AI lets bad developers suck faster — end-to-end skills still win

Important context on the impact of LLMs on software developers.

A lot of developers suck at the non-syntax-generating aspects of software development. Many kind of suck at syntax generating.

And now with AI, they can all suck faster.

Sorry to buck “everyone gets a trophy” culture, just telling you what I’ve seen in my long (😩) tech career and positing *part* of why the feedback about the usefulness of LLMs by devs is so disparate.

Many developers were trained, or prefer, to stay in the “developer” lane — maybe dabble in architecture. Though to be frank, memorizing design theory and applying it to a real world system are two very different things.

But the "translate design into syntax the computer understands" slice of the development lifecycle is shrinking. Fast.

What we need now are people skilled across broader swaths of the solutioning process, if not end-to-end, using AI as power tools.

A lot of developers are used to being the tools in that equation. Those are the ones who will be hit by this paradigm shift the hardest.

But if you’re someone who can...

👂🏻 Ask stakeholder's the right follow-up questions

💡 Design for common paths & edge cases

⚖️ Balance disparate requests into elegant / practical solutions

🤠 Wrangle stakeholders through projects

… you’re fine. There will still be a market for your skills — even as LLMs get their syntax together.

I didn’t come into software through a formal CS path. I stumbled into it — exposed to the sausage-making in fits and starts: databases, scripting, code. The more I learned, the more problems I could solve. Not as a “coder,” but as a problem solver who used tech as a tool. Eventually, my roles shifted to center around building solutions. I was doing Agile before I knew the term... soup-to-nuts solutions.

Those end-to-end skills have never been more valuable.