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 19, 2026
AI agents reshape cost, workflow, and hiring structures
Practical agent deployments are delivering measurable efficiency gains in video QA, launch testing, and media relations, while the shift to AI-assisted verification is already compressing junior technical hiring at frontier labs.
- Verification bottlenecks now run on tacit knowledge, not raw output volume.
- Multi-agent persona testing before launch is becoming a standard practice.
- Junior technical share at one leading lab fell from 27% to 9%.
Decision model cuts Claude video QA cost by 90% · @shannholmberg Prompt spins up five subagent personas to QA your launch · @RoundtableSpace Open-source tool uses AI agents to automate PR and media placement · @coreyganim Anthropic's junior technical hires drop from 27% to 9% of total · @ccatalini Tutorial shows how to write prompts that unlock more ambitious AI use · @sarthakgh
Data as capital asset, not just training input
Data is being reframed from a commodity input into a financializable balance-sheet asset and a strategic moat, with brokered datasets seen as the seed for custom RL environments, evals, and eventually proprietary enterprise models.
- Balance sheets hold unaccounted data that capital markets have not yet priced.
- The viable entry point for new data companies requires rare industry-specific alpha.
- Short-term brokering funds longer-term proprietary model and eval development.
Starting a data company only makes sense with real industry depth · @danielrupawalla Data brokering seen as path to building custom AI models · @Sbhaiwala03 Balance sheet data could become a new source of business capital · @yrechtman Startup raises $9M to give AI agents unified web data API · @mynameisyahia
Equity airdrops bridge crypto ownership models into traditional finance
SpaceX's plan to distribute one share to every American child is being read as traditional finance adopting the ownership-distribution logic pioneered by crypto airdrops, with the stated goal of shifting the economy from income-based to ownership-based participation.
- Distributing equity to minors creates long-term alignment with a strategic company.
- The structure borrows directly from crypto-native token distribution playbooks.
- Whether regulatory or logistical friction will blunt the model remains open.
Capital concentration squeezes mid-tier venture while AI capex narrows
An unprecedented share of venture dollars is flowing to a handful of firms and companies, mirroring AI infrastructure capex where one chip supplier is capturing roughly two-thirds of projected 2027 spend, leaving smaller players competing for a shrinking residual pool.
- US model token consumption stalled eight weeks ago despite continued capex growth.
- Retail participation in crypto trading produces fee losses that dwarf trading losses.
- One profitable inference company trades below 3x forward revenue, signaling mispricing.
US AI model token consumption flatlines after months of growth · @AllVentured Retail traders lost $280M on Hyperliquid with fees topping trade losses · @mdudas Nvidia captures $690B of 2027 AI capex while 499 others share $340B · @zerohedge Profitable AI inference company trades below 3x projected 2027 revenue · @AustinBarack Venture capital concentration leaves smaller firms squeezed out · @infoarbitrage