Code craft
Would you be happy to inherit this?
Correctness, readability, tests, and whether the next person can change it safely.
Bonfire · the Record
A portfolio that says what you can't do yet.
Which is exactly why you can trust the rest of it.
Every builder leaves with a Record: six dimensions, rated and signed, each linked to real work.
The gap
A GitHub is not a portfolio.
It shows code someone wrote, at unknown difficulty, under unknown constraints, with unknown help, reviewed by nobody in particular.
What it cannot tell you: whether they understood the problem or just the ticket. Whether they argued for the simpler thing. How they explained a trade-off to someone non-technical. Whether they caught the agent's confident, plausible, wrong answer. Whether anyone competent ever looked at it.
What's measured
Six dimensions.
One of them is the software. The other five decide whether the software was worth writing, and they are the ones that got scarce.
Product understanding
Do they build the right thing, or just the thing asked for?
Whether they speak up when the request and the real problem have come apart.
Deep communication
Does the team think more clearly because they are in it?
Explaining a trade-off to a non-engineer. Disagreeing well. Leaving a trail others can follow.
Systems thinking
Do they see what this touches?
Failure modes, what else breaks, and the thing that goes wrong in production at 3am.
Architectural design
Will this still look like a good call in a year?
Choosing boundaries, resisting cleverness, knowing which decisions are hard to undo.
Agent orchestration
Do they catch the confident, plausible, wrong answer?
Directing AI, reviewing what comes back, and knowing when to throw it away.
The scale
We rate autonomy, not talent.
Scores out of ten travel badly between companies. Every interview is really asking one thing: how much of my time will this person cost, and for how long? We answer it with evidence.
Needs direction
Does good work once the problem is framed and the approach is agreed. Learning quickly, and worth the time it takes.
Works with guidance
Frames the smaller problems alone. Knows which decisions to bring to someone else, and brings them early rather than late.
Works independently
Owns an outcome end to end. Escalates the right things at the right time, and very little else.
Sets direction
Raises the standard around them. Makes the calls the rest of the pod follows, and is usually right about them.
Why this scale
Nobody is uniformly senior.
Most people are two different levels on two different dimensions, and that shape matters more than any average. A builder who sets direction on communication and needs guidance on architecture is an excellent hire for one team and the wrong hire for another. A single seniority label hides that. The Record shows it.
A worked example
What one actually looks like.
Built the way a real one is, including the parts that do not flatter.
A. Rahman
Bonfire builder · 3 engagements · 11 months
Code craft
Payments migration, 40+ reviewed pull requests. Test coverage
on her module up from 31% to 78%.
Works independently
Product understanding
Pushed back on a requested feature in week two, proposed a
smaller one, shipped it. The original was quietly dropped.
Works independently
Deep communication
Wrote the decision record the client's own team now uses as a
template. Ran the weekly demo from month two onward.
Sets direction
Systems thinking
Strong on the service she owned. Still learning to predict
knock-on effects across service boundaries.
Works with guidance
Architectural design
Good instincts on module boundaries. Has not yet led a design
where reversibility was the main question.
Works with guidance
Agent orchestration
Caught a plausible but wrong data-migration approach from an
agent before it reached review, and documented why it was
wrong.
Works independently
Reviewer's recommendation
Ready for a mid-level product engineering role on a team that already has an established senior. Not yet ready to be the only engineer on a greenfield system, which is what the systems and architecture ratings above are describing. Three months on a service with real cross-boundary complexity would close most of that gap. Hire her for judgement and communication, and expect the architecture to catch up faster than you would assume.
Signed by the principal who reviewed the work, across all three engagements.
Illustrative sample. A real record names its reviewer and links every rating to the artefacts behind it.
Why it can be trusted
Three things that make it evidence.
Every rating links to the work
The pull request, the design doc, the demo. You can go and look.
A named person signs it
Not the studio in the abstract. The principal who was hands-on with the work, with their name on the assessment.
We publish the gaps
A record that only says good things is marketing. Ours say what someone cannot do yet, in writing. That is what makes the rest of it worth reading.
If we vouched for everyone equally, we would be vouching for nobody. The reputation only works if it can be spent, which means sometimes saying the less comfortable thing about someone we like.
light the fire
Want to see real ones?
If you are hiring, we will walk you through the Records of builders who fit. If you are building, we'll talk you through how it works.