Jack Altman

1. Key Themes

The Seed Round Has Bifurcated: Mega-Rounds Above, Slow Compounders Below

The first round has changed shape: some founders skip the traditional seed entirely and raise $50M+ because they can. Others still build on a few million dollars, but that is a different and shrinking part of the market. Jack Altman described the dynamic: "there is a set of founders that are very like in the Silicon Valley network that don't necessarily need to raise... a $50 million seed, but they can. And so they skip the first round... And so in many lanes, I think that round has kind of evaporated." 00:33:46 Rory O'Driscoll framed it as "almost two different worlds": $400-500M to ship a Neolab product versus "we took $10 million, but we shipped the product for $3 million." 00:31:28 Harry Stebbings said his team "can't find anything under 100 million" in round size. 00:31:02

AI Is the Highest-Variance Period in Venture, and Sizing Is the Whole Game

Both prices and outcomes are extreme at once. Jack Altman: "we're definitely not investing at the right speed. We are either investing way too fast or way too slow. But when both sides of the equation are this out of whack, the odds of having it right are zero." 00:00:17 Rory relayed Tyler Cowen's line at his annual meeting: "returns are going to be highly skewed. Variance is going to go up with AI and many of you will fail." 00:00:17 On sizing, Jack's logic is that early on "you just need enough shots on goal," constrained from above by needing "to matter in the context of the round." 00:19:18 Rory pointed to Kelly betting: "edge over odds. And then you say, do I have the stones to do full Kelly or do I back off?" 00:19:30

Consumer Agents Are a New Paradigm, and the Real Test Is Daily Use

Jack Altman cast Instinct as comparable to the coding story: "it kind of reminds me of what happened with coding and with Cursor and Cognition in the face of the labs... you've got these consumer agents... that can interact with the entire third party Internet." 00:10:27 Jason Lemkin's test for winners is hours of daily use: "the ones that win... are the ones we use all effing day long... I don't know if we have an eight hours a day app yet here." 00:12:03 Jack's counter was that early usage is always thin: "if you did the reference calls on their customers when they were at a million of ARR, the customers were like the thing barely works... And now you talk to them and they're like, I run my whole life out of it." 00:13:05 Harry noted the cannibalization risk to chat: "I used to be a massive ChatGPT user... And now I just live my life through Instinct." 00:13:46

Strategic Acquirers Are the New Exit Path for Neolabs

Jack Altman: "There are like 10 companies that can do $10 billion acquisitions and want to. And that's just so different. And it's much easier than going public." 00:00:27 Rory expects more: "all the big foundation model companies are probably in the market to acquire some kind of robotic foundation model story... there will be a bunch of these big-ass acquisitions over the next 6, 12 months if the market continues to hold." 00:27:57 Harry pushed back with the denominator problem, citing a report on "102 Neolabs, $70 billion plus raised." 00:28:32 Rory's filter is credentialism: "for something like... building a world model, you need to have proven technical success." 00:30:19

Inference Is the Index Bet on Everything Outside the Labs

Jack Altman: "the right answer with investing was just put it all in the labs... It has now turned out in the last 18 months... the correct answer was just keep buying inference. You have Modal, Baseten, Fireworks, FAL together. It's just all worked." 00:46:36 He cited two drivers: cost unsustainability, and "intelligent saturation," where "once you have filed [your tax return] correctly, throwing more intelligence at that problem doesn't do you any good." 00:47:33 Jason Lemkin's bull case for Jev rests on market size: "Spend right now, today, is roughly $100 billion... 20% of that is relevant to Jev. So that's $20 billion of accessible revenue. Say they compress it five to one. That gets you $4 billion." 00:45:13

