For software companies
Engineering craft that keeps pace with AI.
Your engineers ship more AI-written code every month. Are they getting sharper or sloppier while they do? We assess the judgment behind the code, practice it with the team, and apply it to one bounded problem at a time. A price agreed in writing, every time.
A short conversation with the engineer who does the work. You leave knowing which of the products fits your team, or that none does yet.
Picture a code review a year from now. AI wrote most of the code. Your engineers read it the way a senior reads a junior's pull request. They know what to keep, what to send back, and why. That's the team that got sharper. The other team ships faster and loses its edge, and review turns into skim-and-approve. The difference is judgment, and whether the team builds, measures and practices it on purpose.
Where our methods come from
I'm Attila Fejér, the founder of Craftosphere. Since 2013 I have coached more than 400 engineering teams. The largest of those programs, at a Fortune 100 financial company, measured its teams on DORA metrics and saw them improve. I'm part of the team that builds nWave, the open-source AI-assisted delivery framework, and we use it for our engagements. The standard for engineering judgment in the AI era doesn't exist yet. I'm building it, and every engagement, with the client's consent, feeds it.
How it works with an engineering team
From a birds-eye-view, every engagement runs the same way.
- 1
We talk first.
A short conversation about where your team is with AI-written code. You leave with a recommendation: which of the products fits, or none yet.
- 2
You get a written assessment.
The Scorecard is two to four weeks of interviews, codebase and review analysis. You get the findings and your three highest-leverage moves. From €8,000.
- 3
You practice, or you apply.
The Judgment Dojo builds the habit weekly. A Precision Sprint fixes one bounded problem alongside your team.
- 4
You own the result.
Reports, playbooks, decision records and the before-and-after numbers are yours.
What to expect
Every product has a fixed scope, a fixed price, and a specified set of deliverables. No retainers, no open-ended advisory. We work with your code and your review process wherever the work needs them, not from a slide deck. Where AI is in the loop we say so, and we show the judgment that made the result trustworthy. We sign an NDA before we see anything, we work on a read-only copy of the code that's deleted at the end, and we assess the team, not individuals. If your team isn't ready for a product, we say so in the first conversation and it costs nothing.
The product portfolio
The first conversation tells you which one fits.
Judgment, assessed.
Engineering Quality Scorecard
A written assessment of your engineering teams' judgment across the four CODA dimensions. Two to four weeks. From €8,000.
DetailsJudgment, improved.
The Judgment Dojo
A weekly practice program for engineering teams, 8 to 16 weeks, on realistic codebases, ours or yours. From €8,000 per team.
DetailsJudgment, applied.
Precision Sprint
Four to eight weeks alongside your team on one specific, bounded challenge. Fixed scope, fixed price, concrete deliverable. From €20,000.
DetailsWhat you risk: very little
What a Fortune 100 program taught us
Three hundred teams moved from Scrum toward extreme programming, a few at a time: two weeks of classroom and six weeks embedded, with about 10% better lead time and change failure rate after six months. What stuck, what didn't, and what it means for AI-written code.
Read the case studyNothing taken on faith
You see the findings before the report is final, the exercises and the playbooks are yours, and the before-and-after numbers come from your own team. At every step you have evidence, not our word.
The engineer behind it
Building software since 2010, coaching engineering teams since 2013: 400+ of them. The standards we teach large engineering teams are the same standards we hold our own work to.
A price that can't run away
The price is agreed in writing before we start. A change in scope is agreed and priced before it happens, and nothing costs anything without your yes.
One stack. Two standards. One open question.
DORA
Measures delivery outcomes: deploy frequency, lead time, failed deployment recovery time, change failure rate.
SPACE
Measures developer productivity: satisfaction, performance, activity, communication, efficiency.
CODA
The Comprehension and Ownership Depth Assessment, in development: a hypothesis we test, that the engineering judgment underneath both is measurable: the judgment AI tools make urgent. Comprehension Ratio, in particular, has no counterpart in SPACE.
