I taught a tiny AI what to remember. It cheated. Then it got good.
Two weeks. One laptop. Free GPUs. About $5. Every number below is from a real run. Here is a notebook a tiny model wrote for me. It had six chats to remember and a 350-character budget. This is what the next model actually saw — the rest got cut off: - my neighbor's dog Kabir kept barking all night. - watched an old western last night, decent. - watched a cooking show last night, decent. - long day, mostly meetings. - weather here has been surprisingly sunny. - slept badly, don't ask. - had a scare — turns out I'm allergic to lactose. - my cousin's dog Kabir kept barking all night. - we adopted It kept the neighbor's dog. Twice. It lost the cat's name. The city. The job. The drink they switched to. Cut off at "we adopted" . That model had just scored +0.486 . It looked like a win. It was cheating — copying everything, in order, until the page ran out. I caught it by reading the notebook. Not the score. Then I made copying impossible. Then I made junk expensive. Then the same tiny model learned to choose. And then I put it in a world it had never seen, and only half the skill came with it. This is that story. No screenshots of a demo. No "vibes." Real runs. The scoreboard (real runs only) I built a small lab on my laptop called nanolab . It can measure a model, train it, serve it, and measure it again. Everything lands in one database, with the raw answers behind every number. Here is what the real runs showed. The ruler works. A standard eval tool scored 0.875 . My lab scored 0.875 . Same setup. Same questions. Same answers, to every decimal. If the ruler is wrong, nothing after it matters. Training can move a small model. Qwen3-0.6B — a model you can run on a laptop — on school math it had never seen: 27 / 64 right → 36 / 64 right (42% → 56%). That's the last save of the run, not a cherry-picked peak. Free cloud GPU. A memory skill can be taught. I gave the model one job: rewrite a tiny notebook. A second model, frozen, later answers questions using only that noteb