Picking Up AI Tools
My industry’s changing. The sky is falling. Software engineering is dead. Or so I keep hearing.
It’s getting old. And of course, it’s mostly coming from folks who hire engineers and don’t love paying engineers. They want near-free labor, even if it means building something once and scaling it a hundredfold after. Spend the money upfront, then rake it in while the marginal cost drops to zero. But no—somehow that’s still not good enough. They want fewer humans, fewer risks, fewer decisions. What they don’t seem to realize is: once building software becomes easy, everyone does it. The market doesn’t get calmer—it gets flooded. And when that happens, it won’t be engineering skill that wins. It’ll be something else.
Right now? It’s whoever knows how to market. The best product doesn’t matter. It’s about who gets in front of customers, who owns the funnel. Maybe AI will change that. Maybe not. Feels like it’ll still reward those who game systems, not those who build great ones.
Anyway, this wasn’t supposed to be a rant about CEOs and startup fantasyland. I meant to write about AI tools. Specifically: how I’ve been catching up on using them.
Getting My Feet Wet Again
I’ve used ChatGPT, Claude, and Grok. Even gave Gemini a whirl back when it was still clunky. GitHub Copilot? I bounced off early—it was basically autocomplete at the time, no chat, no file awareness. I was barely scratching the surface.
But now that I’ve got some time—hiatus, unemployment, whatever you want to call it—I’m going deeper. I’m not “vibe coding” just for kicks (I can still out-code most vibes), but I’m not taking full advantage of these tools either. So I’m poking around. Cursor is interesting. Copilot’s catching up. I still don’t trust it, but maybe it deserves a second shot. Zed’s AI integration looks promising. I’m keeping tabs on Claude 3.5 and GPT-4o. Also stumbled across bolt.diy, which might be worth something. I liked bolt.new briefly, but it hit walls fast.
What I Want From AI
I want more than autocomplete. I want an AI that can live inside a project. One that can remember things. One I can feed markdown specs, architecture docs, goals—and have it actually use them, not just spit them back. Something that edits multiple files, suggests diffs, and lets me review and apply. Something that feels like an assistant, not a search engine.
Maybe that already exists. Maybe I’m just missing it.
But that’s the dream: faster iteration, tighter feedback loops, less mental overhead. And still fully mine.
The Tradeoff
Here’s my worry: that I get dumber.
I’m okay relying on tools for speed. I’m not okay losing the edge. Forgetting how to debug. Losing fluency in the craft. That’s the line I won’t cross. I’ve seen it happen. Like when we stopped memorizing phone numbers—only this time it’s skills, not digits. The solution? I don’t know yet. Maybe regular drills. A mental warm-up. Something to keep the blade sharp.
Still Figuring It Out
I don’t know where this goes. I don’t know what stack of tools I’ll land on. But I know what I want: more velocity, without the brain rot.
What about you? What tools are you using? Are they making you faster—or just making you forget?
Let’s compare notes.
1 thought on “Chasing Velocity Without Losing My Edge to AI”