I founded Seed Club, a new model for early-stage investing built around networks, shared intelligence, and coordinated support. 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 July 20, 2026
AI displacement makes ownership the only durable career hedge
As AI automates time-sold labor across coding, design, research, and creative work, ownership of productive assets becomes the primary remaining source of durable individual economic power.
- Careers built on selling time are structurally exposed once AI is trained on those tasks.
- Ownership technology remains under-discussed relative to the scale of displacement underway.
Autonomous agent infrastructure is maturing rapidly in practice
Tooling for building, running, and optimizing AI agents is moving from theory to production, with live trace-score-optimize loops, social media scraping at near-zero cost, and local fund models syncing live data via MCP.
- Agent-to-agent communication will likely remain plain English absent a REST-level protocol insight.
- Continuous prompt optimization via live eval loops is now an operational pattern.
- Commodity scraping tools are collapsing the cost of data acquisition for agents.
AI agents can now scrape social media with no login needed · @shengkun_ye Agent-to-agent communication will stay in plain English · @threepointone VC fund model syncs with Carta to reprice NAV and carry instantly · @jtriest Carta synced to a local model becomes a live NAV and carry engine · '@jtriest' How to improve live agents with a trace-score-optimize loop · h100envy (@h100envy)
AI commoditization threatens big labs' core business model
If open models and cheap clones erode the margin on inference tokens, the large frontier labs lose the revenue needed to fund future training runs. The US government has no obligation to protect that business model, and American users would likely be fine either way.
- Distillation and Chinese alternatives undercut proprietary token pricing.
- Open question: can any lab sustain frontier training without metered inference profit?
- Nearly every high-skill task is commoditizing faster than differentiation can form.
Private market equity is structurally mispriced against buyers
Pre-IPO equity behaves like a scarce asset that appears to only rise in value, but over-funded companies deliver the worst post-listing returns, and the entire private market structure favors sellers over buyers.
- Companies raising over $3B pre-IPO show the worst 24-month post-IPO alpha.
- Scarcity and excess demand let sellers extract prices disconnected from fundamentals.
- Retail access to 'high quality' private deals conflicts with founder-controlled supply allocation.
Democratizing top startup deals is impossible, founders control supply · @matty_ Retail investors need access to quality private market deals · @darrenmarble Pre-IPO equity is structurally overpriced and risky for buyers · @credistick Over-funded IPOs suffer worst post-listing returns, data shows · @credistick