Think inside your AI world.
What is the context tax?
The time you spend telling an AI what it needed to already know — again. It compounds quietly, per session, and one half of it is structural while the other half is entirely removable.
The context tax is the recurring cost of re-establishing context at the start of every AI session: who you are, what the project is, what stands, what was tried. Community estimates put it anywhere from 'embarrassing' (one developer who actually tracked theirs) to a working figure of 48 days a year — a number worth treating as one practitioner's tracked estimate, not a study. The shape of the cost matters more than its exact size.
Where the minutes actually go
Watch a session start and the tax itemises: restating the goal, re-attaching the files, summarising last week's thread, correcting the model's defaults back to your choices, and — the expensive line — re-arguing decisions that were already made. The first items are re-explaining: facts the session needs. The last is re-litigation: judgement the session should have inherited.
Re-explaining scales with project complexity. Re-litigation scales with how much you have already decided — which means it grows precisely as you make progress. That is why the tax feels heavier on month three than on day one.
What reduces which half
Projects, CLAUDE.md files and memory features genuinely cut the re-explaining half: stable facts, conventions and skeleton context can be loaded rather than typed. The re-litigation half resists them, because what it needs is not facts but standing: the call, the reasoning, and whether it still holds. A model handed the facts will still happily reopen the decision — politely, plausibly, and at your expense.
Removing the re-litigation half
Unl is built at that half. Decisions you settle are kept with their whys and served into future sessions when relevant, so the model treats them as ground already walked: it builds on the call instead of auditioning alternatives to it. The re-explaining half you can file down with good hygiene; the re-litigation half you remove by giving judgement a place to stand.
A fair test for any fix: start a fresh session and push back on your own past decision. If the AI defends it with your reasoning, the tax fell. If it folds and reopens the question, you are still paying.
The context tax splits in two — re-explaining facts, which good context hygiene reduces, and re-litigating settled decisions, which only ends when the decision arrives carrying its why.
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
Is the 48-days-a-year figure reliable?
Treat it as a practitioner's tracked estimate that circulated because it rang true, not as a measured study. The honest claim underneath it is directional: session-start re-briefing is frequent, unbudgeted, and grows with project maturity. Track one week of your own sessions and you will have a better number than any citation.
Doesn't a good CLAUDE.md or Project remove this already?
It removes a real slice — the stable, fact-shaped slice. What it leaves is the decision-shaped slice: files state conclusions, not reasoning or current status, so revised calls go stale silently and the model re-argues what was already closed. That remainder is the half Unl exists for.
Does the tax really grow as the project matures?
Yes, and that is the counter-intuitive part: more progress means more settled decisions available to be accidentally reopened. Day-one projects pay mostly re-explaining; month-three projects pay mostly re-litigation. The fix has to scale with your decision count, which a hand-groomed file does not.
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.