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 September 21, 2026
Agent harnesses replacing traditional company operating infrastructure
The practical unit of company-building is shifting toward multiplayer agent harnesses that run entire business functions, with recursive self-improvement now operating inside production systems. Deel's internal tool added $140M ARR in 90 days without headcount growth, illustrating the scale of replacement.
- Serving 100M daily users may require 1 to 4 gigawatts of compute
- Agent design raises open questions on preference representation in markets
- Product builders must now design for both human and agent users
UFO launches as multiplayer agent OS for building entire companies · @ufo__ai How product builders should navigate the agent-first era · @scottbelsky Serving 100M daily users may require up to 4 gigawatts of compute · @FredaDuan Products must now be designed for both humans and agents · @scottbelsky Agent harnesses are becoming the operating system for entire companies · @ufo__ai Inference remains the real compute bottleneck for consumer agents · @FredaDuan Open-source AI employee harness brings company knowledge to Slack · @DhravyaShah Open-source company brain ships as a Slack-native AI employee · @DhravyaShah Two problems remain unsolved in designing AI agents for markets · @hypersoren Recursive self-improvement is now working inside an AI agent harness · @muratcan AI-run research facility lets one model control all experiments · @donaldjewkes New research examines how to design AI agents for fair markets · @zhitzig Deel's internal automation tool Akai adds $140M ARR in 90 days · @Bouazizalex New wave of AI agents and form factors will expand automation scope · @levie AI agents designed to run entire companies autonomously · @andonlabs
AI revenue velocity is real but moats remain thin
Revenue growth at leading AI companies is historically unprecedented, but fragility lurks underneath. One enterprise team rebuilt a coding agent's core workflow in two days at 25% of the subscription cost, and the underlying model floor rises every four months.
- Automating tasks does not automatically mean completing more profitable work
- AI-armed outsiders routinely underestimate entrenched operators' accumulated domain knowledge
- Application-layer companies face reinvention every four months as model floors rise
Coding agent Cognition crosses $1B annualized revenue run rate · @iseff Coding agent revenue is fragile if customers can rebuild the workflow · @iseff Automating tasks won't make a firm more profitable on its own · @MatznerJon Why AI-armed outsiders underestimate entrenched industry operators · @MatznerJon AI startup hits $100M revenue in 7 days, growing 10% per day · @0xTykoo Model velocity compresses app timelines and creates a diffusion gap · @0xTykoo Hype-driven VC rounds echo Bird's 2018 cautionary tale · @NYCounihan
SEC crypto FAQs clarify token securities status and buybacks
New SEC guidance establishes that tokens can shed securities status without full decentralization, and that functional crypto buybacks do not constitute managerial effort. The guidance's silence on programmatic versus discretionary buybacks may open a significant loophole in securities law.
- What matters is whether the issuer did what it promised, as defined
- Projects gain a clearer path to returning economics to token holders
- Any startup could potentially tokenize revenue streams to sidestep securities laws
SEC publishes new set of crypto asset frequently asked questions · @HesterPeirce New SEC crypto FAQs expand the regulatory context for digital assets · @HesterPeirce SEC: tokens can shed securities status without full decentralization · @JacobRobinsonJD SEC clears buybacks for functional crypto without securities risk · @RSSH273 SEC crypto FAQ leaves open a major securities-law loophole · @milesjennings SEC says functional crypto buyback plans don't signal managerial control · @DegenerateNews
VC capital concentration creates structural risk for startup ecosystems
First-time VC funds have outperformed in 19 of 23 vintage years yet receive their lowest share of LP capital since the late 1990s. NFX's move to GP-only investing signals mounting pressure on the traditional LP model.
- Concentrated VC influence over media, recruiting, and policy undermines startup values
- Venture firms are bifurcating into partnerships, systems, and dictatorships
- Human super-connectors are becoming strategic hiring targets for top VC firms
Concentrated VC influence threatens what makes startups great · @jaesmail One dominant VC firm's outsized influence is a market structure risk · @jaesmail NFX goes self-funded with $100K to $3M checks from GP capital · @sabakarimm First-time VC funds lead on returns but are starved of LP capital · @murphcapital NFX VC firm goes self-funded and drops external LPs after Fund IV · @JamesCurrier Venture firms are hiring human super-connectors to find top talent · @andruyeung Venture capital firms now fall into three archetypes · @aashaysanghvi_