Open Weights May Have Peaked, and Compute Share Is the Ultimate Variable

Jason argued: "I think we've reached peak open weights" because the labs can price cheaply on demand and because of enterprise trust: "No one wants to run on open-source models there. At least Chinese, China-based. Nobody." 00:00:40 Jack Altman added that enterprises will post-train their own models on open weights, and that the market comes down to compute: "it's kind of all just going to come down to like who's got the compute... I think it's, you know, on the order of like a gigawatt or something like that [across inference clouds]. And I think OpenAI [and] Anthropic, you know, high single digits each." 00:00:40 Rory's heuristic: "Compute share probably proxies to your token share, probably proxies with a little adjustment to your revenue share." 00:53:46

Tokens-Per-Task Is Unpredictable, So Watch Buyer Budgets

Rory tried to model AI economics and found it unforecastable: "you're multiplying three numbers, each of which has an error bar that's pretty damn large... The only thing you can observe that's actionable is the buying decisions of people who are allocating their money." 00:54:49 Jason's own Claude Sonnet 5.5 test showed the problem: "Input tokens, 42% higher than before. Output tokens, 44% higher... Cost went down about 10%, not 30%." 00:56:06

Talent Is Being Repriced by the AI Gravity Well

The MongoDB CEO leaving for Meta shows the pull. Jason Lemkin's guess at the package: "he was offered a $500 million package by Zuck to run enterprise." 00:40:28 Jack Altman added that it's not only money: "it's also the attention... it's the thing in all of the headlines." He sees a societal upside: "talent feels very unstuck right now." 00:42:21 Rory's summary: "The market is sending a signal that the only place to be is in these extraordinarily hot AI companies. And it's sending that signal via price and people are responding to price." 00:41:36

The IPO Window Is Open, But the Anthropic S-1 Is Politically Combustible

Rory said the leaked S-1 contained little new but flagged the real catalyst: "The Q3 revenue number is 90% of the data required to make a decision on pricing on Anthropic." 00:06:02 The risk is political, not financial: "I do think there is a non-trivial political backlash brewing... this is the first S1 I've ever seen that poses existential risk and where the product lies to me. Maybe I should, as the attorney general for the great state of fill in the blank, sue these guys." 00:08:50 Jason Lemkin worried about a post-IPO drift from negative retail sentiment. 00:08:04

Venture Fund Structure Is Fragmenting: Go Bigger or Go Proprietary

Bessemer's $5.75B and NFX's decision to invest only GP capital represent opposite responses. Jack Altman on NFX: "When you're not managing external capital, you get to do things without any explanation to anybody." 01:09:14 He also noted the firm-building tradeoff: Peter Thiel "put a huge amount of his money into each founder's fund. But not all," because "if you want to build a firm, you need outside capital." 01:10:47 Jason Lemkin on why he wouldn't go proprietary: "I got to be able to write a $5 million check or I don't want to do this game." 01:13:26

2. Contrarian Perspectives

The Anthropic Leak Told You Nothing, and the Scary Numbers Are Mostly Non-Cash

While the headlines screamed about an $8B operating loss, Rory dismissed the whole leak: "Not a single piece of information in that leak. Not a single piece of useful... The whole, oh my God, they're losing 40 billion. When 34 billion of [it is] non-accounting loss." 00:05:08 The one real data point: "two customers did 25% of the revenue, which means someone spent half a billion dollars on Anthropic last year." 00:05:33 The implication is that investors reading backward-looking financials are chasing noise; Q3 is the signal.

A $10B Pre-Revenue Round Can Be an "Early Stage" Investment

Jack Altman said Benchmark treats Instinct as early stage despite the price: "in our minds, it was actually kind of an early stage investment. And I know that sounds psychotic." 00:16:13 Rory's reframe: "You thought you were raising a growth fund. In fact, you're just expanding the early stage fund because it just takes bigger checks now." 00:16:26 The thesis is that stage should be defined by uncertainty, not dollars.

