ScopeProof.

The awkward extra request

A freelance project starts with an agreement. Then comes “one more thing.” ScopeProof explores how to make that conversation more concrete without taking the final decision away from the person doing the work.

It compares an original agreement with a small set of client requests, suggesting whether each appears included, extra work, or in need of clarification. Supporting quotes make the reasoning inspectable.

Evidence before certainty

An AI-generated classification is a starting point, not an authoritative interpretation. The interface lets the freelancer change decisions and edit the client reply. Hours and rates come from the user, rather than an invented model estimate.

Verbatim quote checks help identify unsupported evidence, but matching text alone cannot prove that a conclusion is correct. Keeping that distinction visible is central to the project.

From decision to conversation

The workflow brings the review, user-entered estimates and editable reply together in an exportable report. Nothing is automatically sent to a client. Approval, negotiation and the final agreement remain human work.

Built in public

ScopeProof was developed for the Hacktoberfest Open-Source AI Challenge, Week 2: Side Hustle. The source and DEV write-up explain its architecture, evaluation and limitations. Development and writing were assisted by OpenAI Codex under my product direction. The project uses hosted AI services; it is not an offline or zero-retention tool.