Portfolio map · updated 19 Oct 2026
Twenty-nine ventures. One operator. AI does the rest.
We build companies that AI runs. Agents do at least 90% of the work, and every venture is marketed through our own rack of real iPhones. Ventures that miss their numbers are shut down. The ones that work are connected, so each makes the others stronger. And every time a smarter AI model ships, all of them get smarter the same day.
Fig. 1 Each dot is a venture, sized by its last two weeks of work. Hover one to see what it connects to and why. Green links carry information; bronze links carry control. The outer ring shows work per day.
Law IDistribution
Most new businesses fail because nobody ever sees them. Ours launch with an audience.
Every venture is marketed through the rack.
The NodeTheory rack is a wall of real iPhones. Each phone has its own SIM card, its own IP address and its own social accounts, and the rack posts carousels and short videos every day. We don’t buy attention from ad auctions. We own the phones that post.
So no venture launches to silence: it has distribution from day one. The same rack is sold as a service through NodeTheory.io, so distribution is a business of its own.
Law IILabor
A company whose work is done by software can grow without hiring anyone.
AI agents do 90% of the work, or we don’t build it.
Code, art, copy, voice, video, posting, outreach, research and support are handled by AI agents. The operator does the 10% that AI is still worst at: deciding what to build, what to cut, and what’s good enough to ship.
Headcount doesn’t grow when the portfolio does.
AI agents · 90
Code · Art and 3D · Copy · Voice · Video · Posting · Outreach · Research · Support
Operator · 10
What to build · What to cut · What ships
Law IIISelection
With AI, starting a venture takes days, not years. Choosing which ones to keep is the real work.
Most ventures get cut. The survivors get everything.
Build
Start cheap. AI agents build the first version in days, so a new venture costs almost nothing.
Test
Each venture gets one test: a number to hit by a deadline, or for a game, a real playtest. We call it the fun gate.
Cut or double
Miss it and the venture is shut down; its code, prompts and accounts are reused. Pass it and it gets more: more phones, more agents, more compute.
Recent decisions
Law IVLeverage
Separate bets add up. Connected bets multiply.
Every venture has to make another one stronger.
29 ventures could be linked 406 ways. We only build the links that move a number: what a customer costs, how fast a venture launches, what a customer is worth, how much it earns.
Each link carries one of two things. Information: leads, audiences, data, content and code passing from one venture to another. Control: AI agents in one venture taking action in another.
Links that move a number12 built of 406
Law VAcceleration
The AI we build on gets smarter every few months. Every venture we own inherits it.
Every new AI model upgrades every venture at once.
AI agents run every link between our ventures. When a smarter model ships, Fable 5.5 and whatever comes after it, every agent gets smarter the same day.
Information links carry more signal. Agents work longer before they need the operator. Links that weren’t worth building with yesterday’s models become worth building. No hiring, no retraining: the whole network gets stronger.
Illustration, not a forecast. Dashed lines are links that become worth building as models improve.
The length of task a frontier AI agent can finish on its own has doubled roughly every seven months since 2019. METR, 2025
The portfolioUpdated 19 Oct 2026
29 ventures in nine sectors, with the rack at the center. Every row is a venture and every square one day of its last two weeks, brighter where more got done. Within each sector, the busiest venture comes first.
2,183
changes in the last two weeks: commits and changed files, across every venture
Where every venture stands
Hosted iPhones that post for you, and the rack behind them
472 changes in two weeksBusiest: NodeControl Console
Games built mostly by AI; one of them will be chosen to ship
761 changes in two weeksBusiest: As Above So Below
Tools that run the network, some of them sold
310 changes in two weeksBusiest: LifeOps
Done-for-you work for local businesses and creators
62 changes in two weeksBusiest: Local Business Directories
Digital products timed to a moment
147 changes in two weeksBusiest: Sunset Syndicate
Subscription apps for the App Store
428 changes in two weeksBusiest: Unafraid · Tailcoach · Redshirt
Single sites for a single business
3 changes in two weeksBusiest: Viridian
Channels and communities
No changes in two weeks
Ventures that do not fit a domain yet
No changes in two weeks
The ThesisScale
| Question | A conventional studio | NodeTheory Ventures |
|---|---|---|
| A new venture needs | a team | a brief and agents |
| Marketing comes from | ad auctions | its own rack |
| Labor grows with | headcount | compute |
| It gets smarter when | it hires | a model ships |
| A failure costs | a year and a team | days, and its parts are reused |
| The next venture starts | from zero | with distribution, an engine and a playbook |
Most companies get stronger when they hire. This one gets stronger when a model ships.
ContactThe operator
If you build, own or run something that belongs in this network, whether a product, an audience, a channel or capital, write to the operator.