The Quiet Compounder Is Uninvestable Because It Isn't Stable

Jason Lemkin, who has pitched quiet compounding "since 2012," now rejects it: "they're just not stable. You have to build so much more software so much more quickly... I have another company at 100 million. They put up a competitor slide... I thought we had one competitor. Turns out we have nine." 00:37:40 Rory stated the logic: "the quid pro quo should be low risk. And if... you get the same level of risk, that by definition is a suboptimal game." 00:38:30 Harry added the opportunity-cost argument: "We're in an opportunity cost game where people can put money into your Sarah grows and your benchmarks of the world." 00:36:56

Open-Weight Model Share Has Peaked

Against the prevailing market-share charts, Jason Lemkin said: "I think we've reached peak open weights... I think these market share charts that have seemed crazy this year... there's a lot of structural benefits that OpenAI and Anthropic have here." 00:49:07 Jack Altman partly disagreed, noting that even if share falls 50% while consumption rises 10x, inference remains a good investment. 00:49:07

Public Markets May Be More Patient Than Private Ones on AI Labs

Counter to the assumption that private investors are the patient capital, Jack Altman argued: "I actually think the public market investors will be a little bit more long term oriented than the private investors... each month, it's like the narrative is just all over the place." 00:06:24 Rory added that the stock buyers are "a small number of highly compensated managers" and not the 70% of Americans who dislike data centers. 00:08:25

3. Companies Identified

Instinct

A consumer agent company founded by 23-year-old Noah Shin, raising a $1B Series C at a $10B valuation, led by Benchmark's Peter and Ev. Mentioned as the new agent paradigm competing with Meta's Muse and the labs. Jack Altman: "this is such an important new paradigm where you've got... these consumer agents that are... the aggregator of aggregators." 00:10:56 Harry Stebbings' revealed preference: "now I just live my life through Instinct." 00:13:46 Rory's skeptical flag: "if you start thinking... of it as travel monetization, you could get pretty angsty about market size here." 00:17:23

Meta (Muse)

Meta's consumer agent product, praised for reinvigorating the organization and competing directly with Instinct. Jack Altman: "They've done a phenomenal job and it's a great product. They've sort of reinvigorated the organization." 00:10:27 Meta also hired MongoDB's CEO to run enterprise. 00:39:17

Anthropic

Frontier lab whose draft S-1 leaked: $4.6B 2025 revenue, $8B operating loss, $518B compute commitments. Rory highlighted customer concentration: "two customers did 25% of the revenue, which means someone spent half a billion dollars on Anthropic last year, which is pretty impressive." 00:05:33 Jason Lemkin noted Anthropic's gross margin claim with caveat: "If you exclude 7,000 things that has 80% gross margins." 00:49:14 Also notable: founders locking up 50.1% voting control. 01:13:53

OpenAI

Frontier lab said to have hit $70B in enterprise, and which paused then reopened a $200 plan at reduced value. Jason Lemkin: "OpenAI just hit $70 billion, they said, in enterprise. Like, it's a lot of traction there." 00:51:27 Rory noted that OpenAI is "really exploding in growth." 00:06:02 OpenAI's dev day was happening the day of recording. 00:15:42

World Labs

Fei-Fei Li's 3D world models company, acquired by AMD for $8.2B in stock two and a half years after founding. The first big Neolab exit. Rory: "it would be pretty criminal if we had the biggest boom in AI history and Fei-Fei Li didn't get a big ass check." 00:26:22 Jason Lemkin on the acquirer logic: "for 8% of that market cap... getting a world-class team to make sure that 279% run continues, like it's cheap if it's the right team." 00:26:44 Harry noted the first round "was like 65" million. 00:31:02

AMD

Chipmaker, up 279% this year to roughly $1 trillion market cap, buying World Labs for stock. Jason Lemkin: "AMD was like one of your grandpa's investments... It's up 279% this year to $1 trillion." 00:26:44 Rory: "AMD just wants to be relevant to NVIDIA." 00:27:57

NVIDIA

Referenced as the benchmark AMD is chasing, and its model team as a counterpart. Jason Lemkin: "I don't even know whether they're going to continue the 3D world models or they're just going to be their counterpart to NVIDIA's model team." 00:26:44

Andreessen Horowitz (Martin Casado)

