Some things I've been thinking about -
This is the second week I’ve experimented with writing a “Some Things I’ve been thinking about -” post on Twitter & LinkedIn.
The idea is for it to be basically a brain dump of whatever has been top of mind for the week -
It was inspired by Jeff tweeting notes from our Investment Committee every week. It’s been really cool to see how many people respond - both publicly and privately - so I thought I’d try something similar.
I’m still figuring out the right cadence for Substack, maybe a roundup every 2 weeks?
Here’s the latest:
Some things I’ve been thinking about (August 4th)
Consumer FinTech
We spent all last week sprinting on diligence for a consumer fintech company. And yesterday, decided to invest. I’m so pumped - the founders are insanely brilliant.
Consumer fintech is a graveyard. We spent a lot of time on the question: what’s actually changed in user behavior?
We think Kalshi/Polymarket, mobile gaming, crypto, and in-app currencies have made money more interactive, social, and experiential. We also think there’s a whole new category of consumer-finance products to build alongside this momentum - and that this company has a real shot at defining it. Looking forward to announcing this one!
AI Services & Physical AI
It feels like 95% of the decks I’ve read this month are either in AI Services or Physical AI.
AI Services is the AI-transformation pitch - usually with some sort of FDE (forward-deployed engineer) flavor. If I weren’t in venture, this is probably what I’d be building because there is so much money to be made. But it’s also a land grab, and we seen a lot of companies that are almost identical on paper. Our #1 question is usually: how aggressive is the founder?
On Physical AI, I was looking at the numbers from robotics investments in Crunchbase, and it’s pretty wild: - $11B into robotics in Q1 2026 - H1 2026 has already out-raised 2023 and 2024 put together) - 41% of this year’s dollars went to just 10 rounds (Helsing, Saronic, Atoms, Skild AI, Mind Robotics, NEURA…)
Situational Awareness
How would it be possible to write a “things I’ve been thinking about” post and not include it? It’s just all so nuts. And consumed my entire Twitter timeline for 72 hours.
For anyone living under a rock:
Leopold Aschenbrenner’s AI fund was up 439% through June. 2.
It then lost 67% in July after its leveraged AI bets moved against it.
The fund was forced to sell most of its public book to Citadel - but is somehow still up around 80% for the year (according to the investor letter).
There are already so many hot takes, so you definitely don’t need mine. But wow. What a journey haha.
Something random
Last week, I ran a Mafia game for startup/venture people.
My husband and I play every few months with friends, so when Founders Fund released their show - I obviously had to run a Mafia venture event.
Game highlight: Jeffrey Chang was one of the Mafias - pretended to be the Seer - and it was so epic. 10/10. When you’re the host, this is truly the best-case scenario for it to be so chaotic and fun.
Claude highlight: I asked whether I should have 2 or 3 Mafia for 14 people, and it decided to run a 30,000-game simulation unprompted (the answer was 3 haha)
If you want to play in one of my future games, fill out this form: https://docs.google.com/forms/d/e/1FAIpQLScrw7pcj7O-6fWu8790s1Z7QMHQX3uLJRgwee_EUQFf5iOwnA/viewform?usp=dialog
Something personal
This is my second week with a Google Fitbit Air - and I don’t hate it.
Pros: no screen, weightless, no subscription
Cons: how is Google one of the most incredible companies in the world and the app is this horrible?
Some things I’ve been thinking about (July 28th)
Horizontal Agent Infrastructure
I’ve seen three “file systems for agents” companies this week.
This gets at what’s valuable in horizontal agent infra -
File systems, sandboxes, browsers, and memory are all clearly useful. But distribution is insanely hard.
Enterprise has the money, but buys on compliance - and compliance is a terrible place to build momentum. Prosumer adoption can create that momentum, but it can be hard to see the path to venture scale. Developer pull works - Exa is a good example - but developers are obviously prickly.
Flybridge wrote a piece called “Agentic File Systems” arguing that companies should pick a regulated vertical and own permissions and audit trails there. I think that’s a good take. It’s also a pretty hard company to raise a pre-seed round for. I don’t have the answer, and this tension is not new, but I’m thinking about it a lot because there is still so much left to build here.
Prediction Markets
I recently googled: “Does winning the World Cup increase a country’s GDP because of betting wins?”
The answer is no haha. There is usually some bump, but it comes from improved sentiment and higher consumption.
I watched the prediction markets closely throughout the World Cup.
Volumes: Kalshi did $31B in notional volume in June, up 70% from May. Polymarket’s international book hit a record $10.8B. Rothera - the Susquehanna/Robinhood JV - launched in June and did $2B in its first month.
Fees: Kalshi has generated $1.15B in fees since launch, with $850M coming this year. Now that the retail market has been activated, more people are starting to think about the liquidity layer.
USV leading Pascal’s Series A feels like part of the same thesis. The company was founded by ex-dYdX traders and is building for professionals who are underserved by Kalshi and Polymarket on order types and execution.
Physical AI
There have been a bunch of huge physical AI rounds announced in the past couple of weeks: -
Senra Systems : $65M Series B for wire-harness manufacturing in aerospace and defense. Lowercarbon and Interlagos led, with Sequoia, Founders Fund, a16z, and Dylan Field in.
Walden Robotics : $300M seed at a $1.1B valuation, spun out of Toyota Research. Founded in January and already running inside a Toyota plant.
TerraFirma : $100M Series A led by Kleiner. Two ex-SpaceX founders turning excavators and bulldozers into robots.
Atoms: $1.7B from a16z for Travis Kalanick ’s new company.
They’re across three different areas -
Walden is focused on general-purpose robotics, where one robot can learn different tasks and work across environments.
Atoms is the opposite. a16z’s argument for investing $1.7B is that specialized robots will beat humanoids at almost every real job. TerraFirma’s thesis is similar.
Senra is focused on industries like aerospace and defense, where the work is hard to automate and the workforce is aging. Its software makes wire-harness manufacturing easier to learn and more reliable, cutting training from 18 months to four weeks and bringing first-pass yield up to 99%.
That is basically a bet that the robots won’t arrive before the workforce leaves.
Something simple/random I built this week
Scraper to search for IMAX Odyssey tickets for my Dad in New York. Lincoln Center is actually sold out (as reported).
His best bet ends up being driving 35 minutes to West Nyack haha. Yikes.
I saw it non-IMAX and am telling myself I didn’t miss out on anything.
Something personal
I’d basically stopped working from an office because I use Wispr Flow all day and I can’t Flow around other people.
But to save my sanity, I’ve started going back in - and it does feel nice to be part of society again.
I’m a General Partner at Chapter One, an early-stage venture fund that invests $500K - $2M checks into pre-seed and seed-stage startups.
If you’re a founder building a company, please feel free to reach out on Twitter (@seidtweets) or Linkedin (https://www.linkedin.com/in/jamesin-seidel-5325b147/).