Service · 01 01 of 10

IDEA to Solution

Define the question before the build. We sketch the solution shape with you, then build the bespoke version — no vendor templates, no scope creep.


Most “idea” engagements arrive already half-translated into a solution someone wants to sell. The valuable conversation is the one before that — where the question is still unclear and the path is still up for grabs. We do the unclear part with you, then build what comes out of it.

What this looks like in practice

We start with a structured discovery: a few conversations to surface the actual decision the work needs to support, the constraints that aren’t obvious, and the parts that aren’t problems despite looking like they should be. By the end of week one we have a sketch of what to build and what NOT to build. Week two onwards is the build.

For a similar shape of work delivered as a public artifact, see the AI self-replication reproduction — a research-paper claim translated into a working harness anyone can run. Different domain, same arc: read the question, sketch the shape, build the thing.

Who this fits (and who it doesn’t)

This works when you have a real problem and an open mind about its shape. It doesn’t work when the brief has already been written, or when “the answer is obvious; we just need someone to execute.” If you already have full clarity, you don’t need this engagement — you need a contractor.

The first conversation

Free, 30–45 minutes, by phone or video. We’re listening for: what would change if this worked, what’s been tried, who else is affected. If we’re a fit, we propose a discovery cycle (one to two weeks) priced fixed, with a clear deliverable.

From the Field Journal

  1. Research

    Source-Backed Memory for AI Coding Agents

    JEV Context Fabric keeps exact, hashed source text as coding-agent memory and marks retrieved excerpts as untrusted and possibly stale.

  2. Research

    Building a Research-Driven, Evidence-Based AI Application

    Building Typewright, an alpha compiler for typed AI decisions, with research-cited design, hashed evidence, and an honest negligible live gain.

  3. Research

    Finding Where AI Decisions Actually Improve Existing Software

    An offline evaluator that finds judgment calls in code, gates every change behind approval, and runs studies allowed to say no. Five of six did.

  4. Research

    Mapping a Lean 4 Computer Vision Library

    How I documented a small Lean 4 colour and sensor library with twelve architecture views, including a review view of its limits.