Project · Creator and maintainer
FrameShift
An open reasoning system for finding the right problem before optimizing a solution.
Updated
Context
Engineering requests usually arrive as solutions in disguise. Select a 5 kW DC/DC converter, change the motor, add a battery: each one names a mechanism before anyone has agreed on the outcome it is supposed to produce.
Constraint
Accepting the framing is the expensive mistake, and it is invisible at the time. Work then optimizes a solution to a problem nobody stated, and the reasoning that led there is not recorded anywhere it can be challenged later.
Contribution
I am building FrameShift, an interactive system that separates observation, need, requirement, constraint, assumption, and proposed solution, reframes the request across abstraction levels from component to business, builds causal hypotheses carrying evidence and counter-evidence, and records why a decision was made.
Approach
The project is specification-first. The repository holds the durable decisions, the domain language, and machine-checkable contracts, and says plainly that it does not yet contain a production application. Human checkpoints are part of the loop by design, so the tool structures judgment rather than laundering model output into authority, and a reasoning trace is portable across models and runtimes.
Outcome
The work so far is a specification and a set of contracts. What it is meant to produce is a decision record that survives the change of model, tool, or person that usually erases the reasoning behind a choice.
Evidence
The public repository holds the architectural decision records, the domain language, and the contracts; the backlog is tracked in its issues.