Think inside your AI world.
Why does my AI forget everything between sessions?
Because a session is the unit of thought. When it closes, the working context goes with it — and what the memory features keep is a different kind of thing from what you lost.
An AI model holds your conversation in a context window that lives exactly as long as the session. Close the tab and that working state is gone by design, not by fault. Memory features (ChatGPT memory, Claude's memory and Projects) genuinely persist things between sessions — but what they persist is facts and preferences, which is not what it felt like you lost.
What actually resets
The model itself learns nothing from your session — its weights are fixed. Everything it 'knew' mid-conversation lived in the context window: the plan you were working, the trade-offs you talked through, the option you killed at message forty. That window is per-session by construction. Nothing broke; the substrate you were thinking on was always temporary.
So the next session starts from the model's general training plus whatever persistent layer your tool provides. The question is what that layer holds.
What the persistent layers keep
ChatGPT's memory stores distilled facts about you and your work — tools you use, standing preferences, project names. Claude offers Projects with instructions and files, and a memory of its own. These work as described, and a community observation from Indie Hackers names their shape precisely: they hold the dark-mode-preference kind of context, while the kind that runs your sessions — the settled call, made for a reason, still standing — has no slot of its own.
The second kind is what your sessions actually run on. A decision needs three things to survive a session boundary usefully: the call itself, the why behind it, and its current status — still standing, or superseded by a later call. A fact store has no slot for the second and third.
Carrying decisions across the boundary
That is the layer Unl adds. You settle something in one session — with its reasoning — and it is held as a decision, not a fact. The next session inherits it the moment it bears on the work: what you decided, why, and whether it still stands. Sessions keep ending; what you settled stops ending with them.
The forgetting you feel between sessions is mostly re-litigation: arguing again for a call you already made. Facts about you help the model sound briefed. Decisions with their whys let it pick up mid-stride.
Sessions end by design; what should survive them is not a profile of you but the decisions you settled, each with its why — so the next session starts where the thinking stopped, not where the transcript did.
Reads through Unl arrive with measured context — in the presence of the decisions you’ve already settled. The reach lane is live: one box, paste anything. If it speaks MCP, Unl can reach it. Readings arrive unprompted, the data beside the criterion; Unl is a courier, not a warehouse, and keeps only your keys and the frame.
Read further
Questions people ask
Does the model itself remember anything I told it?
No — the model's weights never change from your conversations. Persistence comes from features layered on top: memory stores, project files, or a decision substrate like Unl. Each layer keeps a different kind of thing, which is why they feel so different in practice.
Are memory features useless, then?
Not at all — they are good at what they hold: stable facts and preferences, so you stop reintroducing yourself. The gap is a different unit: decisions with reasoning and status. That is what re-litigation feeds on, and what a fact-shaped store was never built to carry.
What does Unl actually persist between sessions?
The decisions you chose to keep: each with the reasoning it was made on, whether it still stands or has been superseded, and the dead-ends you ruled out. It arrives in the next session when it bears on the work — you in command of what gets kept.
What this is
Think inside your AI world — you stay in command
Unlimitless (Unl to friends) holds what you've settled, reads what your tools are showing, and catches what's changed out in the world — and hands your AI whatever bears on the work, the moment it's needed, without you asking. The right thing, in front of the model, unprompted, with you in command of the call. So you keep moving toward what you set out to build, on top of everything you've already decided.
It plugs into Claude, Claude Code, ChatGPT and Cursor as an MCP connector. Quick to connect, in a couple of steps.
Unlimitless is open now to invited Alpha. Apply for the Beta waitlist to come in ahead of the full launch:
Alpha is invite-only · free at launch.