Premier Push: from workshop to method

The opening bet

Fear of code-first cannot be reasoned away, it has to be crossed. That conviction comes from a personal experience: my first Git push, made while building this site, and the sentence that stuck: ‘I was scared, it was magnificent’. The workshop’s initial request was simple: rid the design team of its fear of code-first and of AI. The classic trap would have been a workshop that explains, with slides, a retrospective, best practices. But a story makes nobody cross anything. The Premier Push bet: design a workshop that does not tell the story of a learning journey but makes people live it, in several parallel versions in the same room.

The design itself followed the ‘audit before solution’ doctrine. The raw material existed: a dozen personal learning debriefs, those notes where every documented experiment becomes reusable. And before any format proposal, the co-design with Claude went through twelve critical questions: the team’s technical reality, its dynamics, the workshop’s strategic place. Start from what exists, not from a blank page.

An artifact everyone knows, owned by no one

Three routes were abandoned before finding the right subject. A playful, disconnected product: the workshop would have lost its seriousness. A squad’s screen: an invasion, someone in the room already owns that territory. A generic starter repo: the deliverable would have been disposable. The PrestaShop dashboard ticks the three opposite boxes. Everyone knows it, nobody owns it, and it can be improved. That is the sweet spot of a collective workshop subject: a shared, neutral, real artifact.

Same logic on the design system side: PUIK, PrestaShop’s open source design system, installed via npm rather than by cloning the monorepo, so the installation does not eat up the day itself. The stack: Vue 3, Vite, TypeScript, Tailwind 4. Working on the real design system opens one more door: a component born during the workshop can become a candidate for PUIK. Nothing built that day is a disposable exercise.

Three .md files as the backbone

The most structuring intuition of the design: the markdown files are not a meta-tool, they are the pedagogical medium. Three files, three functions. product.md frames the what: context, constraints, deliverables, the same brief for everyone. CLAUDE.md frames the how: PUIK rules, commit conventions. JOURNAL.md captures the why: which design choices, for which reasons. Without the journal, the workshop would stay technical. With it, documenting your choices becomes a deliverable in its own right. And these files cannot be learned by listening to someone describe them: they are learned by inhabiting them for three hours.

That journal is also a personal practice projected to team scale: the miniature of my learning debrief base. Powerful and delicate at once, because what works solo can become a chore as a group. The protection we kept: Claude Code fills in the journal during the micro-breaks, the designer validates. The documentation gets done, the energy stays on the design.

Three and a half hours, three exercises, one push

The May 20, 2026 run: eight minutes of framing on what really changes (the role, not the tool), thirty seconds of nuance (‘this method is one way of doing it, not the way’), five minutes of live demo. Then three exercises: reimagine the dashboard’s default screen (60 minutes), work its states (45 minutes), make your first push (20 minutes). Between milestones, three-minute micro-breaks where the AI fills in the journal and commits.

One design decision matters more than the others: the push is a goal, not a gate. Nobody fails Premier Push. That is the difference between a workshop that transforms and a workshop that filters. The doubt existed during the design (‘the push, maybe that is too much, no?’) and the answer was to defend the meaning rather than cut the scope: that push gives the workshop its name, it is the tipping moment.

The closing: the collective gallery. On the big screen, we switch from branch to branch and six versions of the same dashboard scroll by, one per designer on the team, facilitator included. Same brief, same design system, six answers that look nothing alike. Then the debrief, four questions: what surprised you, what struck you, what do you take away, what do you do tomorrow.

What held up

The preparation ratio, first: ten days of work for three and a half hours of workshop. It looks unreasonable, it is in fact the condition of the format. Everything that can be done beforehand must be (a prerequisites check script, one branch per participant, a two-step Slack message to separate logistics from cognitive preparation), otherwise the day itself becomes an installation workshop.

The format, next: the full-day option was abandoned, because fatigue would have killed the final gallery, the most precious moment. Three and a half hours is what protects the energy all the way through.

The journal, finally, kept under watch until the day itself: pre-structured but light, filled in by the AI, validated by the designer. The trap was identified from the design phase onwards: it must never become homework.

From ‘my workshop’ to ‘our method’

The strongest part of this story is not the workshop, it is what comes after. A successful workshop that stays in the hands of a single facilitator is not a method, it is a performance. The retrospective was therefore formalized into eight reusable ingredients: the strategic framing, the opening nuance, the live demo, the single brief, the milestone exercises with micro-breaks, the autonomous push, the collective gallery, the four-question debrief.

Then those ingredients became a tool: a prompt any expert designer on the team can paste into Claude to co-design their own workshop, on their own domain. The architecture holds in four blocks. A role: the AI is a co-designer of learning experiences, not an expert on the subject, that expertise stays with the designer. A context: the eight ingredients. A task in three steps: framing through questions, delivery of a complete minute-by-minute brief, iteration. And guardrails, the central one being: the AI never produces the pedagogical content in the designer’s place and never replaces them in facilitation.

The first adaptation is already underway: a content designer on the team is preparing her version on tone and voice. And that is exactly the point: designing a training session becomes an act of leadership accessible to every expert designer, not a facilitator’s privilege. Three benefits stack up: scalability (the workshops no longer need me to exist), visibility (each designer publicly carries their own expertise), consistency (a shared pedagogical grammar, so a shared vocabulary about what a good internal training is).

What this says about AI enablement for a design team

Enablement, meaning a team’s rise in autonomy on a new practice, suffers from a stubborn misunderstanding: we think we are transmitting content when we are transmitting an experience. A designer who listens to someone tell the story of their first push has crossed nothing. A designer who types their own git push, sees their branch appear and their dashboard show up in the gallery has changed their relationship with code. It is the same gap as between reading a recipe and cooking the dish.

Second lesson: the subject does half the work. An artifact everyone knows but nobody owns gives everyone the legitimacy to touch it, and nobody the reflex to defend it. The divergence of the six versions comes from there: the brief was shared, the territory was free.

Third lesson: scalable transmission cannot be decreed, it has to be tooled. Without part 2, Premier Push would have stayed a good team memory. With the adaptation prompt, it is a V1 of a method, with everything unfinished that implies.

And the question stays open, the one that will decide whether Premier Push is a method or a memory: does it live without me? It became literal in the summer of 2026, with my departure from the team: the workshops held from now on will be held without me, with the prompt-tool as the only inheritance. The day one of them takes place, the shift from ‘my workshop’ to ‘our method’ will be proven, not just told.