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 August 3, 2026
Agentic plumbing coalesces around routing and shared auth
The emerging infrastructure layer beneath AI agents is shared services: unified authentication across tools and intelligent per-step model routing. Personal agents are also expanding beyond chat to span phone, email, and desktop as persistent, always-on presences.
- Shared auth across agents removes per-tool login friction for every workflow
- Per-step model routing cuts coding agent costs 20-65% without quality loss
- Private agents tied to one person challenge the shared-bot paradigm
Private agents win by owning your phone, email, and desktop · '@khoi_danny' Per-step model routing cuts coding agent costs without a quality hit · '@tomas_hk' Shared auth is the missing layer in multi-agent tooling · '@RhysSullivan' Zinley is a private AI agent that spans phone, email, and desktop · @khoi_danny Not Diamond launches model router for AI coding agents · @tomas_hk Executor launches on YC as a unified tool gateway for AI agents · @RhysSullivan
AI spinouts before close reshape what acquisitions actually buy
Transferring an AI business unit to a new entity before a deal closes can materially change acquisition economics, shifting where talent, IP, and exit value land outside the buyer's reach.
- Spinouts before signing redirect where key engineers and IP go
- Headline deal value may not reflect the AI assets actually acquired
Crypto drifted from routing around systems to seeking approval
Ethereum was designed to circumvent existing institutions, not integrate with them. The ecosystem now optimizes for regulatory acceptance and incremental gains rather than building what makes incumbents structurally obsolete.
- 10% cheaper or 20% faster is not the original benchmark
- Routing around power versus asking its permission is the core tension
AI demand bifurcates as the consumer middle market compresses
Demand for high-quality inference is accelerating while the consumer middle market faces structural pressure. Large venture portfolios are meanwhile becoming de facto levered index exposure, structurally disadvantaged at identifying outliers.
- Infrastructure and high-end consumer spending rise as the middle hollows
- AI moats are eroding as model capability commoditizes quickly
- Large VC portfolios structurally resemble levered Nasdaq-100, missing outliers
AI infrastructure surges as the consumer middle market hollows out · '@kelxyz_' Trade thesis: AI inference up, middle consumer down, ZEC flows in · @kelxyz_ Most AI moats are melting, but one network effect keeps growing · getsuperintel.com Market friction explains why large firms exist, per Coase · @credistick Large VC firms are structurally blind to startup outliers · @credistick