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 November 9, 2026
Agentic coding demands empirical, not engineering, discipline
Producing software with AI agents is structurally closer to ML than to traditional engineering: outputs are probabilistic, iteration is empirical, and the intuitions from conventional software craft often lead astray. Best practices are still being discovered.
- Agent experience (AX) is now a first-class design concern alongside UX and DX.
- Implicit state machines are a new surface for agent-aware code improvement.
Build AI-native competitors, not tools for incumbents
Enterprise AI transformation takes roughly a decade due to legacy systems and organizational inertia, giving AI-native startups a structural window to build ground-up replacements. A small, AI-fluent team now operates closer to an entire company's output than ever before.
- Selling AI tools to incumbents is slower and harder than replacing them.
- The moat is lean teams with no legacy playbooks, not capital or headcount.
Agents crossing into professional liability and accountability
The first malpractice insurance policy covering AI agent actions marks a structural shift: agents are being held to professional accountability standards rather than treated as software tools. Insurance infrastructure for autonomous actions is now being priced and issued.
- Measuring agent outcomes with evals is prerequisite to insuring them.
- Professional specialization erodes as agents insource previously rare skills.
VC capital concentrated late, starving early-stage discovery
Of roughly $500B deployed in venture in 2026, only $10B reached emerging managers, while most flowed into later-stage deals resembling private equity. Small funds remain the primary source of alpha discovery but are structurally undercapitalized.
- 71% of unicorns remain private, compressing DPI across the asset class.
- LPs are bypassing GP relationships to pursue direct co-investment discounts.
Just $10B of $500B in 2026 VC reaches emerging managers · @adeoressi Most unicorns never exit, with 71% still private, Stanford data shows · @IlyaStrebulaev Small VC funds source unicorns that megafunds later back, data shows · @arian_ghashghai LPs want co-investment discounts, some bypassing GPs altogether · @MeghanKReynolds