It has been a wild near half decade ride on the GenAI coding tool beast.

If I’d ridden a bucking bronco like I fantasized about as a kid it would serve as a clean overlaid transposition of my mind with my ass. 🐂 🤠

I first used a GenAI coding tool in October 2021, when I was added to the GitHub Copilot Technical Preview. It was interesting, but barely better than IntelliSense—which is a tool collecting dust in my garage now that I think about it. I continued using it and as its performance improved it became my daily driver while continuing to use Neovim to get my digits dirty.

A ChatGPT Overflow

In May 2023 I subscribed to ChatGPT and began to see GenAI’s true potential. It quickly generated working code, which I happily copy-pasted, and gradually replaced my manual StackOverflow workflow other than the reading part. Answers to my coding problems came faster, though not always accurate.

Maybe StackOverflow will find its way back to public consciousness after numerous declarations of its death.

Skipping Claude for Windsurf

Next up was Windsurf. I used it on a contract project in December of 2024 and was able to produce a complex AWS Step Function workflow and a UI that interfaced a lambda which processed a CSV file in an hour that would have taken me several hours to produce from scratch. I started taking the technology even more seriously and bounced between GitHub Copilot and Windsurf for a while. I briefly used Claude which had features similar to ChatGPT at the time with no code assistance.

A Long Blinking Cursor

In March of 2025 I started using Cursor. I was immediately impressed with the agent capabilities and the workflow. I was able to generate code and use its agent capabilities to automate several steps in my software development workflow. I was also able to generate documentation and commit messages with ease. I started started using Cursor as my daily coding driver, and began.

I briefly tried using Google Antigravity and it wasn’t an improvement over Cursor, so I quickly stopped using it. I also tried Loveable and Replit out of curiosity, and quickly ducked out of both. I’m not having the WYSIWYMG (What You See Is What You Might Get) ™️ experience. I still have a touch of PTSD from continuously reinstalling Front Page extensions at Internet In a Mall.

A Kiro of Code

In February of 2026 I tried AWS Kiro which was a forked VSCode with the flavor of Cursor paired with Speckit, OpenSpec, and BMAD which I had already evaluated and found to be overweight GenAI coding tools though I get the allure. Coaxing GenAI coding tools on a straight and narrow path is fraught with peril if nigh impossible, so the spec driven development approach seems reasonable.

AFAIK, Kiro was the first tool to embrace the spec-first approach. As I was vibe coding a keyword matching algorithm with little success I went back to the drawing board with Kilo and it improved the results, but it wasn’t a satisfying experience so I went back to Cursor.

Hello Again, Claude

After using Claude Code from time to time after its release in February of 2025 I started using it for my daily code driver. As Simon Willison mentioned in a recent Lenny Podcast, Claude better matches my taste for how code is supposed to look. Despite my tangos with Cursor, VSCode, and the like—my true love is the CLI. I’ve started using tmux with multiple windows running Claude as my primary workflow. I’m still not ready to let The Mayor take over my day-to-day coding operations.

Catching The nWave

A couple months ago I stumbled across an nWave post, having spun away from Gas Town heeding Steve Yegge’s warnings in January.

I started using nWave in earnest on March 21st 2026. After doing so, my ability to deliver high-quality code with GenAI shot up like a rocket. 🚀

nWave requires no upfront Markdown scaffolding and outperforms vibe coding by a wide margin. It stays on the rails where Gas City flies off them. I’m still keeping my eye on the wasteland, I believe it points toward the future, but after seeing it and OpenClaw in action, I feel compelled to ground myself in software development fundamentals.

nWave bakes in software fundamentals like outside-in TDD, adversarial review, and mutation testing. It also works with OpenCode if the recent Claude Code hack scared you straight, and support for ChatGPT and Codex is in on the way.

The workflow is clean. Load your CLI, type /nw:new, and you are on your first wave. Describe the feature you want to build; nWave asks follow-up questions scaled to how much up front detail you provide.

I have oscillated among /nw:discuss (jobs to be done), /nw:design (architecture deep dive), /nw:distill (acceptance tests), and /nw:root-why (unwinding sticky bugs). I have not yet needed the other commands, but I plan to lean on /nw:refactor once I finish building my open-source small-business management K3 cluster.

If you’re interested in learning more about “why nWave”, jump into this conversation about the fallacies of “spec-driven-development” (SDD) on LinkedIn. And, so you don’t miss it, look at this data-driven article which refuses claims that SDD provides an code intent audit trail.

I like Andrea Laforgia’s mental model of nWave as Expectation Driven Development.

What’s Next?

More dirt roads and wastelands will appear in the coming months, but nWave feels like a solid ride.

I enjoy observing how people are reacting to GenAI. The history of AI and ML shows this pattern repeating: hype, backlash, then real adoption. The spectrum of behavior rainbows from “You’ll pry my code from my cold dead hands.” to “We’ll build robot-building robot factories with a Minority Report interface.” I hang out somewhere in the middle. I’m looking forward to Gas Town’s mayor reporting on its garden-building directive in my back yard, but I’m still picking weeds and watering my plants by hand.

Just about anyone can build a viable software product with these tools as evidenced by the glut of mostly useless tools flooding every zone—a new AI memory manager seems to drop every other day with the likes of Milla Jovovich getting in on the action. I suspect there will be zombie-themed GenAI coding solutions shambling to a Git Repo near your.

We’re in the “Hey! Look ma, no code.” phase of this general-purpose technology wave. I plan to enjoy watching the showoffs fly into ditches, but I won’t miss having fun as a newbie alongside you and everyone else joining the party. 🧪 🎉

References