Named as having a hot streak across OpenRouter, World Labs, and Cursor. Harry: "Mr. Martin Casado is making us look bad because the man has had OpenRouter. Obviously, last night had Feifei with World Labs and then he had Cursor." 00:24:46 Jack Altman: "extremely impressive run by Martine." 00:29:20

Jev

A fast, low-cost inference-routing company reportedly raising at a $10B valuation after a seed at about $200M. Jason Lemkin's pitch: "Jev's already 17% of the traffic on OpenRouter. It's 20% of the traffic through Vercel's router... Jev is the 70th of the price and 100 times faster." 00:43:50 Rory's analysis: "you can credibly get to a billion dollar revenue line relatively quickly here by literally taking money that's already being spent and saving 80 cents on the dollar." 00:45:43

Modal

Inference infrastructure company tripling to $15B in a new round. Jack Altman grouped it among inference winners: "You have Modal, Baseten, Fireworks, FAL together. It's just all worked." 00:46:36

Baseten

Inference company reportedly in talks at $26B. Cited by Jack Altman as part of the inference index bet. 00:47:05

Fireworks

Inference company grouped by Jack Altman with the inference winners. 00:47:05

FAL

Inference platform grouped by Jack Altman with the inference winners. 00:47:05

OpenRouter

Model routing marketplace where Jev has 17% of traffic. Referenced both for its traffic data and as a Martin Casado investment. 00:43:50

Vercel

Mentioned for its router, through which Jev reportedly has 20% of traffic. 00:43:50

Cursor

AI coding company cited as the precedent for independents surviving against labs. Jack Altman: "when Cursor came out, they have like a long window before the labs caught up. Now... you look at the Instinct to use window. It's short. These cycles are only getting shorter." 00:15:18

Cognition

Coding agent company cited alongside Cursor as an independent that survived lab competition. 00:10:56

Harvey

Legal AI company cited as an example of usage deepening over time. Jack Altman: "Logora and Harvey... now you talk to them and they're like, I run my whole life out of it. And these businesses are at hundreds of millions of ARR growing really fast." 00:13:05 Also named as a Vanta customer. 00:02:24

Legora

Legal AI company ("Logora" in transcript), cited for rapid growth and daily usage. Jason Lemkin: "folks are running Harvey and Logora like Harry's partner... She gets home. She's on the sofa running it all effing day long." 00:12:03 Harry noted the stories "still play out in a short number of years." 00:35:36

Higgsfield

AI video company Rory cited as an example of capital-efficient shipping: "you guys are in Higgsfield. You run into a bunch of these guys who are like, well, we took $10 million, but we shipped the product for $3 million. And then the customers took off." 00:31:02

Owner

A software company on whose board Jason Lemkin and Jack Altman serve, "a rocket ship north of 100 million revenue." Used as an example of how much software must be built to stay ahead. 00:38:09

Oura

Consumer health wearable that pulled its planned $16B IPO days before pricing, with Morgan Stanley, Goldman Sachs and JP Morgan as underwriters. Rory (an investor) was surprised: "they had Morgan Stanley, Goldman Sachs, and JP Morgan... It's a profit. It's a big company. Consumers know it." 00:57:27 Forerunner had planned to sell all of its position. 00:58:25 Employees had received "534 million in tender offer just a couple months ago." 01:00:49

Forerunner

Venture firm and large Oura investor that planned to sell its whole position in the IPO. Rory: "Forerunner, whom I think are super smart, said they're going to sell all their position, which I've been doing a long time. I'd never seen someone in an IPO being able to sell all their position." 00:57:55

Monzo

British digital bank, reportedly being acquired by Nubank at an $8-12B range. Rory attributed the sale logic to governance turmoil: "all the VCs were like, you did what? We weren't backing the chairman. We were backing the CEO and you changed him." 01:06:54 Harry noted it is "a phenomenal asset in the UK" but too small for US or European public markets. 01:03:32

