What I understood
about Studio.

Studio isn't a dashboard. It's the cockpit where creators run their business, and every screen that lags, every number that looks wrong, erodes the trust that keeps them publishing on Dailymotion instead of elsewhere. From the outside, three battles stand out:

If I've misread any of this, it's the first thing I want to learn in our conversation.

I understand the whole chain, from the creator's click to the chart that renders it.

Ten years, compressed into a timeline you can scrub:

• “He's never worked in video.”True. I've worked in energy, another domain where real-world data streams in 24/7 and non-technical users need it legible. Domain knowledge takes weeks; the instincts for data-heavy product take years. I'm bringing the years.✓
• “No GraphQL in production.”Correct, and I won't pretend otherwise. My data layers were REST. But the hard parts (fetching, caching, normalizing client state, typing the payloads) are my daily work. The query language is a bridge, not a wall, and crossing it is written into week one of my 90-day plan below.→
• “Rennes-based; we're in Paris.”Hybrid-ready from day one. Full remote, two days a week in Paris, or two days every other week: whichever rhythm the team runs on works for me. I've spent two years working in a distributed team: async-first habits, written culture, updates shared before you ask.✓
• “Another candidate claiming to use AI.”This site is my answer. Built with AI in the loop, reviewed line by line: architecture audit, TDD write-up, and a 37-test suite in the repo. I use AI the way your posting asks: fast, and with known limits. The /guide page shows the receipts.✓

Every codebase rewrites the plan that meets it. But here's how I approach the first 90 days, the same way I'd approach a feature: understand, ship small, then compound.

Think this plan is wrong? Good: challenge it with me in 30 minutes.

I haven't seen your codebase, so these aren't recommendations. Each one fits in a sentence, can be tested in under a month, and produces a number. That's how I like to work.

1. Where the analytics views spend their render time, and whether memoization or server-side aggregation is the cheaper win.
2. Whether a performance budget in CI (bundle size, route-level vitals) would catch regressions before creators do.
3. How error states reach creators, and if a triage of the top 10 frontend errors would move a support metric.
4. Whether long-range charts downsample before rendering, or render raw and pay for it. (I've fought this exact battle.)
5. What percentage of dashboard components live in a shared design system, and what duplicating the rest costs per feature.
6. Whether GraphQL fragments are colocated with components, and if typed codegen is fully leveraged.
7. Which tests the team doesn't trust: flaky suites are more expensive than no suites, because they train people to ignore red.
8. How accessible the core creator flows are: keyboard, contrast, screen readers. Creators are more diverse than personas admit.
9. Whether AI-assisted development has guardrails the team agreed on (review norms, generated-code ownership) or just habits.
10. What the slowest recurring frontend task is for the team, and whether tooling, docs, or deletion fixes it.