The New Middle Class of Software Builders

2026-05

If you were not aware, there is a subset of people working in the tech industry right now who are primed to get a lot of value out of AI coding tools… and having a metric shit ton of fun doing so!

These are the kind of people who have a passion for building software and understand many of the systems involved, but for whom a certain level of coding ability escapes them.

They include (but are not limited to):

• Data / Analytics folk
• Designers
• Docs
• Engineering managers and people in leadership roles whose coding skills have lapsed
• IT and SysAdmins
• More frontend-y Engineers
• Product Managers
• Program / Project Managers
• Quality Assurance / Quality Engineering
• Solutions Engineers / Sales Engineers and technical sales people
• Support
• Technical Marketers
• Technical Recruiters

Some of these people would get tarnished with the label “non-technical”, but the truth is that they were always operationally technical enough to excel in their roles. Often, being a level of abstraction away from the code makes them better at their jobs. They just lacked the last 20-40% needed to independently ship software.

If you’re in this niche of people (I am), it feels like this very specific point in the evolution of AI tools was made for you.

Sure, any normie can probably figure out how to use AI website builder tools, or even some of the more “dynamic” consumer-oriented AI tools. But if you want to build almost anything, you need a certain level of technical understanding of the systems involved in order to build with minimal coding and/or vibe coding. Claude Code, Codex, GitHub Copilot, etc. are all still too close to the intricacies of writing code for truly non-technical people to grapple with.

The ‘Who can build software end-to-end’ abstraction layers:

1. One year ago: Dedicated software engineers needed
2. Today: People who understand enough about the underlying systems
3. Next year: Perhaps anyone who can clearly articulate a problem or outcome to an LLM
4. ???

Any movement between 3 and 4 is super interesting to me. As I can’t imagine a world where good software can be built without someone being able to clearly articulate a problem or outcome to an LLM. Surely good prompts are the final barrier? …But I could be wrong!