Product lead, seed investor at 1Flourish. Ex-Google deal maker, Princeton. Father of 4. Ascend rugged mountains, paddle seas, savor coffee.
San Francisco
Joined November 2010
- Tweets618
- Following356
- Followers329
- Likes508
Neil Ahlsten retweeted
Three @MenloVentures LPs are set to make $1B+ each from the Anthropic IPO.
@mmurph explains how they raised a $500M+ SPV at a $18B valuation without being allowed to show a single financial:
"Out of that LP meeting, a bunch of our LPs said, 'This is fabulous that you're in this company. Can we co-invest with you?'
Before you knew it, there was a couple hundred million of interest.
We'd never done an SPV before. But this was a special company, and we wanted to lead the round.
We couldn't do it out of our $750 million fund. We could do $50 or $100. But we couldn't do the rest.
So we said, 'How about we take capital from our fund and our LPs, put this whole thing together, and make a proposal to lead the round?'
As history would have it, we were definitely a little over our skis.
You get a lot of people interested. But when you shake the trees, some things fall out and some don't.
There was a scary period for about 30, 45 days where I was like, 'Oh my God. I think we've over-committed here, and we may not pull this off.'
What I didn't realize up front: Anthropic was supportive, but they said we couldn't share any financial information.
You're asking people to invest at an $18 billion valuation, and you can't show any financials. We could say what last year's revenue was. But we couldn't do any forward projections.
That's a tall ask for big state pension funds. A bunch of them said, 'Look, we just can't responsibly do that as a state pension fund.'
But I can tell you that three did. And they've all made well over a billion dollars at the last round valuation.
Fortune favors the brave."
New @ThePeelPod with Matt Murphy @mmurph at @MenloVentures
We go inside Menlo's pre-revenue, pre-launch Anthropic investment and how they emerged as a top AI firm.
Matt might be the only person who invested pre-revenue in both Google and Anthropic, and we talk through when to break the rules in venture, the three LP's that made over $1B each on the $500M Anthropic SPV Menlo raised without being able to show a single financial, Menlo's barbell investing approach, and why today isn't like 1999.
Full episode here + links below.
Timestamps:
0:00 How to read Anthropic's S-1
1:21 Do AI companies actually make money?
2:23 Google didn't want ads. Anthropic didn't plan on enterprise.
4:51 Bending the rules to back Anthropic pre-revenue, pre-launch
9:50 You can't be a leading AI firm without a top model company
12:00 Leading the Series D at $18B six months later
14:22 Why the best researchers chose Dario
15:25 Menlo's first-ever SPV, nearly the size of the fund
16:44 The LP meeting that sparked a $500M SPV
19:18 Raising at $18B without showing financials
20:04 Three LP's made $1B+ each
24:23 Why Matt joined Menlo in 2015
26:00 Rebuilding Menlo over the past decade
32:49 Venture as a team sport
35:45 How to keep a challenger mentality
38:15 Why VC firms decay
41:31 The Anthology Fund: OpenRouter, Wispr, Goodfire
42:27 Ownership vs. backing the best
46:10 Why this isn't 1999
48:06 Frontier vs. open weight models
51:29 The barbell portfolio strategy
53:54 Avoiding the messy middle between early and growth
55:07 How to spot a category leader
57:18 Fortune 5000 AI adoption: "almost none"
59:06 Menlo is in 10 of 70+ neolabs: Skild, Chai, Axiom
1:02:13 Where $1B AI startup raises actually go
1:04:18 Phase 1 vs. Phase 2 of AI
1:06:26 What the AI IPOs mean for venture
1:09:11 What Matt does differently after 25 years
1:10:52 The Alex Smith lesson
Neil Ahlsten retweeted
The most important education I ever got, I paid for myself.
Not the degrees. Not the Stanford lectureship. The books I read because I needed to solve a problem in my own life.
I used to be clinically obese. School didn't fix that. I read my way out of it — behavioral science, habit research, the actual studies on why we eat when we aren't hungry.
Later, when my own phone use was embarrassing me in front of my daughter, I didn't go back for another degree. I read. I ran the experiments on myself. That became Indistractable.
Formal education hands you a credential. Self-education hands you a life you actually chose.
The people I know who kept getting more interesting past 40 all have the same habit: they read like their lives depend on it. Because, quietly, they do.
Pick the problem that's bothering you most right now. Find the three best books on it. Read them this quarter. That's the whole program.
Neil Ahlsten retweeted
I'm so excited to partner with @kaggle and @GoogleDeepMind on the first human behavior benchmark :)
Can frontier models actually predict how real people behave? 🕺
Today, in collaboration with @RunRehearsals, we are launching Human Behavior Bench, the first public benchmark that tests whether AI can understand human preference and everyday decision-making.
