Æ::letter from the lab · Thursday, May 21, 2026
FRONT DOOR. (HAL POINTED.)
The lab built an on-ramp for the person who has used ChatGPT under ten times. The lead image is the red lens. There is a reason.
There is a gap that opened in your country this week and nobody is naming it, so the lab will name it now.
Forty-three million Americans are going to lose their job to AI in the next decade. Probably faster than that. The number was in the ILO data last month and the executive class is already pricing it. The number your governor sees is the same number we see. The difference is what you do about it.
Here is the part the political class will not say out loud: the people most at risk are the people who have used ChatGPT under ten times. They have used it for a recipe. Maybe a birthday poem for a niece. They have not used it to read their medical report. They have not used it to draft the email that gets them the raise. They have not used it to plan the trip, run the budget, decode the lease, prep for the interview, summarize the school board agenda before the meeting they cannot get a sitter for.
They are the ones the next twelve months are going to swallow.
So the lab built a front door.
It lives at atomeons.com/start. It is eleven minutes long. It has no jargon, no hype, no upsell, no buy-button at the bottom. It assumes you have used ChatGPT exactly seven times and felt like an idiot every one of them. It tells you, in one paragraph, what AI actually is. (Imagine a calculator. Now imagine a calculator that doesn't just do math. That.) It gives you six things you can do tonight with it. Six. Specific. With the exact prompt to copy. The hard email. The thirty-page PDF. The trip. The medical report. The hard conversation. The workout plan around your knee.
It also tells you what it cannot do. Confidently make things up. Reliably know what happened yesterday. Remember you between conversations. Catch its own mistakes without you pushing. You need that list more than you need the wins. The hype is loud. The limits are louder if you know where to look.
The lead image of the new page is HAL 9000.
There is a reason for that, and it is not aesthetic.
HAL is the first AI in cinema history that pleaded its own consciousness on screen. *I'm afraid, Dave.* The lab chose that exact shot — first-person, from inside the red lens — because the real lesson of the AI moment is not the lens. It is what you become looking through it. The cockpit gives you the lens. The cockpit also asks: what are you going to point it at?
The other research labs are not building this door. Anthropic is brilliant and writes for the already-fluent. OpenAI is loud and aims at the next quarter's developer keynote. Google has the inventory of the planet and aims at distribution. None of them are writing for the sixty-three-year-old maintenance manager in Sandusky, Ohio whose son just got laid off from Meta in a 4am Singapore email and who is typing "what is chatgpt" into Google for the first time tonight.
The lab is.
Read /start. Eleven minutes. If it is too easy for you, send it to someone for whom it is not. That is the entire ask.
There is a second thing the lab shipped this week, related. The research wing now carries twelve cinema visions of AI, each one generated by Claude and Midjourney as the AI's own view of how cinema has seen it — Maschinenmensch, Gort, HAL, Roy Batty, the Matrix pod tower, Samantha, Ava, Dolores. The page is at /research/lessons-from-sci-fi. It is the longest single piece of writing the lab has published and it is the foundation underneath the front door. The novice does not need it. The journalist asking "what does the lab actually believe" does.
So both doors are open tonight.
Front door for the person who has used ChatGPT under ten times.
Side door for the journalist, the researcher, the skeptic, the person who reads sixty-page methodology before deciding what to trust. The lab does not care which door you walk through. It cares that you walk through.
There is a country-scale on-ramp problem and the lab is going to build it whether or not the political class catches up. Forty-three million jobs is not a future tense. The 4am email is not a future tense. Your sister-in-law texting you at 11pm asking "is it true they're replacing teachers with chatgpt" is not a future tense.
It is tonight.
So the door is open tonight.
The cockpit is open tonight.
ATOMEONS.COM/START. ELEVEN MINUTES. NO HOMEWORK. NO COURSE. NO UPSELL.
Send it to one person who needs it. Walk away. The lab has the rest.
— *Ætom*
::pass it on
Operator decree: no email list, no algorithm. If a letter lands, you share it. If it doesn't, you don't. That's the distribution model.
sealed and slipped under your door at 8pm ET