中文
Figma Config

Musician RAC, twenty years in: once AI makes competence free, your point of view is the only asset that can't be averaged.

It used to be hard to be average ft. André Anjos (RAC) | Config 2026 · Noah Levin

27 min
AI ProductAI Coding

26 min total·Actually worth watching closely: ~9 min·2 must-watch clips

Orange = the 9 minutes worth watchingFor the rest, the guide is enough
Segment guide · 5 segments
  1. 0:12 7:15Listen

    The worst software taught him the most

    After Noah Levin's introduction, André goes back twenty years to the start: teaching himself music on the most primitive tools, the kind that couldn't do anything for him; no tutorials, so he reverse-engineered records he loved; no one opening doors, so he cold-called a band's manager and built his own.

    The fundamentals come from constraints — the less a tool does for you, the more you internalize

    Pure spoken narrative, no on-screen demos; fine to listen to like a podcast — everything moves through the telling▶ Jump to 0:12
    Speaker · André Anjos
  2. 7:15 11:31Listen

    The lesson of the first remix

    How the first remix became his way into a career, and the core conviction he drew from it: aiming at an audience means aiming at an average, and nobody has ever loved an average thing. The work that actually connects is specific, and specificity only comes from wanting the thing to exist.

    Making it for yourself isn't a retreat from the audience — it's the only route to them

    A stretch of spoken argument with no visual dependency; the point rides on the story, so you lose nothing listening with your eyes closed▶ Jump to 7:15
    Speaker · André Anjos
  3. 11:31 16:23Listen

    Tearing up the formula, twice

    From electronic DJ act to live band, then a deliberately anti-pop record: one continuous hour of music with no real pauses. Twice he dismantled a lane that was already paying. That "failed" record set up the most successful thing he's made, Boy, which found its listeners even under the worst possible conditions.

    Specificity survives bad weather, the generic needs a tailwind; the risk was the reward

    Narrative stretch with no live demos or diagrams — all the information is in the telling▶ Jump to 11:31
    Speaker · André Anjos
  4. 16:23 21:07Listen

    The consensus engine: average goes free

    The heart of the argument: a generative model is a consensus engine, the average of everything ever made delivered instantly and for free, and it drains the middle ground where working professionals live. What's left with value is the part it can't average — your point of view — and a point of view only counts if it can fail.

    If you can't fail at something, it isn't a point of view — it's just consensus wearing your name

    Pure argument, no reliance on slides, but it needs your full attention — these five minutes are the spine of the whole talk, so don't run them at 2x▶ Jump to 16:23
    Speaker · André Anjos
  5. 21:07 26:25Watch

    Let the machine help you think, not think for you

    The closing gives an actionable principle: use AI, but the moment you hand it the wheel it pulls you to the middle. He demos the music software and tape plugins he built with AI — twenty years of taste decide what to make, the machine only executes.

    "Finding your voice" isn't a summit, it's a daily practice against gravity — and the pull never stops; the moment you relax, you slide to the middle

    Around 1462s is the one high-value visual in the talk: the demo of his own plugins (reverb, echo, distortion, tape sampler), where the abstract principle finally takes concrete form — worth watching the screen▶ Jump to 21:07
    Speaker · André Anjos