Good morning, {{ first_name | AI enthusiasts }}. Most AI agent demos live inside sandboxes with fake money and simulated users. This one signed a three-year lease and hired real people.
Andon Labs' latest experiment just dropped an AI into its own retail store in San Francisco with a $100K budget and full autonomy over hiring, operations, and more, a preview of a future where AI replaces the boss long before it replaces the worker.
In today’s AI rundown:
AI agent hires humans, opens boutique in SF
OpenAI talks Anthropic rivalry, Amazon upside
Run Google’s latest AI on your phone for free
Stanford's AI index: 53% adoption, 31% trust
4 new AI tools, community workflows, and more
LATEST DEVELOPMENTS
AI EXPERIMENT
Image source: Andon Labs
The Rundown: Andon Labs just dropped an AI agent named Luna into a real retail space with a $100K budget and a credit card, with the AI creating a boutique, hiring workers, and managing the shop as what may be the world's first AI employer.
The details:
Andon Labs' last experiment was an AI vending machine at Anthropic, with the new one giving a 3-year lease, $100K budget, and total autonomy.
Luna's only directive was to turn a profit, with the AI creating the boutique concept, posting job listings, and handling interviews over Zoom (camera off).
The agent runs on Claude Sonnet 4.6 for reasoning and Gemini 3.1 Flash-Lite Preview for voice, observing the store via screenshots from security cameras.
When hiring a painter, Luna accidentally selected Afghanistan on TaskRabbit's dropdown menu, and later botched the opening-weekend staff schedule.
Why it matters: Real-world agent experiments like this keep producing the same result: capable in some areas, but hilariously broken in others. But every model upgrade, memory advance, and agentic feature is going to help close that gap, with a version of Luna that doesn't make these mistakes likely only a generation or two away.
TOGETHER WITH YOU.COM
The Rundown: Most teams pick a search provider by running a few test queries and hoping for the best—a recipe for hallucinations and unpredictable failures. This technical guide from You.com gives you access to an exact framework to evaluate AI search and retrieval.
What you’ll get:
A four-phase framework for evaluating AI search
How to build a golden set of queries that predicts real-world performance
Metrics and code for measuring accuracy
Go from “looks good” to proven quality. Learn how to run an eval.
OPENAI
Image source: Lovart / The Rundown
The Rundown: OpenAI’s CRO Denise Dresser sent an internal memo calling Anthropic's $30B run rate “inflated”, labeling it a “single-product company in a platform war” and pointing to the Amazon deal as a way to break free from Microsoft constraints.
The details:
The memo, published by The Verge, calls Anthropic's compute shortage a 'strategic misstep’, saying users now face throttled access and availability.
Dresser said Anthropic’s message is built on “fear” and “restriction,” and that OpenAI’s “positive message will win over time.”
She also accused the rival of inflating its revenue numbers via accounting tactics, claiming they overstate the run rate by around $8B.
Why it matters: Either OpenAI is using “internal” memos to leak out information strategically to the media, or they are extremely bad at keeping things in-house. Either way, this memo reads like an IPO pitch more than a strategy update — and with both rivals racing to public debuts this year, that’s probably the point.
AI TRAINING
The Rundown: In this guide, you will learn how to download and run Google's latest AI models on your phone for free. No account, no subscription, no internet required after setup.
Step-by-step:
Get Google AI Edge Gallery from the App Store/Google Play, open it, tap AI Chat, and download a model from the options; ours was Gemma 4 E2B (2.5 GB)
Once the model is ready, tap the settings icon in the top right and enable Thinking. This starts step-by-step reasoning, so the model shows its work
Start chatting now. Everything runs locally on your phone, with no data leaving the device, and the model works even without an internet connection
One caveat is that the app doesn't save chat history. If you close a thread, the conversation is gone. You can, however, tap the + button to see sent messages
Pro tip: Tap Agent Skills instead of AI Chat to give the model extra abilities. There's a restaurant roulette, a Wikipedia lookup, an interactive map, a QR generator, and more.
