Nobody actually sits down and declares an idea bad. What really happens is quieter and faster than that: you run a silent cost estimate — how many evenings, how much setup, how much risk of a wasted weekend — and the idea loses to that estimate before it's ever given a fair trial. The idea doesn't die on its merits. It dies on arithmetic you did in about four seconds, standing in the kitchen.
This isn't a new pattern, and it isn't really about ideas at all. Every generation's real leap wasn't "the same work, faster" — it was some estimate quietly getting cheaper. Publishing a pamphlet used to require a printing house; then it didn't. Capturing a moment used to require a trained painter and hours of sitting still; then it didn't. Prototyping a device used to mean a machine shop; then it didn't. Nobody asked to do the old job quicker. What changed was which ideas suddenly cleared the bar of "worth attempting" — and that bar moves every time the honest cost of finding out drops.
Ground that in something real, because a belief is only worth holding if it maps to an actual thing you can point to. A new idea for a piece of software carries a specific version of that same estimate: will trying this put my real project at risk, and will testing it require a weekend I don't have. DF Code's shape answers both directly rather than by promise. A new idea gets its own workspace, with its own git worktree and its own thread history — trying it never touches the project you actually depend on. And an idea doesn't need your undivided Saturday to move forward, because agents assigned to that workspace can keep working on a schedule of their own, the Conductor pattern, rather than only when you're sitting at the keyboard. That's not a feature list. It's the estimate itself getting smaller.
Once the estimate changes, the question you should be asking changes with it. Stop asking "is this a good idea." You can't actually answer that from the kitchen, standing there guessing. Start asking "what would it actually cost to find out" — and notice, more often than you'd expect, that the honest answer is smaller than the estimate you'd been quietly carrying around for months.
Name the thing you've been shelving. Not because it's definitely good — you don't know that yet, and neither do we. The point was never to convince you it's good. It was only ever to make finding out cheap enough that you'd actually try.
Q: Why do most side-project ideas never get started? A: Not because they're judged and rejected — because a mental cost estimate (time, setup, risk of wasting a weekend) kills them before they're ever actually tested against reality.
Q: What does DF Code change about that cost estimate, specifically? A: Starting a new idea gets its own workspace, with its own git worktree and thread history, isolated from your main project — and agents that can keep working on a schedule of their own, so testing an idea doesn't require setting aside a dedicated block of free time.
Q: Is this just about working faster? A: No — it's about which ideas clear the bar of "worth attempting" once the cost of finding out drops, which is a different question than doing the same work quicker.
Q: What's the underlying belief here? A: That every generation inherits the last one's tools, and the real opportunity isn't using them to do old work faster — it's using them to make previously-not-worth-it ideas worth trying.