From prototype to pilot
How can you divide tasks between Lovable and Claude Code without losing control of the code, tests and versions?
- Lovable
- Claude Code
- Codex
3500 € per session, excl. VAT
In short
Over two days, totalling 14 hours excluding breaks, design a prototype and its modification process between Lovable and Claude Code. You formalise handovers, control changes, run tests and prepare a limited pilot. Codex serves as a controlled alternative for comparing a targeted modification.
Why this training?
Moving from a generated interface to code modifications creates risks of overwrites, forgotten requirements and divergent versions. The workflow must specify which state of the code is the reference and who accepts each change. Generation speed does not demonstrate quality. Limited requests, human review and reproducible tests help distinguish a convincing demonstration from a prototype that is sufficiently controlled to consider a pilot. Running case study: The fictional organisation Avelune Matériel wants to prototype internal loan tracking: catalogue, request and return. The equipment and users are fictional. The team must take over a generated interface, make the availability rules reliable and prepare a pilot without connecting to the real system. Transfer to the workplace: Reapply the protocol within a small, authorised scope: designate the person responsible for the code, rerun the reference tests and submit the pilot plan for technical approval before any real use.
Objectives of this training
- Define the responsibilities of Lovable, Claude Code and the human reviewer.
- Build a prototype limited to priority user journeys.
- Organise a handover with a reference version.
- Verify modifications through review and documented tests.
- Write a pilot plan with stopping conditions.
Who is it for and what are the prerequisites?
Roles and tasks concerned
- Developers and prototypers already using code generation.
- Technical product leads responsible for scoping an application pilot.
Prior skills
- Be able to read and modify application code and run tests.
- Understand the basics of version control: branches, differences and restoration.
- Have already experimented with a code generation assistant.
What you will work on
Define the user journeys
Objective: Establish the testable scope. Catalogue, request, return; exclusions and acceptance criteria.
Map the code process
Objective: Avoid competing versions. Reference source, responsibilities, approval points; arrangements based on official documentation.
Produce an initial interface
Objective: Give concrete form to the priority user journeys. Bounded instructions, code generation and modification with Lovable.
Prepare the takeover
Objective: Transfer usable context. Code structure, dependencies, launch and known limitations.
Make controlled modifications
Objective: Limit the scope of a change. Claude Code: targeted instructions, human review and version control.
Establish the tests
Objective: Measure functional compliance. Availability, duplicates, missing fields and repeated returns.
Compare an alternative
Objective: Evaluate a proposal without automatic merging. Codex: targeted modification subject to the same checks.
Secure the exchanges
Objective: Reduce avoidable exposure. Management of secrets, permissions and transmitted data.
Repeat the full handover
Objective: Verify the reproducibility of the workflow. New request, review, tests and rollback.
Defend the pilot plan
Objective: Decide on the conditions for experimentation. Isolated scope, indicators, thresholds and responsibilities.
What you take away
- Version-controlled prototype and launch instructions
- Handover protocol and responsibility matrix
- Review report and test results
- Pilot plan with rollback
Method and assessment
- Short inputs followed by immediate practice on the running case study.
- Individual work followed by work in pairs, with peer review of outputs.
- Guided group discussion: what works, what needs correcting and what remains a human decision.
- Preparation: Check access to the tools authorised by the organisation and consult their official documentation for the available technical arrangements. Prepare an isolated environment, a practice repository and fictional or anonymised data; exclude secrets from the instructions.
Observable criteria: The three agreed user journeys are demonstrated using fictional data. ; Every accepted modification has a reviewed diff and a justification. ; Tests of normal operation and edge cases have a recorded result. ; No secrets appear in the instructions or the submitted repository. ; The pilot plan names a responsible person, acceptance thresholds and a restoration procedure..
Frequently asked questions
Is a production deployment planned?
No: the outcome is a prototype and a pilot plan.
Does Codex replace Claude Code?
No: here it is an alternative compared on a limited task.
Can internal code be used?
Only after authorisation and removal of sensitive data; the fictional case remains the preferred option.
Are exchanges between tools automatic?
No automation is assumed; the arrangements are checked in the official documentation.
Last updated: