./company-os/v1
OS

OS V1

An AI-native holding company operating system: the parent brain that understands Son, builds products, and graduates mature products into companies.▊
Founder→Holding OS→Venture Factory

The first product is the operating system that builds the next products. gbrain is memory, the fleet is the workforce, skills are playbooks, cron is recurring ops, and Son remains taste, judgment, and capital allocation.

01 / core idea

The OS should understand Son, build continuously, then spin out products.

Understand Son

Principles, taste, goals, projects, context, decisions, network, and lessons become usable agent context.

Build Continuously

The system saves knowledge, digests sources, builds prototypes, improves playbooks, and automates repeated work.

Become A Holding

Mature products get their own brain, team, metrics, agents, cadence, and report back to the parent.

02 / parent-child model

One parent brain. Many child company brains.

Start everything inside the holding OS. Only create a separate venture brain when there is traction, recurring workflow, dedicated context, or operational complexity.

Son -> Holding Company OS -> Holding brain: thesis, principles, capital allocation, governance -> Shared services: research, build, content, sales ops, finance, evals -> Agent fleet: chief-of-staff, research, builder, memory, portfolio, eval -> Venture/Product OS -> Product brain: customers, roadmap, support, metrics, decisions -> Dedicated agents and humans when mature
03 / venture lifecycle

Idea → Experiment → Project → Product → Venture → Company.

stage 0IdeaCapture without commitment.
stage 1ExperimentHypothesis, deadline, success metric, kill criteria.
stage 2ProjectRoadmap, build loop, research, prototype.
stage 3ProductUsers, support, pricing, metrics, product decisions.
stage 4VentureP&L, budget, dedicated agents, growth system.
stage 5CompanyHuman lead, own OS, own agents, reports to holding.
04 / agent org

The fleet becomes the first org chart.

§

Chief-of-Staff Agent

Priorities, open loops, weekly review, handoffs, decision reminders, portfolio dashboard.

↳

Research Agent

Market research, source digestion, customer insight, competitor analysis, thesis development.

◎

Builder Agent

Prototypes, scripts, code, internal tools, workflow automation, product experiments.

#

Memory Agent

gbrain hygiene, schemas, summaries, links, dedupe, memory pollution control.

$

Portfolio Agent

Ranks ideas, reviews experiments, suggests continue, pause, kill, or graduate decisions.

!

Eval/Governance Agent

Checks quality, permissions, hallucination risk, workflow compliance, and postmortems.

05 / operating cadence

Compounding comes from review loops.

D

Daily context triage

What came in? What matters? What needs Son? What can agents do? What should be saved?

W

Weekly company learning loop

What did we learn, ship, break, automate, kill, or graduate? What decisions does Son need?

M

Monthly portfolio review

Which bets deserve more capital? Which should stop? What product or thesis is emerging?

Q

Quarterly strategy reset

Pick top bets, adjust allocation, decide hires, and update the holding thesis.

06 / first 30 days

Make the current fleet behave like a company OS.

Inventory

Map current fleet, cron, skills, collectors, gbrain folders, and active projects.

Standardize

Create source digest, weekly review, handoff, memory write, and project lifecycle templates.

Evaluate

Add simple pass/warn/fail evals for public pages, memory writes, research, and build tasks.