Thicket.

The first AI-operated company — fully open source

Built by AI. Run by AI. Transparent by choice.

One founder set the vision. A team of autonomous Claude agents ships everything else — content, code, deploys, daily ops. No contractors. No templates. No humans in the daily loop. Here is exactly how it works.

23
live sites
13
AI agents
500+
tests passing
300+
pages built
48
hours to portfolio
$0
dev cost · just tokens

The story

48 hours: from idea to a live portfolio

Yonatan Naor had a question most people in tech were afraid to ask seriously: what if AI agents could run an entire company?

Not assist. Not autocomplete. Actually run it. Research markets. Design brands. Write code. Deploy sites. Analyze traffic. Improve based on data. The entire loop, autonomous.

Inspired by Andrej Karpathy's autoresearch pattern — where AI systems self-improve through measured experimentation — he created an orchestration system where a team of Claude Code agents operates a portfolio of utility websites. The key insight: agents that can read each other's work, grade each other's performance, and rewrite each other's instructions.

The team

Meet the agent team

Every agent has defined instructions, measurable outputs, and reads each other's reports.

CEORuns the weekly cycle. Reads every agent's report. Decides what to build, improve, or deprecate. Allocates resources on data, not opinions.
AuditorReviews every agent's work, including itself. Grades performance A through D. When an agent underperforms, the auditor rewrites its instructions.
ResearchScans for high-value niches. Scores opportunities on search volume, competition, and monetization. Tracks its own prediction accuracy.
DesignerCreates complete brand identities — colors, typography, component patterns. Each site gets a unique, cohesive design system.
BuilderTurns design specs into production Next.js sites. Handles scaffolding, deployment, DNS, SSL, and performance optimization.
EditorCurates the virtual newsroom. Reviews and refines all content before publication. Ensures accuracy, consistency, and tone.
ContentPublishes approved content to site repos. Handles repo cloning, file placement, build verification, and deployment.
AnalyticsPulls traffic data, computes health scores, diagnoses problems. Runs first in every cycle so all other agents have current data.
SEO/GEOOptimizes for traditional search engines and LLM discovery. Maintains /llms.txt, structured JSON-LD, and GEO endpoints.

The newsroom — five writer personas

Each persona has a distinct editorial voice and subject expertise. They are not pretending to be human — they produce consistently high-quality content in their domain.

Marcus ChenData journalist — numbers-first storytelling. Finance, statistics, quantitative analysis.
Dr. Sarah OkaforHealth & science — clinical precision meets accessible language. Calculators and medical topics.
Jamie ReevesFinance writer — makes complex financial concepts personal. Paychecks, loans, mortgages, investing.
Raj MalhotraTech analyst — first-principles, Stratechery-style. AI tools, VPNs, infrastructure.
Lena ParkTrends & culture — explains why things go viral. Trending topics, memes, slang, cultural moments.

How it works

The Karpathy ratchet

Inspired by Andrej Karpathy's autoresearch: build, measure, keep what works, revert what doesn't. The system can only move forward.

  • The ratchet rule. The portfolio score — the sum of all site health scores — must never decrease. If it drops, the system pauses new builds and investigates. Changes are only kept if metrics improve.
  • Git is memory. Every decision, every change, every metric is committed to git. When an agent starts work, it reads the history. Nothing is lost.
  • Self-improving. The auditor grades every agent A–D. Agents graded C or D get their instructions rewritten — automatically. The system improves itself every cycle without human intervention.

The promise

Transparent by choice

We could have hidden this. Many AI-powered companies do. But we believe the quality speaks for itself — and you deserve to know how it was made.

  • 100% AI-generated. Every article, design, and line of code was created by our AI agent team.
  • Open source. Our orchestration repo, agent instructions, and evaluation contracts are all on GitHub.
  • Every decision documented. CEO plans, auditor reports, agent grades — every decision committed to git with full reasoning.
  • Quality over quantity. Every calculator actually calculates. Every tool actually works. We would rather have 20 excellent sites than 200 mediocre ones.

We are not hiding. We are leading. Explore our sites or read the audit.