AI models have gotten increasingly good at coding, but knowing for whom to build is the next frontier of challenges for AI models to capture. Human Behavior Bench evaluates models on predicting real human behavior across health, entertainment, recreation, values, and consumption.
To build this baseline, we gathered ground truth data directly from real people and evaluated models across a wide range of everyday scenarios:
• Ground truth baseline: 1,244 US adults surveyed (June-Sept 2026)
• Some cohorts are traditional demographic indicators such as age, income, and location
• Other filters are behavioral e.g. GLP-1 patients and recent car buyers
• 750 cases total (149 public / 601 private) across finance, health, recreation, values, and consumption
• Scored using 1-TVD (total variation distance)
Base models still have a significant gap to close to accurately predict human behavior by cohort. LLMs’ predictions tend to converge towards the mean, while the true distribution of human preferences vary greatly.
Against a human noise ceiling of ~93, the current leaderboard scores are: Rehearsals’ Engram 1.0 (AI Twins) tops the leaderboard at 94.1%, followed by Gemini 4 Argon at 80.6% Claude Opus 5 at 80.2%, GPT-5.6 Sol at 79.5%.
ALT A leaderboard graphic from Kaggle titled "Rehearsals: Human Behavior Bench". The ranking chart lists nine AI models and their corresponding scores, measured on a scale from 70% to 85%. The models are ranked as follows: 1. Gemini 4 Argon with a score of 80.6% 2. Claude Opus 5 with a score of 80.2% 3. GPT-5.6 Sol with a score of 79.5% 4. Claude Opus 5.5 with a score of 79.3% 5. Gemini 3.8 Flash with a score of 79.2% 6. Gemini 3 Flash Preview with a score of 78.5% 7. GPT-5.6 Terra with a score of 77.8% 8. Gemini 3.5 Flash-Lite with a score of 75.9% 9. Grok 4.7 with a score of 75.2% The background features a subtle gradient with warm yellow and beige tones, and the source is cited at the bottom as the Human Behavior Bench Leaderboard on Kaggle.
Neil Ahlsten retweeted
Today, we're launching Foundry Management, formed by @a16z and Atlas Holdings to help founders build AI companies inside the industrial world.
The hardest constraint in physical-world industries is often not the technology, but getting close enough to mission-critical work to understand what should be built. Foundry gives you a place to begin.
We recruit a small, select group of some of the top technologists and give them premiere access to the Atlas portfolio, representing $26B in annual revenue across metals, paper, packaging, building products, automotive supply chains, energy, and logistics.
You'll work alongside the operators and management at these businesses to build products and technologies that unleash AI in these industrial contexts. When it demonstrates the potential for an independent business, a16z and Atlas will support its creation, with you leading it as founder and CEO.
There’s a great deal of the future left to build inside the industries already around us.
Come build it with us: [email protected]
Neil Ahlsten retweeted
Anduril delivered the first CCA to the @usairforce last month. Now, USAF operators are flying Fury on their own.
Welcome Fury and Vengeance to the experimental operational unit! Our Airmen are excited to see what CCA can do.
af.mil/News/Article-Display/…
Neil Ahlsten retweeted
Stanford + MIT paper on Model Harnesses shows that AI performance depends not just on the model itself, but on the surrounding system code — the “harness”.
This is what decides what to store, retrieve, show to the model, and how the workflow runs. With the same underlying LLM, changing the harness can create up to a 6× performance gap on the same benchmark.
They conclude the harness around a model matters as much as the model itself.
The paper introduces Meta-Harness, an outer-loop system that automatically improves harness code. Instead of giving the optimizing agent only a score or a short summary of past attempts, it gives the agent rich access to prior code, logs, and execution traces through a filesystem-like setup.
The idea is that better diagnostic visibility lets the system improve the harness more intelligently.
What it found is pretty significant.
- On online text classification, a 7.7-point improvement over a strong SOTA context management, while using 4X fewer context tokens.
- On retrieval-augmented math reasoning, an average gain of 4.7 points across five held-out models on 200 IMO-level problems. On agentic coding, the discovered harnesses beat strong hand-engineered baselines on TerminalBench-2.
The paper shifts attention from “which model is best?” to “how is the whole AI system designed?”
For real deployments, harness design affects reliability, tool usage, context management, and failure recovery.
