Reflection
I went quiet after ISART. Not idle, quiet. Everything since has been one project pulling the next tool out of me, and every side quest fed straight back into the main one. This is the whole chain, because the chain is the only honest answer to why it took this long.
The short version, if that's all you want: I set out to build one thing, it kept needing tools that didn't exist yet, so I built them, and the tools turned out bigger than the thing. Five projects grew out of one. From outside it reads as drift, overcooked as usual, and I know it. But every fork was forced, and they're all arriving at the same time now.
The foothold
I came out of ISART in 2022 with two things already running, both still paying out from the League years. The first was a year of Music QA on Dying Light 2. I started in September 2021, in Olivier Derivière's studio in Paris, alongside my classmate Theodore Demirdache and following the interns before us, Thomas Lanza and Mattéo Siscaro. It was my first real AAA pipeline and my first Wwise sessions the size of a house. The job itself was playing and recording the game next to Wwise's profiler and flagging every anomaly, plus a lot of value-tuning on the Parkour Music System with Olivier and his assistant Maxime Gautier. Then a four-episode series explaining the game's music, cut in After Effects with Reaper and OBS. What it actually gave me was bigger than the credit: a way of thinking about interactive music that turned out to be the exact obsession I'd been carrying, how far audio and visuals can be pushed into each other, now confirmed at AAA scale by someone doing it for real.
The second was KCX. Three days scoring the Karmine Corp x Orange sponsor film with Loic Goraguer on visuals, a Valorant / Psyops-leaning bassline built out of u-He Diva and Zebra, landing on a VI VII i instead of a triumphant major because the road to worlds isn't triumphant yet. Small in the grand scheme, but Karmine had flown me in for the showmatch a few months earlier, and that invite told me a bit of the 2015 - 2018 Yasuo status was still warm out there. Finishing both, I noticed what I actually missed. Not the game. The stream. The sharing, the live back and forth, making a thing with people right there for it.
The bar
So I started turning it over. There had to be a cocktail in here, one build that pulled in every skill I had at once instead of one at a time. Player, editor, composer, systems guy, all the roles I'd run separately, fused into a single thing where the player and the author stop being two people. I was at a bar with Mattéo, who'd gone from ISART classmate to Olivier's assistant. I pitched it as a joke: imagine [ // CLASSIFIED ]. He laughed. Then I stopped. Wait. What if.
Off the rails, as usual
It went off the rails the way it always does with me. Inside a few weeks the joke had a LoL API layer, an Unreal project, Wwise, a pile of Python holding it together, a Yasuo model I was rigging in Blender, a full narrative cinematic arc wrapped around the whole thing, and music I was writing at the caliber I'd just spent a year standing next to. The full panoply. I worked it hard for about a year, then tore it down and rebuilt the core in Unreal C++ instead of blueprints, half for clarity and half because the LLMs had finally crossed the line where I could architect something that size and have it hold. Somewhere in there the Python around Wwise stopped being glue and became its own thing: I'd effectively rebuilt a working Wwise authoring layer that drove the tool through its API and its XML. A friend got me over the first wall. After that it was me and the models, and the code just kept getting bigger.
Aware of my own degeneracy
I get lost in phases. Some of them useful, some less. I know how this looks from outside, it's the same pattern every time: a bounded task cracks open and I follow the crack until it's a project. Lightning in a bottle, and I'm aware of my own degeneracy. It's part of the process.
Wait, I'm cooking here
This is where the forking got serious. I'd stopped to do a full pass on my Reaper theme, knocked it out in a week, and figured it could use a theme adjuster. Halfway through the adjuster I caught myself: wait, I'm cooking here. And if I'm cooking, it's finally time to build the region playlist I'd wanted for years. That became the first tool in ARKITEKT. Building that, I figured I was warm enough on architecture to just do Gestalt, the modular rack I'd been circling. Gestalt kept slamming into the same wall, which was Wwise, so it grew into Dirigent, which now openly means to replace it. Wwise is top of the line and genuinely industrial, and I've been fighting its restrictions since day one of the main project for what I specifically need, so Dirigent is me building the one I actually want. And the part that pays for the whole detour: Gestalt was built for music, and its modulation math is what drives this site's visuals right now. The rack I made for tracks is running the hero.
The break that shipped as an app
Four or five months rushing the Reaper tools and I hit a wall, needed to actually stop. The problem is I eat like garbage when I hyperfocus, so within a day the break turned into a systems problem: I need a system for this. Nutritional black hole, then cooking, then a framework for cooking, then a SaaS with a registered trademark. Cookodex. Even the recovery came out of me as a build. Point the thing at dinner and it still runs.
The tracks underneath
Through all of it the music never actually stopped. When I procrastinate and it isn't games, maybe one time in four it's another track for the project. Twenty to thirty of them by now, four to eight basically finished. Same as the League years, when the thing running under everything without me noticing was already the music. Even the procrastination pointed back at it.
Where it stands
Right now Gestalt, ARKITEKT and Cookodex are all inside a month or two of shipping. The main project waits on those launching, plus another month or so on Dirigent, and there's still runway, so it stays where it's been the whole time. Classified. Two weeks ago I came off the Cookodex sprint, looked at how much web I'd built along the way, and decided the portfolio was the next thing to finish, because it's about to have people to welcome. This site is the most recent fork. It's also the front door for the rest.
Convergence
I can feel the whole thing converging now, and I like the shape it's taking. I'll name the cost, because it's real: I surfaced years later than I could have, and the pay-off, the money included, is that much later with it. But I came out the far side more capable than the version of me who'd have run straight at the one project and shipped it clean in 2023. I don't regret the detours. Lightning in a bottle, and it was part of the process the whole time.
Draft, your voice goes over it. The punchline at the bar stays // CLASSIFIED and TRANSFERENCE isn't named, your call whether to expose it. To confirm on the rewrite: the exact year of the bar and the project start, the four-to-five months of tool-rushing, the twenty-to-thirty tracks, and the "two weeks ago" beat, which will date. Before/after or process shots from any of the tools would carry the chain without a caption.