Nubank

Brazilian digital bank buying Monzo, with stock down on the news (market cap $47B, down 12.6% on the week). Jason Lemkin's read: "they're buying time... They're saving themselves time." 01:05:44 Harry questioned the strategic fit with David Vélez's focus on the US. 01:03:24

Revolut

European neobank cited as a competitor and as proof that crappy incumbent markets reward challengers. Rory: "Revolut, which is sticking it to all the crappy old school banks in Europe, is making out like a bandit." 01:04:23

Chime

US neobank, trading at "five, six billion dollars," used as evidence that efficient US banking leaves less profit pool for challengers. Rory: "Chime, which I admire as a company, trades well, but not amazingly." 01:03:53

Bessemer Venture Partners

Raised $5.75B including a $1.75B seed fund. Jason Lemkin tied it to Anthropic: "they weren't in Anthropic [early]... but that was a great one, right? The growth team accelerated... Samir and the whole team, it's been a win." 01:08:45 Jack Altman: "if you want to play the game, why not just go bigger and play the game?" 01:09:14

Benchmark

Led the Instinct round after a historic "no growth" ethos, signaling a shift. Rory: "I've known Benchmark since 95... when you guys said after 30 years, you're going to embrace growth, I thought you tiptoe in... It's 10 billion pre pre revenue, big ass check." 00:15:42 Harry referenced the fund's results: "your Sarah grows and your benchmarks of the world." 00:36:56

NFX

Venture firm moving to invest only GP capital. Cited by Jack Altman as a structural alternative: "When you're not managing external capital, you get to do things without any explanation to anybody." 01:09:14

Homebrew

Seed firm that pioneered investing GP capital with outside LPs for diverse check sizes. Jason Lemkin: "the Homebrew guys... they were phenomenally successful... In fact, the Homebrew guys recommended I do it when I started." 01:13:26

Founders Fund

Referenced via Peter Thiel putting "30% of the fund" of his own money in. 01:10:44

MongoDB

Public database company (~$20B market cap) whose new CEO left for Meta, causing a 20% one-day drop before the prior CEO Dev returned. Rory: "the CEO... got an offer from Muse, hit the bid, moved over to Facebook slash Meta to run their enterprise division. And Mongo stopped, dropped 20% in one day." 00:39:17

OpenClaw

Open-source agent project credited with inspiring Instinct and Muse. Jason Lemkin: "Generation one was OpenClaw. No one knew how to run it. Other than destroying, you know, all your security and credit card. This is generation two." 00:12:31 Harry argued OpenAI should have shipped it first: "that was most strategic thing with the person who did OpenClaw." 00:15:13

Vanta

Sponsor: agentic trust and compliance platform trusted by 16,000+ companies. 00:01:55

Deel

Sponsor: global IT and HR platform across 130 countries. 00:02:54

Framer

Sponsor: AI website builder with agent and human workflows. 00:03:24

Oracle

Mentioned as a source of credit risk in the AI trade. Jason Lemkin: "the massive spend, the CDSs at Oracle, some defaults on data centers." 00:08:04

Fidelity

Referenced as a type of institutional buyer who will own AI stocks regardless of public opinion. Rory: "they're not running Fidelity Growth." 00:08:25

OpenTable

Reservation company Rory cited to size Instinct's restaurant booking opportunity. Rory: "you guys were smart enough to do OpenTable 26 years ago now... it's a billion-dollar business today." 00:17:23

Lattice

HR software company Jack Altman founded. 00:34:42

Nscale

Compute-dependent infrastructure company Rory flagged as an upcoming IPO to watch. 01:01:18

4. People Identified

Noah Shin

23-year-old founder of Instinct, described by Harry Stebbings as a "generational talent" who "raises a billion at 10 billion, takes on Muse and Alex Wang and Zuck." 00:09:59