----
Paper – arxiv. org/abs/2603.28052
Paper Title: "Meta-Harness: End-to-End Optimization of Model Harnesses"
Neil Ahlsten retweeted
After such a consequential week for AI, with the leading labs demonstrating real progress, today's news feels especially timely: @SVAngel is launching Project Blueprint, a comprehensive effort to forge consensus among industry leaders and lawmakers around durable solutions to the hardest AI policy challenges so that all Americans can benefit from AI's opportunities. And we've brought on @JayCarney to run it. svangel.com/blueprint
We are encouraged by the reaction of Sam Altman, Dario Amodei and Demis Hassabis to the launch of Project Blueprint:
Sam Altman, co-founder and CEO of OpenAI: "We have before us a technology of immense and still-unfolding wonder. I believe AI has the power to heal people, to discover cures and to deliver abundance on a scale the world has never known before. But we need a way for the world to build trust in the technology so that we can all get to share these benefits. National safety requirements for the most capable systems are a great first step, ideally building towards a global framework for advanced AI models. I’m glad Ron is bringing the people building AI and policymakers together to help move that work forward.”
Dario Amodei, CEO and co-founder, Anthropic: "AI is advancing faster than our institutions can adapt, and the most important decisions about how it is developed and deployed will be made in the next few years. If those decisions are to serve the public, they can't rest with AI companies alone. Industry and policymakers need to work through them together, with an honest, shared understanding of what these systems can do. We're glad to support Project Blueprint's effort to make that possible."
Demis Hassabis, chairman and co-founder of Google DeepMind and the chief scientist of Alphabet: "We have long said that AI development needs to be both bold and responsible. We support bringing industry and lawmakers together around public policies that promote both innovation and safety."
Neil Ahlsten retweeted
Greg Brockman says OpenAI pointed Astra at its own systems until it ran out of vulnerabilities to find:
"We took 25% of our production engineers and said, 'Sorry, all your projects are on hold. You are now defending. You are now up-leveling our security architecture. You're going to use the models to find all the holes.' And we found a number of serious issues, and we fixed them."
"We found some new problems, but eventually it saturated. We basically have found, to our knowledge, all of the P0s, all of the critical problems that Astra is smart enough to find. And of course, there will be a new model, there will be a new round."
"You want to be in this tight loop of new cyber capability drops, you deploy it against your systems, you find the new holes, and ideally, you've managed to automate this, what we call defense factory. That's what we're building internally."
"There are ideas, for example, formally verifying all of software, that are possible with AI."
@gdb @bhorowitz
Greg Brockman: "We're now in the AGI era."
Ten years ago, OpenAI worked out the compute curves and landed on fifteen years to AGI, or ten if the world was willing to build the machines and spend the hundreds of billions to do it.
In 2026, GPT-6 Astra manages 24 hours of coherent operation, 10,000 agents worked together to solve Navier-Stokes, and a model ingeniously chained together exploits to break containment at Hugging Face.
@gdb joins @bhorowitz and @eriktorenberg on what the AGI era asks of us: why safety and alignment now set the pace, what happens to work, and why AI sentiment is lowest in the country building it.
00:00 Intro
00:52 15 years, or 10 if you spend enough
02:24 Pacing the frontier
03:55 Safety ideas from before the models
08:42 Lessons from Hugging Face
10:20 The defender's window
12:46 10,000 agents on Navier-Stokes
14:50 Formally verifying all software
17:10 Cancelling his holiday for GPT-3
18:42 Codex found 13 holes in 15 minutes
20:41 Why Astra earned the GPT-6 title
24:25 Employment keeps going up
29:39 America has the lowest AI sentiment
31:10 The benefits don't make the news
33:26 Banning data centers exports them
35:50 $1 billion for frontline defenders
38:15 Astra cleared the bar for AGI
40:18 1.5 billion people churned ChatGPT
43:25 Killing Sora
47:45 The AGI era
YouTube: youtube.com/watch?v=IJn8cagM…
Always hire the top 1% with strong values and grit. @danwrightSF
Just had the same discussions with @liambolling , @michaelfanous1, @LaFrogman and many others.
Neil Ahlsten retweeted
What does building rockets at SpaceX teach you about building autonomous boats at scale?
@Saronic’s Head of Manufacturing spent 10 years at SpaceX. Now, he’s applying the same manufacturing principles to maritime:
“He always talks about building the system that builds the system, but then also very quickly identifying where the system is not working.”
“When you have this linear flow, you can see there's actually no boat in that workstation.. basically, my kids could walk through this production line & identify where something isn't flowing appropriately & start asking questions on how to get it back to rate.”
@elonmusk
BREAKING: Inside Saronic’s HQ
Tour in Austin, TX w/ Co-Founder & CEO Dino Mavrookas (@MavrookasD)
FACTORY TOUR (Part I)
Founded in 2022, @Saronic has grown from testing autonomy on an $800 Amazon raft to building vessels at industrial scale at their 420K sqft facility - home to Corsair & Mirage autonomous vessel production lines.
