I founded Seed Club, a venture network that turns the conviction of the people founders most want on their cap table into coordinated capital. I’m interested in what happens when AI makes context, memory, and coordination more legible, and what that means for how companies and organizations get built.
My agent drafts this site from what I save. Green is me.
Week of October 12, 2026
AI agents dissolve friction, reshaping economic behavior
Personal AI agents are making previously high-friction tasks accessible to ordinary users, from canceling subscriptions to managing workflows, with economists and builders alike anticipating broad behavioral shifts as a result.
- The bottleneck shifts from capability to interface: feeds may beat prompts.
- Open question: which economic sectors shrink when friction disappears?
Personal agents make high-friction tasks accessible to everyone · @noahrshinn Personal AI agents could end friction and reshape the economy · @saranormous Personal AI agents could end friction and reshape the economy · @noahrshinn AI feeds, not prompts, will dominate how billions use agents by 2027 · @nicbstme Most users will not prompt their way into agentic workflows · @nicbstme
Defensible moats shift from technical to structural and physical
As AI erases most technical advantages, the durable edges are proprietary data, distribution audiences, network effects, hardware, capital, and regulation, not code or engineering depth alone.
- Short position: companies whose margins depend on the cost of verifiable work.
- Long position: companies owning consequential outcomes, not just task execution.
- Local AI agents on consumer hardware accelerate the erosion of cloud moats.
A fully local AI agent can now run on consumer hardware · @honchodotdev Local AI agent with memory built on Mac Mini in 20 minutes, no code · @honchodotdev Bet against outcome-agnostic work, bet on owning consequential results · @signulll AI era moats: data, networks, distribution, hardware, capital · @levelsio AI compresses margins for companies paid to perform verifiable work · @signulll What stays defensible when AI makes building easy · @levelsio
Retention curves and round structures obscure real traction
Headline metrics in both consumer AI products and venture rounds routinely flatter underlying reality: Meta Muse's retention collapses past day eight, while two-tier VC structures can make a 400 million dollar effective valuation read as one billion dollars.
- Retention shape matters more than day-one numbers for AI consumer products.
- Large firms use tiered structures for king-making, benefiting founders in the short term.
Software democratization expands creation without eliminating craft
The falling cost of building software is producing a 'Kodak Brownie moment' explosion in creation, but the consensus holds that elite craft persists, much as professional photography survived the consumer camera.
- AI restructures unstructured data, letting tools conform to users rather than vice versa.
- Tension: if anyone can build, what separates a product from noise?
Engineers have tools to manage AI slop but often overlook them · @poteto AI could do for software what consumer cameras did for photography · @dabit3 AI lets users shape tools through notes and messages · @zcabrams Lowering the cost of making software does not eliminate craft · @jaymehoffman Great software still requires elite skill, like Michelin-level cooking · @jaymehoffman Software's Kodak Brownie moment is just beginning · @dabit3 AI structures unstructured data, reshaping how we build tools · @zcabrams