TaskMeter AI:
One-Click Estimates.

01 // The Problem

Estimates
From Memory.

Every freelancer tracks time, and almost nobody uses that history when quoting the next job. Estimates still come from memory and optimism, and the tracker becomes a record of how wrong the last estimate was. I had the same problem with my own client work, which is why I started building this for myself first.

02 // What I Built

History That
Plans Ahead.

The first version is deliberately small: time tracking and clear reports, without invoicing. On top of that sits the assistant. It keeps the task list tidy by grouping and cleaning up entries, and when you describe a new task it finds the most similar work you have tracked and turns it into an estimate in one click.

03 // The Hard Part

Estimates You
Can Argue With.

A single number from an AI is easy to produce and impossible to trust. So the estimate is a range, and it shows the past tasks it was built from, with how long each one actually took. The language model's job is matching task descriptions that mean the same thing in different words. The arithmetic stays in code, so the same history always gives the same answer.

04 // The Product Decisions

Scope Cut
On Purpose.

Invoicing was the obvious next feature and I left it out. Getting invoicing right for Spanish freelancers means tax compliance rules that would have doubled the build before anyone used the core idea. The same thinking shapes what I advise clients: ship the part that proves the product is worth building, then decide what comes next from real use.