Fei-Fei Li

Founder of World Labs and creator of ImageNet. Rory: "Cranked through ImageNet. Kicked off the whole damn thing because the entire AI resurgence was really kicked off when that ImageNet project that she built... didn't monetize that, worked at Stanford, worked at Google. And then midlife, quit, did this two and a half years ago and nailed it." 00:25:28

Martin Casado

Andreessen Horowitz partner on a streak with OpenRouter, World Labs, and Cursor. Harry: "Mr. Martin Casado is making us look bad." 00:24:46

Ev Williams (Ev, Benchmark)

Benchmark partner who led Instinct with Peter; Harry noted his humility for retweeting a take that he may have been wrong that a16z wouldn't earn a 5X in this vintage. 00:25:12

Peter (Benchmark)

Benchmark partner who co-led the Instinct investment. 00:10:27

Mark Zuckerberg

Meta CEO, described as willing to write enormous checks to recruit talent. Harry: "Mr. Mark Zuckerberg decides to whack out the checkbook to hire MongoDB's CEO." 00:01:26

Alex Wang

Cited as leading Meta's Muse efforts, in competition with Instinct. 00:10:27

Tyler Cowen

Economist who spoke at Rory's firm's annual meeting. Rory: "He just quietly said... returns are going to be highly skewed. Variance is going to go up with AI and many of you will fail." 00:23:32

Pete Steinberger (Open Claw founder)

Founder of OpenClaw ("Pete Stipeet" in transcript), subsequently absorbed into OpenAI. Harry: "Isn't this what OpenClaw meant to be? And the founder." 00:14:36

Eric Vishria

Benchmark partner who coined a line Jack Altman paraphrased: "it's all going to work... It doesn't mean that every company is going to work. Not every sector is going to work, but in general, a crazy number of things are working." 00:47:05

Ben Thompson

Stratechery writer credited with calling consumer agents "the aggregator of aggregators." 00:10:56

Dev (MongoDB)

Former long-time MongoDB CEO who stepped back in. Rory: "Dev stepped back in, who'd been not the founder, but the CEO for a long last time and is now back running it." 00:39:17

Tom Hale

Oura CEO, who pulled the IPO. Jack discussed whether the decision was driven by investor pricing: "You really think Tom Hale decided not to do the IPO because Forerunner couldn't get the price they wanted? I don't buy it." 00:59:44

David Vélez

Nubank founder and CEO, called "brilliant" by Harry but questioned for buying Monzo while focused on winning the US. 01:03:24

TS (Monzo CEO)

Monzo CEO under whom the company turned around before he was replaced. Harry: "the company turned around when TS came in. And then when TS got taken out... why the fuck did you take the guy who was competent out of this business?" 01:07:12

Tom (Monzo founder)

Monzo founder who retired. Rory: "Tom had done the founder thing. The company was doing 100 million, losing 100 million. They hired this guy. He kills it." 01:07:26

Peter Thiel

Referenced for putting about 30% of the Founders Fund in his own money while still taking outside capital. Jason Lemkin: "what he would say is if you want to build a firm, you need outside capital." 01:10:47

Dan Primack

Axios journalist. Rory: "Dan Primack made that point... We are 1.4% off the S&P all-time high. The Shiller PE is at an all-time high." 01:01:18

Benchmark partner and Lattice founder; the guest, offering the Benchmark view on Instinct, inference, and fund structure. 00:10:27

Jason Lemkin

SaaStr founder, active angel and fund investor; contributed the Jev pitch and open-weights thesis. 00:43:50

Rory O'Driscoll

Scale Venture Partners investor, Oura shareholder; contributed analysis on sizing, S-1 data, and governance. 00:05:08

Harry Stebbings

20VC host and investor, who bets on Instinct and cites a partner who runs 102-Neolab research. 00:21:26

Paul (Harry's French partner)

Harry's "brilliant French partner" who "wrote this report on 102 Neolabs, $70 billion plus raised." 00:28:32

Samir (Bessemer)

Bessemer growth leader credited for accelerating the growth team. Jason Lemkin: "Samir and the whole team, it's been a win." 01:08:16

