Often, people who build tools don't end up using them. They build something, maybe use it a few times out of obligation for the sake of dogfooding, then drift back to whatever they were using before. The tool becomes a fancy portfolio project and never really integrates into their workflow. But what I've been mulling over recently is when someone builds something they genuinely prefer using. Not something they're proud of in the abstract, but something that's actually better for their workflow than the alternatives. This is surprisingly rare.

Companies may famously eat their own dogfood, but the phrase itself gained cultural relevance from Microsoft, where Paul Maritz was desperately trying to increase internal adoption of a product employees were avoiding. The gap between building something you're theoretically proud of and something you actually reach for daily turns out to be vast, and most tools quietly fall into disuse even among their creators. Even among developers, while they may approach their own creations with more patience, they're not immune to these abandonment curves.

And I think one of the reasons why most tools fail to transform from interesting experiment to essential part of workflow because they are built to eliminate friction rather than amplify joy. They promise to remove obstacles between you and the work, but rarely make the work itself more compelling. They help you finally get around to tasks you've been avoiding, but they don't transform those tasks into something you actively want to do. It's you building elaborate systems for avoiding the thing you claim to want to do. It's the difference between building a better alarm clock versus designing a life that makes you actually love mornings.

This isn't universally true for all tools, of course; tools in many domains truly are about friction reduction, where the work itself is inherently unpleasant and the goal is simply to minimize suffering. But for creative and intellectual work, where engagement itself is often the bottleneck, this distinction matters most, because work that requires sustained creative attention — writing, coding, designing, composing — the quality of engagement directly shapes the outcome. For these domains, tools that only reduce friction often fail to create any lasting commitment or impression because they treat symptoms (inefficiency) rather than the root cause (disconnection from the work itself).

So, the tools that become indispensable are the ones that make you actually want to engage with the work itself, that turn the process from something you endure into something you seek out. Even in a janky state, the tool has to solve your real problem — which is often not the friction itself, but the fact that the work feels like something happening to you rather than through you. Eliminating friction helps when the desire is already there, but most abandoned tools reveal that you're trying to optimize our way into caring about something you fundamentally experience as a chore, when what you really need is to transform the nature of your engagement with it.

I've been building an AI text editor called Unslop for the past few months. The first feature I built was inline edits. You select any phrase, sentence, or paragraph and make surgical changes to specific parts. You can do multiple selections at once, so the text never gets fully rewritten by AI. Instead of the usual “rewrite this” workflow, where the model spews a sloppified version of your entire text, Unslop let me stay inside the writing. I can perform surgical edits without losing control of the tone or rhythm. It makes me an essential part of the loop and all edits are driven by intent, not the spray-and-pray approach typical of someone using ChatGPT or Claude to give them an essay, or worse, a LinkedIn post.

Naturally, this immediately solved a real problem I had, because good writing is all about making specific, highly intentional choices. You don't want the LLM to rewrite everything, only augment you in making spot fixes while keeping everything else exactly as you wrote it.

That first feature already did 80% of what I wanted. Everything else I’ve added since has just been refinement. The brunt of the work was done once the tool crossed that invisible line from “interesting experiment” to “thing I instinctively reach for.” It was genuinely more useful for this specific use case than any already available alternative.

Perhaps, another way to look at it is seeing creative tools as control surfaces with tight feedback loops. Tools stick when they increase the density and granularity of meaningful actions under your authorship. Unslop’s multi‑select + localized mutate + reversibility is a new degree of freedom with a low blast radius, so the work feels done “through me,” not “to me.”

As a general principle, for low‑motivation, procedural tasks, a tool that removes steps will win the trial. But in creative work, where desire is the bottleneck, the hierarchy flips and tools that tighten your action‑feedback loop or “flow” beat tools that merely sand edges. Friction reduction drives initial adoption, but agency drives retention and preference. Heck, even in boring domains, people stick with the product that leaves them feeling in control: legible state, direct manipulation, reversible actions, and fast feedback when they poke it. So, the tools you'll actually prefer a year later is the one that lets you see what changed, undo it safely, and steer outcomes.

The insight here is that for you to use the creative tool you're building, you need real frustration with existing tools, and your first feature has to cross the threshold of “minimum viable magic” that unlocks one decisive degree of freedom for a narrow job so the local fitness landscape tilts toward repeat use. Most attempts to build creative tools fail one of these conditions.

But the thing that makes a tool stick is whether, early on, it manages to capture the essence of what actually increases your engagement with the work. Once that happens, iteration is inevitable, because you’re not imagining a user anymore. You are one. And you use because it's already better than alternatives at something specific that you care about. Everything else follows from that.