Problem-first, no scaffolding
The audience has software experience, so we do not burn classes installing things: that is pre-work, documented step by step.
Training service
Turn your team into AI-First builders.
Dilux AI Builders Lab is a hands-on program where each participant builds a real application — the same one, several different ways — until they feel first-hand the difference between vibecoding, specifying and orchestrating agents.
- 11 modules, live classes and work between them: the real practice does not happen in class.
- Teaches how to use Dilux Agentic Workflow in your own repositories, end to end.
- Ends with a final project in production and a Demo Day against a rubric.
11content modules
8live classes
20 hintensive
1real app in production
The idea
The same app, built several ways
Knowledge travels, code gets rebuilt. Each method starts from scratch in its own repository, from the same PRD, the same guardrails and the same skills. What gets compared is the method, not the starting point.
Pablo builds live
Classes are real demonstration, not slide reading. You see the mistake, the fix and the decision as they happen.
Incremental final project
Each module produces a piece of the deliverable. By Demo Day the app is already built in parts: no final panic sprint.
Peer review in trios
Pull-request review between peers, with a short checklist and AI-assisted review. It is the real practice and it is also course content.
Rubric from day 1
The final project spec and the Demo Day rubric are handed out in the first class. Every class has an explicit purpose.
Everything is recorded
Access to recordings and materials. Missing a class does not put anyone out of the program’s rhythm.
Syllabus
Eleven modules, from idea to production
High-level content. The detail of each lesson is tuned to your team’s stack and starting point.
AI-First Development
What genuinely changed in how software gets built, and how to write a PRD an agent can actually use.
Professional vibecoding
Directing the agent with judgment: context, iteration, and the limits of working without a safety net.
Build with a safety net
Git, your own skills and discipline. The same app, now versioned and with tools the participant built.
Spec Driven Development
Specifying before implementing, with Spec Kit. The project constitution and what a team gains by writing one.
Agentic Orchestration
Dilux Agentic Workflow installed in the participant’s repo, customized, and a full first feature with Claude. Assessed module.
DAW with OpenCode
The same method ported to another tool. You see what belongs to the framework and what belongs to the agent.
DAW with Copilot
A third port. By now the difference between method and tool no longer needs explaining: it shows.
GitHub Workflows and code review
Continuous integration, pull requests and AI-assisted code review — the flagship skill of 2026.
AI-First QA
Testing and evals: how to test what an agent built and how to measure whether it improves or degrades.
Monitoring and adoption
Deployment, observability and the hardest part: getting the whole team to adopt the way of working.
Practices and trends
Bonus module: what is emerging and has not settled yet, with the judgment to separate signal from noise.
Demo Day
Each participant presents their application in production and defends it against the rubric handed out on day one.
The syllabus adapts: if your team already specifies well, M4 gets shorter and orchestration and evals get deeper.
For your organization
Building AI-First teams, and full agentic ones too
The program does not end at “your people know how to use an agent”. It ends with teams able to build the software your company or professional practice needs, at the level of human intervention each kind of work demands.
- AI-First teamsPeople who direct agents with judgment and method, and know when to do it by hand.
- Full agentic teamsFlows that run end to end with supervision at the gates, not on every line.
- In your contextPrivate cohorts for your company, with your stack, your repositories and your real cases on the table.
- 100% remoteLive classes with recordings and materials included, so your team joins from wherever it is.
- Pairs with the consultingTraining and assessment combine: first you know where you stand, then you train what is actually missing.
Who is behind this
About us
We do not sell a theory: we ship to production the very thing we teach.
Training
Request a cohort for your team
Tell us how many people you are and where they start from. We will put together a proposal with the syllabus tuned to your context.
- Private cohorts for companies, or seats in open cohorts.
- Recordings, materials and final project included.
- Combines with the maturity assessment if you want to measure before and after.
Done — your request is registered.
You are on the private preview list. We will email you at the address you provided as soon as a seat opens.
Request number: —
Frequently asked questions
Who is it for?
People with software experience: developers, tech leads, product managers and adjacent roles. It is not an introduction to programming.
What is the format?
Eight live classes, twenty intensive hours, 100% remote. The work between classes is the real practice, and that is where the project gets built.
Do we need to pay for tools?
Dilux Agentic Workflow is free. Coding agents require a subscription or API keys from the provider you pick; the list and the alternatives are handed out in the pre-work.
Can it be combined with the consulting?
Yes, and it is what we recommend most: the assessment defines what your team needs, and the training covers exactly that instead of a generic program.