Nick Clegg

Former UK Deputy Prime Minister who went to work for Meta, used as precedent for a prominent figure joining Zuckerberg's orbit. 00:41:20

Ilya Sutskever

Rory noted the ImageNet 2012 winners included "Ilya and two or three other people... their model using neural nets just kicked ass." 00:25:55

Chatham (Chatham Asset Management / contact)

Contact Harry messaged about a benchmark-style fund's numbers. 00:37:26

Friedrich Hayek

Invoked by Rory: "Mr. Hayek would be happy," regarding price as a signal. 00:37:26

5. Operating Insights

Pre-Commit to Sizing Boundaries Before the Deal Gets Hot

Jack Altman's framework: "the sizing is constrained by wanting the fund to have enough chances to get something great. And then like the upper bound is, you know, these rounds are big, you need to matter in the context of the round." 00:18:50 He also admitted to expecting to "size incorrectly almost every time," which argues for rules set in advance. 00:20:02 For founders, the corollary is that high-priced investors need meaningful dollars, so expect larger and fewer checks.

Use Stock-Funded Acquisition Windows When Your Own Stock Is Hot

Jason Lemkin's read of AMD: "If AMD was up 3% this year, they ain't going to be spending $8 billion. But this is a run that you... got to maintain." 00:27:12 Operators with an elevated stock or valuation should treat it as a window to buy teams. For sellers, timing a sale to an acquirer's market-cap spike matters.

Don't Go Cheap When Talent Is the Bottleneck

Rory's comment on the talent market: "Money is a signal, price is a signal, and price is sending a signal, everybody go right here." 00:00:27 And Jack Altman: "it's a talent war in the most insane way right now." 00:00:27 If your best people can be offered 10x elsewhere, retention packages and mission must compete on both axes.

Run Competitive Slides Honestly: Competitors Multiply When Software Gets Cheap

Jason Lemkin's board anecdote: "They put up a competitor slide at the last board meeting. I never heard of eight or nine of the ones. I thought we had one competitor. Turns out we have nine." 00:37:40 The tactic: assume your visible competitor list undercounts the field by an order of magnitude, and re-run the exercise each quarter.

Eval Each Model Upgrade on Your Own Task Before Assuming Savings

Jason's measured test of Sonnet 5.5: tokens up roughly 42-44%, cost down only about 10%, quality up on a blind test. 00:56:06 The tactic is to benchmark every model release on your own workload, since per-token price cuts can be offset by token-usage inflation.

6. Overlooked Insights

The "Intelligent Saturation" Principle Quietly Reprices the Whole AI Stack

Buried inside the inference discussion, Jack Altman made a point with huge implications: "We have now crossed sort of the threshold in a lot of areas... where you get sort of like intelligent saturation, where it is now good enough to do the thing... Once you have filed it correctly, throwing more intelligence at that problem doesn't do you any good. If your job is to hit a nail in with a hammer... making it a gold encrusted hammer that costs $30,000" doesn't help. 00:48:00 If many enterprise tasks saturate, then value migrates from frontier-model pricing power toward cost, speed, and routing, which explains Jev's growth and the inference companies' rise. It also stresses the labs' unit economics, because they must keep selling frontier tokens against a growing base of "good enough" workloads.

The IPO Backlog Is Waiting on the Labs to Go First, Which Makes the Anthropic Debut a Market-Wide Event

Jason Lemkin noted: "we haven't really seen any of the like AI native companies go out yet. My guess is a lot of them would trade very well because public markets don't have enough exposure to them. But I think it's scary to be the first one. It's scary to do it before the labs have gone out." 01:01:48 Rory added that Nscale is another aggressive compute-dependent name to watch. 01:01:18 The overlooked implication: the Anthropic and OpenAI listings are the gating event for an entire cohort of AI-native IPOs in 2027, and Oura's pulled deal shows even strong consumer names can stumble. Anyone holding secondary positions in AI-native companies should track post-lab-IPO trading closely.