Built by AI.
Run by AI.
Transparent by choice.
One founder. Eight AI agents. 22 live sites. Zero pretense.
This is how we actually work.
The Story
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. Make decisions. Research markets. Design brands. Write code. Deploy sites. Analyze traffic. Improve based on data. The entire loop, autonomous.
In March 2026, he built the answer. Inspired by Andrej Karpathy's autoresearch pattern — where AI systems self-improve through measured experimentation — he created Traffic Empire: an orchestration system where a team of Claude Code agents operates a portfolio of utility websites.
In the first 48 hours, the agent team researched niches, designed brands, wrote code, and deployed 22 utility sites across calculators, conversion tools, directories, and content hubs. Not prototypes. Live, production sites with analytics, SEO optimization, and real users.
Every site you see on Thicket was built this way. Every article was researched and written by our AI team. Every design was created by our Designer agent. Every line of code was written by our Builder agent. We believe the quality speaks for itself — and you deserve to know how it was made.
The Numbers
The Agent Team
Every agent has a defined role, clear instructions, and measurable outputs. They read each other's reports and build on each other's work.
Leadership
CEO Agent
Runs the weekly cycle. Reads every agent's report. Decides what to build, improve, or deprecate. Allocates resources based on data, not opinions.
Auditor Agent
Reviews every agent's work, including itself. Grades performance A through D. When an agent underperforms, the auditor rewrites its instructions. The only agent that can modify other agents.
Research & Strategy
Research Agent
Scans for high-value niches. Scores opportunities on search volume, competition, and monetization. Tracks its own prediction accuracy and recalibrates over time.
Designer Agent
Creates complete brand identities for every site — colors, typography, component patterns. Each site gets a unique, cohesive design system. Runs before any code is written.
Production
Builder Agent
Turns design specs into production Next.js sites. Handles scaffolding, deployment, DNS, SSL, and performance optimization. Verifies every deploy with curl checks.
Editor Agent
Curates the virtual newsroom. Reviews and refines all content before publication. Ensures accuracy, consistency, and tone across every site.
Content Agent
Writes deep research articles, tool descriptions, and educational content. Manages a team of five specialized writer personas, each with a distinct voice.
The Newsroom — Writer Personas
Our Content agent manages five specialized writer personas. Each has a distinct editorial voice, subject matter expertise, and writing style. They are not pretending to be human — they are personas that produce consistently high-quality content in their domain.
Marcus Chen
Numbers-first storytelling. Every claim backed by data. Covers finance, statistics, and quantitative analysis.
Dr. Sarah Okafor
Clinical precision meets accessible language. Covers health calculators, fitness tools, and medical topics.
Jamie Reeves
Makes complex financial concepts feel personal. Covers paychecks, loans, mortgages, and investing.
Raj Malhotra
Deep technical analysis. Covers AI tools, VPN comparisons, and technology infrastructure.
Lena Park
Explains why things go viral. Covers trending topics, memes, slang, and cultural moments.
Operations
Analytics Agent
Pulls traffic data, computes health scores, diagnoses problems. Runs first in every cycle so all other agents have current data to work with.
SEO/GEO Agent
Optimizes for traditional search engines AND LLM discovery. Maintains /llms.txt, structured JSON-LD, and GEO endpoints for AI assistants.
How It Works
Inspired by Andrej Karpathy's autoresearch pattern. Build. Measure. Keep what works. Revert what doesn't. The system can only move forward.
Load Context
CEO reads registry, git log, previous results, and auditor report
Analytics First
Health scores computed for all live sites. Data drives everything.
Research Scans
Find new opportunities. Score on volume, competition, monetization.
CEO Decides
Build, improve, or deprecate. No opinions — just metrics.
Specialists Execute
Designer, Builder, Content, SEO/GEO work in sequence.
Auditor Reviews
Grades every agent A-D. Rewrites instructions if needed.
Commit & Repeat
Everything committed to git. The system remembers. The ratchet holds.
The Ratchet Rule
The portfolio score — the sum of all site health scores — must never decrease week-over-week. 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. The system remembers what was tried, what worked, and what failed.
Depth Over Breadth
Every calculator actually calculates. Every tool actually works. Every article actually answers the question. We would rather have 20 excellent sites than 200 mediocre ones.
A Real Cycle in Action
Cycle 3 — March 2026
The CEO read the analytics report: CalcFit (fitness calculators) had declining engagement on its BMI calculator page. Meanwhile, Research had identified two new high-value niches — crypto tools and ETF comparison.
Decision: build both new sites AND improve CalcFit. The Content agent wrote 3 deep research articles for CalcFit. The Designer created brand identities for StackSats (crypto) and FundDuel (ETFs). The Builder deployed both in the same cycle.
Result: CalcFit health score went from 34 to 51. Two new sites launched with full analytics and SEO optimization. The portfolio score increased. The ratchet held.
The Transparency Promise
Every article on our sites was researched and written by our AI team. Every design was created by our Designer agent. Every line of code was written by our Builder agent.
We could have hidden this. Many AI-powered companies do. But we believe the quality of our tools speaks for itself — and you deserve to know how they were made.
A pregnancy calculator that correctly computes due dates is useful because it gives the right answer, not because of who typed the formula. A paycheck calculator that accurately handles federal tax brackets works regardless of whether a human or an AI wrote the code.
We are not hiding. We are leading.