I build things that make a real problem smaller.

Why it exists

HotkeyX started as a daily annoyance. I'd be deep in Cursor or Claude Code, mid-thought on something I was building, and I'd type something lazy like “make this cleaner” or “fix the logic here.” The agent would guess wrong. I'd burn three iterations, or I'd stop coding to write a proper spec, which broke whatever flow I was in. I did this twenty times a day. At some point I thought: why isn't this a keystroke?

The thing most people get wrong

When AI gives you mediocre output, you blame the model. You wait for the next one. But the model usually isn't your bottleneck. Your prompt is.

The gap between “fix this” and a tight instruction with real context and constraints is the gap between three rounds of garbage and one good answer. Better prompting beats a better model almost every time, and it's the one variable you actually control. Everyone is waiting for smarter AI. The leverage was sitting in how they asked the whole time.

How it works, and how it got built

Select any text, anywhere on your Mac, hit a hotkey, and a second later your rough thought becomes a precise instruction. Same shortcut in Cursor, Claude, ChatGPT, your terminal, an email. No switching into “prompt-engineering mode.” It's the clarity of someone who writes specs for a living, bolted onto every input field you already use.

I built it solo, nights and weekends, around everything else. The hard parts weren't the AI. They were the boring things nobody sees: wrestling the macOS Accessibility layer to read selected text system-wide, and getting the round trip under a second. Latency was the whole game. Slower than a heartbeat and the magic dies, and people just go back to typing rough prompts.

Where it's going

In a year, every input field on your machine carries the clarity of a great prompt, in any tool and any language, without you ever thinking about it.