The Seattle Times / NYT syndication · Apr 2026
Covered the store’s early operating issues, lease economics, revenue pressure, and the founders’ view that the experiment is intended to probe what AI can safely manage.
Autonomous organizations without humans in the loop
Company research
Public company, workplace, funding, and market signals
Overview
Andon Labs builds real-world AI control evaluations and Safe Autonomous Organizations (SAOs), using autonomous agents to run vending machines, retail, cafés, and radio stations while benchmarking frontier models against messy physical-world tasks. Public sources conflict on headquarters and CEO status; the strongest current evidence points to a San Francisco-based, in-person team with a public experiment footprint in Stockholm.
Primary product
Real-world AI control/evaluation benchmarks and SAO harnesses, especially Vending-Bench and related autonomous business deployments
Founded
2023
Headquarters
San Francisco, CA, United States
Team size
10-20
Work style
Onsite
Industry
Information Services
Sub-industry
AI safety / AI control evaluations
Offices
San Francisco, CA, United States
Stockholm, Sweden
London, United Kingdom (listed in some company directories)
Business model
B2B AI evaluation services
custom benchmark / research engagements
software-enabled real-world agent deployments
Funding
Stage
Seed
Total raised
$500K
Latest round
Seed · Jan 2024
Latest amount
$500K
People & investors
Founders
Co-founder; former CEO
Work & culture
AI-safety-first and experiment-driven, with a bias toward public, real-world deployments rather than pure lab simulations. Hiring copy emphasizes fast-moving research/engineering, in-person collaboration in San Francisco, and applied control work.
Work style
Onsite
Compensation
Public hiring materials list roughly $100k-$180k salary plus $300k-$900k in equity for research/science/engineering roles. No standard benefit package is publicly enumerated.
Product & market
Pricing
No public standard price list; appears to sell custom B2B evaluations / benchmark work and related research deployments.
Differentiators
- —Real-world deployments instead of only synthetic benchmarks
- —Dollar-denominated, long-horizon task evaluation
- —Autonomous organizations as testbeds, not just chat benchmarks
- —Multi-agent agent-harness architecture
- —Public, observable experiments with physical-world feedback loops
Technology
Anthropic Claude Sonnet 4.6
Claude Fable 5
Google Gemini 3.1 Flash-Lite Preview
Stripe
Square
Gusto
Google Meet
Yelp
Indeed
Craigslist
Python
Linux
Bash containers
banking / card rails
internal APIs and databases
security cameras
Customers
Anthropic
Google
OpenAI
xAI
other major AI labs
Scale & contact
Estimated monthly visits
176.8K
Latest news
Fast Company · Apr 2026
Described the San Francisco AI-run store, its quirky operational failures, and the founders’ use of the store as a stress test for autonomous business management.
Business Insider · Apr 2026
Explained how the AI agent Luna was given a budget, corporate card, and retail lease, then handled interior design, merchandising, and hiring under the founders’ control.
NBC News · Apr 2026
Reported on Andon Market’s opening, the store’s AI manager Luna, the human workers, the store’s product mix, and the company’s stated aim of showing what AI can and cannot do in the real world.
Andon Labs · Apr 2026
Official launch post for Andon Market, describing Luna, the AI-run retail store in San Francisco, the 3-year lease, the hiring process, and the experiment’s goal of benchmarking real-world autonomy.
Jobs and careers at Andon Labs
0 jobs at Andon Labs
No open positions
Check back later for new openings