There's a shift happening in how ideas become things. With the era of agents upon us, we are seeing a step change in the path from ideation to outcome. And the most interesting part isn't the tools themselves. It's what the tools stop asking of you.
Let's take the example of basic graphic design. For years, making something visual meant learning a tool first. Affinity Designer. Linearity Curve. Even PowerPoint pressed into service. These are powerful applications — skilled people make extraordinary things with them. But for someone with a design instinct and without the muscle memory to operate these environments fluently, the tool becomes the obstacle. Most of the creative energy gets spent navigating menus, not navigating ideas.
That changed recently. A combination of voice transcription and capable LLM-powered agents like Claude Code has opened up a vista I didn't even know existed. Over the past few weeks, on and off in my free time, I've been dabbling with making logos for various side-project needs. I've been going through fifteen, twenty iterations, in rapid successsion, changing direction often substantially, exploring the narrative arc of what a mark is trying to say. In each of my recent projects, attempting to craft a logo quite simply the conversation was the design process. I described what I wanted to feel. I named what wasn't working and why. I talked about tension and balance, about whether something felt too visceral or not grounded enough. And the tools responded at that level of abstraction.
There's a show — Mad Men — about an advertising agency in the 1960s. What always struck me about the creative directors in that world was that they never picked up a pencil to make a layout. They articulated a feeling. They named a tension. They said: this is what it needs to evoke. And someone translated that into form. That relationship — between the person who feels the thing and the person who makes the thing — used to require another human in the middle. Now it doesn't, at least not always.
What's powerful about this isn't automation. It's abstraction. The outcome is still entirely a function of taste — of values and sensibilities and the judgment calls that only you can make. But the conversation has moved up a level: from how do I do this to what is this trying to say. That's the shift. And it's not a small one.
The same thing is happening in software development — perhaps even more dramatically. The conversation has moved from syntax to intent. From debugging loops to describing outcomes. The person building an app is increasingly the person who understands what the app needs to do, not necessarily how the code achieves it. That's a different person than the one who could build apps before.
The activity timeline above is my own. The sparse stretch from spring through fall was a period of sporadic side-project work — Cursor-assisted, heavy on copy-paste, a lot of back-and-forth between idea and execution. The kind of output that reflects how much friction sits between having an intention and realizing it. Then a small cluster over Thanksgiving: a week off, four working days, a shift to Gemini's models and a new editor, Antigravity, and more meaningful progress on an iOS app I'd been building than I'd made in months of intermittent effort before it. Then March 2026. That's where the graph gets dense. Wispr Flow for voice, Claude Code as the primary coding agent — and the output changed materially. Personal websites updated, new features shipped on the iOS app, a publishing workflow automated through GitHub Actions. Same hours. Same nights-and-weekends life of someone with a full-time job. What changed wasn't the time available. It was how much could happen inside it.
Writing is no different — and perhaps the most surprising example of all. Getting a thought onto a page has always had two problems: finding the thought, and finding the words. The second, it turns out, is the easier one to offload. A kernel of an idea — a tension you've been noticing, a framing that feels half-right — can now become the opening of a conversation rather than a blank page. You push on a paragraph that doesn't land. You redirect something that drifts. The back and forth is genuinely collaborative, not just generative: the ideas still have to be yours, the judgment calls still have to be yours. But the activation energy required to move from "I've been sitting with something" to "here it is, written down" has dropped considerably. Blogs, essays, the act of thinking out loud in public — these, too, are becoming more accessible to the person with something to say and less dependent on the person with time to craft every sentence.
Every meaningful abstraction in history has worked this way. When spreadsheets arrived, they didn't just make accountants faster — they made non-accountants into people who could model their own finances. When desktop publishing arrived, it didn't just accelerate print shops — it created a generation of self-publishers who had never touched a press. The new capability rarely looks like the old capability made quicker. It looks like a different population of people doing things that were previously out of reach.
We can't yet see clearly what this abstraction will unlock. The new things that become possible when the barrier between feeling and making collapses — those things are probably not imaginable from where we stand today. What seems clear is the direction: the tools are becoming fluent in intention. And that changes who gets to make things.
The most interesting things haven't been made yet — by people the old tools never let in. This fluency of intent is their invitation in.