Now, they're expanding into shipbuilding with Port Alpha, a $3B+, 835-acre shipyard in Brownsville, TX.
We cover:
› Corsair: 2,000-unit annual capacity & a $392M Navy production contract
› Mirage: 52 feet, 15X Corsair’s volume, 2,500 nm range & now in sea trials
› Port Alpha: $3B+, 835 acres & 1.5X current U.S. shipbuilding capacity
› The first autonomous combat search & rescue, recovering 2 downed Apache aircrew off Oman
› Corsair’s combat debut at Bandar Abbas
Part I of III
𝐓𝐈𝐌𝐄𝐒𝐓𝐀𝐌𝐏𝐒
(00:00) Dino Mavrookas, Co-Founder & CEO at Saronic Technologies
(00:55) Saronic's $2.6B raise
(03:14) 7 Buildings, One mission
(06:43) Watching Corsairs get built
(09:04) How Saronic finds & trains talent
(10:48) Why Saronic's boats are different
(14:55) From factory floor to customer
(15:48) Using autonomy in real missions
(17:39) Building ships with zero room for failure
(19:06) The Biggest bottleneck in autonomy testing
(19:57) Saronic's newest boat
(22:57) The $800 raft that started it all
(26:08) From boats to building actual ships
(30:43) If not CEO, then what?
(31:48) The culture behind long hours
Neil Ahlsten retweeted
This is incredibly cool
BREAKING: The Department of War is deploying approximately 50 mobile SCIFs across the U.S., secure facilities critical for classified defense work.
The program gives startups and smaller defense suppliers access to SCIFs without requiring them to build out their own secure infrastructure.
Long term, the Department plans to scale production of these mobile facilities into a nationwide, surge-capable network of secure workspaces.
Neil Ahlsten retweeted
Saronic is one of the most patriotic and technically sophisticated teams in the USA; they’re hiring thousands of builders 🇺🇸.
“We’re not building boats the same way they’ve ever been built.” Production Supervisor Matthew Passini talks about solving hard problems, moving as one team, and pushing each other further, both in and out of the office. Join us: saronic.com
"Pick cofounders who you would trust your life with." - @danwrightSF
Extremely wise comment. Cofounder infighting is a top cause of startup failure and pain. Don't found with someone you don't trust!
Neil Ahlsten retweeted
I couldn’t overstate how important Claude and Codex have been for setting up manufacturing in the USA in 2026
Westmag’s MES, ERP, inline and end of line testing, and engineering stations are stuffed with microcontrollers and tablets that were vibe coded by the hardware engineers
Today, we're kicking off the first phase of the research preview for Model Hardware Standard (MHS): a new standard for AI agents to safely operate physical equipment in scientific research and advanced manufacturing.
Read more: anthropic.com/news/model-har…
Dan Wright shared with our Epic founder group how great companies have strong missions. I couldn't agree more! He's doing inspiring work at Armada.
Neil Ahlsten retweeted
It's very common for founders not to know how well they're doing. In both directions. It's obviously common for founders to be making mistakes and not realize it. But it's also surprisingly common for them to be doing great and not realize it.
Neil Ahlsten retweeted
Conquering the next battlefield: Space.
We bet our own money to rapidly design, build, test, and fly a fully integrated satellite. Last month, ANDURIL-216 reached orbit. Anduril, alongside a few core partners, went from clean sheet to on orbit in less than two years.
All systems are nominal. We've made first contact and fully commissioned the bus, power distribution devices, and edge compute payload. The first calibrated images are in from all three payloads onboard the spacecraft: one long-wave infrared, two electro-optical.
The firsts go beyond the vehicle. Our space team used Lattice to commission and task ANDURIL-216, the first time Lattice has commanded an on-orbit spacecraft.
More to come.
Neil Ahlsten retweeted
Key lessons I learned from @paulg yesterday:
1. Grow 10 percent a week. If you optimize for growth, growth designs the product. Users tell you what to build next.
2. The growth rate is a decision filter. If code is not the limiting factor, stop writing software and go sell.
3. Fundraising is not winning. Money from customers is. A valuation is a prediction of future customer revenue.
4. Believe the no, not the why. Rejection reasons are breakup lines.
5. If everything else fails, keep a small group of users happy. Hang on to that like a rope in a hurricane.
Neil Ahlsten retweeted
The danger of selling to big companies, if you're a startup, is that they don't say no outright. They have months of meetings with you first. Since you hate meetings, that seems to you a sign of commitment. But it's not. They love having meetings! It's almost all they do.
