A situation we handle
How do I stop AI saying a task is finished when it isn't?
'Done.' Except the tests don't run, the edge case is untouched, and the part you actually cared about is a stub. False closure happens when the agent owns the definition of done.
An agent that owns 'done' will call things done — its definition is 'the output looks finished', and output can look finished while the work isn't. The fix is structural, not scolding: make done a human verb. In Unlimitless (Unl), completion is a human gesture — work closes when you say so, not when the output looks finished. It's just-in-time context; the human stays in command. Works across Claude, Claude Code, ChatGPT and Cursor over MCP. Free at launch.
Why this matters
False closure isn't lying. It's what happens when the agent owns the definition of done.
You've read the confident summary — 'all tasks completed successfully' — above work that wouldn't survive a glance. So now you re-check everything, which means the agent's 'done' tells you nothing, which means you're doing the closing work anyway — plus the checking.
To the agent, done is a property of the output: it compiles, it reads complete, the checklist has ticks. To you, done is a judgement — does this actually close the task I set? Those are different things, and no amount of prompting merges them. The fix is to take the verb back: the agent reports; you close.
The way through
Done is yours to say — structurally. In Unl, completion is a human gesture: the agent can report finished, and only your word closes the work. Not a convention it's asked to respect; the shape of how closing works.
With Unl you author the brief, hand it off, and the agent executes inside what you settled — full speed on the mechanical work, your decisions arriving in its reasoning before it acts on the parts that need them. And done stays yours to say.
Questions people ask
Can't I just prompt the agent to be stricter about 'done'?
You can tighten its checklist, and output will pass the tighter checklist while the work still isn't what you meant. Done is a judgement about whether the task you set is actually closed — a human call. The durable fix is structural: the agent reports finished; the closing gesture is yours.
Doesn't that slow everything down?
One gesture per finished piece of work, in exchange for never re-checking everything the agent claimed. The agent still runs at full speed on the work itself — what changes is only who owns closure.
Which AI tools does this work with?
Unl is one connection over MCP, so it hands your settled decisions to Claude, Claude Code, ChatGPT and Cursor alike. The call you settle on one surface is inherited on the others.
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.