SPACE was designed before LLM coding tools became common. The question of whether engineers understand what they ship wasn't as important then. It is now. CODA is meant to add the judgment layer to both if the hypothesis holds, and every Scorecard cites SPACE and DORA.
Questions engineering leaders ask
The questions we hear most often.
Engineering teams: our products
The Engineering Quality Scorecard is Craftosphere's two- to four-week assessment of your engineering teams: interviews, codebase and review analysis, across the four dimensions of CODA, the Comprehension and Ownership Depth Assessment (Decision Velocity, Decision Confidence, Comprehension Ratio and Review Depth). It produces a written report with the findings and a prioritized leverage plan. From €8,000. More about the Scorecard.
The Judgment Dojo is a weekly practice program for engineering teams who want to build judgment on purpose, not through on-the-job osmosis. Teams of 6 to 12 engineers meet weekly for 8 to 16 weeks on one of four tracks: AI Code Judgment, Architectural Thinking, Refactoring Judgment, or TDD and Quality Practices. From €8,000 per team. More about the Dojo.
A Precision Sprint is a four- to eight-week engagement where a senior engineer from Craftosphere works alongside your team on one specific, bounded challenge, such as establishing an AI code review practice, shifting quality culture, or stabilizing architectural decisions. Fixed scope, fixed price, concrete deliverable. From €20,000. More about the Sprint.
With a short conversation. You tell us where your team is with AI-written code, we tell you which of the products fits, or that none does yet. It costs nothing and there is no follow-up sequence.
Engineering judgment training is built for senior and mid-level engineers (typically 5 to 15 years of experience) who use AI coding tools daily but want to maintain and build the decision-making skills AI can't replicate. It also serves team leads and staff engineers who need to set quality standards in teams where a growing percentage of code is AI-generated.
A Craftosphere Engineering Quality Scorecard takes two to four weeks depending on team size. It produces a written report across the four CODA dimensions and a prioritized leverage plan. A follow-up Scorecard measures the change after a Judgment Dojo or a Precision Sprint.
Engineering teams: CODA
CODA, the Comprehension and Ownership Depth Assessment, is the framework Craftosphere is building to measure engineering judgment. Its four dimensions form two tension pairs:
- Speed vs. Depth: Decision Velocity (how quickly the team reaches architectural decisions) and Decision Confidence (how well those decisions hold up over time).
- Output vs. Ownership: Comprehension Ratio (whether developers can explain and modify the code they ship) and Review Depth (how thoroughly AI-generated code is reviewed).
Gaming one dimension of a pair visibly degrades the other, the same design principle that makes DORA metrics effective. CODA is in development. Today it's the qualitative frame of every Scorecard.
DORA, SPACE, and CODA are at three different altitudes. DORA measures delivery outcomes. SPACE measures developer productivity. CODA measures the engineering judgment AI coding tools make urgent. CODA is built to work alongside DORA and SPACE, and both are cited in every Scorecard.
Engineering teams: your code and your people
Yes. We sign your NDA before we see anything, or ours if you prefer. Who on our side sees the code is written into it.
Read-only. We work on a clone, because an assessment and the AI-assisted parts of it need the whole codebase at hand, and we delete it at the end of the engagement. We never push, and we never need write access.
During an assessment or a sprint, yes, under Craftosphere's own accounts and only with your written consent for that engagement, naming the tools. We tell you which model touched what. If you say no, we work without them, but it takes longer.
No. The Scorecard assesses the teams' practices, not people. Nobody gets a personal score and no individual data goes to management. Engineers are interviewed to understand how the teams work, not to be measured.
Our notes and the findings, so a follow-up Scorecard can measure the change. Tell us and we delete them instead. If you agree in writing, we also keep anonymized data for the research behind CODA, with nothing that identifies your company or your people. Without your consent, nothing is kept for research.
Bring your codebase. Leave with a recommendation.
A short conversation about where your team is with AI-written code. You leave with a recommendation, even if it's "not yet".
No newsletter, no follow-up sequence.