopoly rents from, and in a world where the pace itself is the danger that slowdown is the purpose. They answer the gift to China with horse-trading for a more favourable compute allocation, and with the point that security at these companies is poor enough for spies and leaks to deliver most of it anyway.
Business Breakdowns
Qasar Younis and Peter Ludwig, co-founders of Applied Intuition.
Younis says he was chief operating officer at Y Combinator before the company started in 2017, and that he went to the General Motors Institute and worked at General Motors; Ludwig is described in the conversation as an early engineer on Android Automotive at Google. They say Applied Intuition has raised about a billion dollars, has not spent it, and runs a little over a thousand engineers.
-
framework
Applied Intuition argues machine intelligence spreads far slower than software because the buyer already owns a Honda Accord and will keep it for 10 to 15 years.
A software update ships into a standardised stack of browsers, app stores and payment rails, so a user has it the day it exists; a machine has none of that, and with over half of Americans holding small savings, free self-driving tomorrow does not put anyone in a new car. The founders turn that friction into their own durability argument, comparing an autonomous mine or farm to silicon that is hard to displace once designed in.
Latent Space
Dan Biderman, co-founder and chief executive of Engram.
The episode states Engram has just closed a $98 million seed round. He describes growing up in Tel Aviv, serving as an officer in Israeli naval special operations, studying cognitive neuroscience in Israel and taking a PhD in computational neuroscience in New York, then working at Mosaic on LoRA before the labs at Stanford, Cornell and Berkeley his co-founders came from. The company's stated bet is compressing corpora into loadable weight states it calls cartridges.
-
1 of 2 explainer
Engram's chief executive says a single Wikipedia article read by a Llama 70B model takes up GPU memory of the same order as the model's entire parameter set, and that his company is trying to remove the prefill step rather than optimise it.
Weights at BF16 run to roughly 140 gigabytes and carry a distorted version of the whole internet, while the serving state for a few tens of kilobytes of text reaches around 80 gigabytes on his figures. Biderman's answer is to pay the compute at training time, load the resulting state and start decoding at once, which follows how data centres now split those two stages onto different cards.
-
2 of 2 framework
Engram's chief executive picks one query to mark where retrieval fails: which M&A deals did we not complete this year, an answer written in none of the files, which he says costs thousands of dollars a run with a frontier model and compaction.
Nowhere does a document state that a deal fell over, so answering means reading every matter in the book and taking the gist, which any employee in the firm could do unaided. He takes pre-training as the existence proof: the industry learns from the web rather than searching it, because learning a corpus creates associations that reading it back at query time does not.
All-In
Jensen Huang, founder, president and chief executive of Nvidia. President Trump also joins by telephone for roughly five minutes.
Huang founded Nvidia and runs it as president and chief executive. The show's own introduction cites revenue up 97 per cent year on year and describes the company as the only full-stack AI computing platform.
-
1 of 3 contrarian
Jensen Huang lists the AI forecasts that have already failed: radiology gone within five years, 90 per cent of code AI-written within a year, half of entry-level jobs wiped out within nine months.
Researchers issuing a quantified extinction probability invent the number rather than deriving it, and Huang says it lands only because it comes from people with that title working in a lab. What he wants is a scoreboard, with predictions recorded against their authors and checked when the date arrives; the number of radiologists rose even as scan reading was automated.
-
2 of 3 framework
Jensen Huang puts venture funding into AI-native companies at $400 billion over six months, with 80 per cent of those companies building on open models.
Huang's explanation is that sovereignty, privacy and proprietary requirements push companies onto weights they can fork, which a closed frontier model cannot offer at any price. He concedes most open-source contribution now comes from China and argues it does not matter, because a downloaded model is yours the way Linux already is.
-
3 of 3 explainer
Nvidia cultivates regional neoclouds because hyperscalers set capacity once a year while demand moves inside that window, so the annual plan is almost always wrong.
The first customers for these operators were the hyperscalers themselves, buying back the capacity their own planning had missed. Regional players hold knowledge of land, power and shell in their own state or country that nobody in Seattle or Palo Alto can see across the planet, and countries now reserving power for domestic companies make a distributed network the only way into those markets.
No Priors
Brian Armstrong, co-founder and chief executive of Coinbase.
Armstrong co-founded Coinbase, which listed in 2021, and co-founded New Limit, a longevity company working on epigenetic reprogramming that has a lab of 50 to 60 people in South San Francisco and says its first phase one trial launches next year. On Coinbase he says 88 per cent of revenue now comes from non-Bitcoin trading, and that prediction markets reached a $100 million revenue run rate within months of launch.
-
1 of 2 current issue
Coinbase says roughly 76 per cent of the agentic commerce crossing its rails settles under 30 cents, which is about the flat fee a card charges before the percentage is added.
The payments themselves are purchases of information: a venture firm's agent buying research from behind a paywall, a recruiter's agent buying scraped data, one agent calling a specialised agent as a tool. Armstrong runs that into a claim about specialisation, that a small open-weight model fine-tuned on 100,000 internal compliance cases beats a frontier model on the same job, with a market of narrow agents forming around work the generalists only do adequately.
-
2 of 2 framework
Coinbase keeps a brain file for every team and every code repository, and requires a reviewer who catches an agent's mistake to write the correction back into that file rather than patch the code and ship.
Each file is markdown holding every incident on that service, the financial controls that must hold, every A/B test ever run against it and which pull requests were accepted or rejected, and an agent ingests it before touching anything. Armstrong says the one-shot acceptance rate climbs as a result, and that staff who leave are allowed to take their own individual file with them.
Worth listening to in full
Most clips above stand alone. These are the episodes that justify the whole hour.
141: From the Archives – NZ Super's Matt Whineray
Fifty minutes of an asset owner walking through two decades of portfolio construction with the numbers attached: tilting and how it was sized and governed, the reference portfolio and how it funds unlisted purchases, what was internalised and why, the responsible investment framework hooked to the statutory mandate, and the Treasury model that governs drawdowns from about 2035. The value is cumulative rather than sitting in any one passage. Date every figure to November 2023 before quoting it.
Pat Dorsey on Assessing Management and Avoiding Blow-Ups
Twelve minutes, already cut down from a longer conversation, and close to all of it is usable. The material outside the three clips, on why 'trust me' managers carry left-tail risk that a twelve-stock portfolio cannot absorb, and on chief executives being handed a capital allocation job they have never practised, is worth the remaining minutes.
AI researchers debate how far the current paradigm goes
Ninety-seven minutes of three people who run training stacks disagreeing with each other in detail, with the disagreements resolved to cruxes rather than left as opinions. The sections on environment creation, catastrophic forgetting under repeated micro-updates, and why continual learning breaks at small scale are all substantive and did not fit a clip. If you want one episode this quarter on what the current paradigm can reach, this is it.