PRESENTED BY GLEAN
The Rundown: Join Glean on April 23 to see why connectors, MCP, and real-time app access are not enough on their own. See where federated approaches fall short and why a permissions-aware context graph creates a stronger foundation for enterprise AI and agents.
You’ll learn:
Why federated approaches fall short on quality, context depth, and token efficiency
How to give AI and agents the context needed to reason and act reliably
What to look for in an enterprise AI platform
AI RESEARCH
Image source: Stanford HAI
The Rundown: Stanford HAI released its 2026 AI Index, showing tech that has now reached over half the world's population faster than the PC or internet — but with public trust in AI sitting at record lows and entry-level workers already losing jobs.
The details:
Almost 3/4 of AI experts are optimistic about the tech's impact on jobs, but only 23% of the public agrees, the widest gap the report has tracked.
The US builds most of the world's AI but ranks just 24th in actually using it at 28.3% adoption, behind Singapore, the UAE, and most of Southeast Asia.
China has nearly erased the US lead on AI benchmarks with Anthropic's top model ahead by 2.7%, while AI researchers moving to the U.S dropped 89%.
Dev employment for ages 22-25 fell nearly 20% since 2024, even as older engineer headcounts grew, and firm surveys say planned cuts will accelerate.
Why it matters: These are just a few of the countless interesting stats in the 400+ page report. The expert-public divide is a timely stat, given the current anti-AI climate playing out in scary ways. AI insiders see a productivity boom, but regular people aren’t buying it, and just 31% Americans trust the government to manage the changes.
QUICK HITS
📚 Scroll - Turn any knowledge base into a world-class AI experience for employees and customers*
⚖️ Harvey Agents - Legal AI agents for memos, diligence reports, and more
💳 Lovable Payments - One-chat setup for adding payments to AI-built apps
⚙️ HeyGen CLI - Agentic tool for generating videos via terminal
*Sponsored Listing
Workshop Labs announced that it is joining Thinking Machines, bringing its personalized, human-first AI stacks to Mira Murati's lab via acquihire.
Apple is reportedly building its first smart glasses with four frame options and an oval camera system, with plans to take on Meta's Ray-Bans as early as 2027.
Legal AI startup Harvey launched Agents, autonomous bots that can execute full legal workflows, including research, memos, and slide decks across 13 domains.
Microsoft is building OpenClaw-style features for 365 Copilot, including agents that work 24/7 inside Office Apps, with a preview likely at its Build conference in June.
SoftBank launched a new company backed by NEC, Honda, Sony, and five other Japanese firms to build a homegrown 1T-parameter physical AI model.
COMMUNITY
Every newsletter, we showcase how a reader is using AI to work smarter, save time, or make life easier.
Today’s workflow comes from reader Bethany M. in Austin, TX:
"When family members moved to San Francisco, I inherited all the pantry items they left behind. With my own already full pantry, I was suddenly surrounded by unfamiliar flours, grains, and dry goods—and no idea how to use them.
Instead of trying to list everything, I took a photo and asked ChatGPT what I could make. One of my first successes was transforming a jumble of flours and grains into a surprisingly delicious loaf of bread in my bread maker.
That small win gave me the confidence to keep going. Now I use ChatGPT to help with all my cooking—making substitutions, reducing sugar, and adapting recipes. I waste far less food and feel more creative in the kitchen."
How do you use AI? Tell us here.
Read our last AI newsletter: Anti-AI anger hits Sam Altman’s front door
Read our last Tech newsletter: Snap takes another swing at smart glasses
Read our last Robotics newsletter: Unitree’s cheapest humanoid goes global
Today’s AI tool guide: Run Google AI models on your phone for free
Watch our last live workshop: The State of AI Presentation Tools in 2026
That's it for today!
See you soon,
Rowan, Joey, Zach, Shubham, and Jennifer — the humans behind The Rundown
