AienOSPERSISTENT MACHINE INTELLIGENCE

#StillBuilding

If you built it with AI, you built it.

I'm an engineer. I don't type most of my code anymore. I design the system, I direct AI agents, and I make them prove what they claim. Programming and engineering are way different than coding. Nobody needs to know how to write the letters.

People who build with AI get judged for it, and it has been measured. This page is my small push in the other direction: show the work, say how you made it, and keep leveling up.

What the research found

  • In four preregistered experiments with 4,439 people, those who used AI at work were rated lazier and less competent than people who got the same help from other sources, and they were less willing to disclose it. Reif, Larrick and Soll, PNAS, May 2025
  • 1,026 engineers reviewed identical code. The author's rated competence fell 9% when reviewers believed AI helped: 13% for women, 6% for men. Harvard Business Review, Aug 2025
  • The good news: in the PNAS study the penalty was not significant among evaluators who use AI weekly or daily, and a Microsoft study found no detected penalty for disclosed AI use where AI use was already normal. Microsoft Research, VL/HCC 2026
  • 32% of employees who use generative AI keep it secret from their employer. Ivanti, 2025 Technology at Work Report

Familiarity seems to help. So the best thing anyone can do is show real work, made by real people, and say how it was made.

Engineering is the next level

If you vibe coded your first app, you're on the board. The next level is knowing why it runs, then proving it does what you say: tests, security checks, reading the change before you accept it. Take as long as you need. Nobody starts there.

I wrote down how I do it, with my agents, in Build with AI, a free guide with no signup. Start here shows what I'm building and how to check it.

The concerns are real

This isn't a claim that AI is beyond criticism. Criticize the tools, the companies and the output. Just don't shame the person.

  • Credit and consent. Artists and writers whose work trained these models without asking have a real grievance. One AI company agreed to pay authors $1.5 billion to settle claims over pirated books used in training. NPR, Sep 2025
  • Security. 45% of AI-generated code samples failed security tests with OWASP Top 10 flaws. Veracode, Jul 2025. Ask your AI to review its own code, then check it.
  • Trust. 46% of developers actively distrust the accuracy of AI tools. Stack Overflow Developer Survey 2025. Distrust the output and check the work.
  • Jobs. Some work will change, and nobody honest can say how much. Mocking people for learning the tools protects nobody's job.

Hand coders, hand artists and hand writers are builders too. This isn't a contest.

Show me what you built

Post it on X with #StillBuilding and say which tools you used, or reply to me at @drakestapleton. I'll ask in public before I share anything, and I'll take it down if you ask.

I count posts by hand from public replies. This page adds no tracking of its own. The site's X pixel runs only if you say yes to it, and privacy explains that choice.

Disclosure: my AI agents drafted this page and its graphics. Every source is linked so you can check it. Source on GitHub.