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 5, 2026
Agent infrastructure replacing sandboxes and traditional product metrics
The underlying architecture for AI agents is shifting away from ephemeral compute sandboxes toward durable context workspaces, while the product metrics built around UI engagement become obsolete when agents complete tasks with minimal user interaction.
- Sandboxes are too expensive and unreliable for production agent workloads.
- Retention and session metrics cannot capture value delivered by autonomous agents.
- Shared wikis secured over private networks emerge as multi-agent coordination layer.
Agentic AI forces a rethink of how product engagement is measured · @mignano Agent workspaces may replace sandboxes as AI infrastructure matures · @jhleath Shared wiki approach streamlines coordination across multiple AI agents · @iamBarronRoth Agentic products need metrics that reward outcomes, not time in UI · @mignano Agent infrastructure shifts from VMs to durable context workspaces · @jhleath A shared wiki secured via Tailscale lets multiple AI agents coordinate · @iamBarronRoth Anthropic launches multiagent orchestration feature in public beta · @ClaudeDevs
Crypto investing edge requires unique access or information asymmetry
Durable returns in crypto depend on whether an investor has genuine information asymmetry or privileged access, with on-chain gas monitoring and value-accrual analysis cited as repeatable edges over waiting for aggregated reports.
- Token structures linking staking to equity conversion create novel valuation puzzles.
- Novel protocols from trusted developers with tokens attract outsized demand in bull markets.
- On-chain gas data surfaces low-cap opportunities before human analysts report them.
Monitoring gas consumption across chains helps spot low-cap crypto gems · @binji_x Staked token holders can convert to equity after 4 years · @Evan_ss6 Key crypto investing question: do you have unique access or an edge? · @rargulati Novel protocols with trusted devs and tokens see massive demand · @AvgJoesCrypto Studying token value accrual models like Hyperliquid, Aave, and Venice · @0xZergs Tokens let users trade on enterprise value as ownership incentive · @bread_
AI capability gap widens as model layer commodifies
A sharp divergence is forming between the small population experiencing rapidly accelerating frontier AI and the majority still encountering rudimentary chatbots, while the model layer itself commodifies and compute demand surges across the stack.
- Non-frontier models saturate everyday use, compressing model-layer margins.
- Frontier model competition intensifies even as commodification accelerates below.
- Public mental models of AI capability lag practitioner reality by years.
Gap in AI literacy widens as most users stay on outdated free tools · @karpathy AI model layer commodifying as compute demand surges across the board · @Silicon_Data Non-frontier AI commoditizes the model layer, lifting compute demand · @Silicon_Data AI capability debate fractures as experiences diverge sharply · @karpathy Claude Opus 5.5 can replace After Effects for motion design work · @flxrnc Anatomy tool turns text descriptions into interactive 3D machine models · @wheresryan22 Will software engineering skills still matter in the age of AI? · @poteto
Venture capital norms fracturing under valuation and structure pressure
Pro rata honor systems, fund extension dynamics, and elite founder pricing are publicly breaking down, signaling a market where old informal rules no longer hold and early-stage round sizes have ballooned well past historical discovery thresholds.
- Fund extensions and tranched rounds are read by LPs as toppy-market signals.
- Elite pre-seed valuations now routinely reach $5M to $10M before any traction.
- Pro rata rights operate on eroding trust rather than enforceable structure.
Timing is an underrated factor in how early-stage VCs pick startups · @dunkhippo33 Fund extensions and tranched rounds signal a toppy VC market · @MeghanKReynolds Discovery-stage VC checks still active but round sizes have grown · @TurnerNovak Elite founders now command pre-seed rounds of $5M to $10M · @alexiskold Venture pro rata rights operate on a crumbling honor system · @credistick Old unspoken rules of venture capital are publicly breaking down · @rubinovitz Emerging managers should maximize ownership and plan for secondaries · @PeterJ_Walker Early Silicon Valley VCs were visionaries, not benchmark chasers · @Sethwinterroth Discovery-stage VC checks still active but round sizes have grown · @TurnerNovak