AI reinforces cowboy development

In the early days, technical co-founders were used to shipping fast. It meant SSH’ing into the production server and pushing straight to production. It got them to product-market fit. But now there’s a team in place, and suddenly everything feels slower. So they revert to old habits. Skip the process, pull requests, and pair programming. Software development in the Wild West.

The founder proudly shares how productive they’ve been. They shipped a feature in a day. Meanwhile, the team was “wasting” time in refinement sessions. But the fallout always lands with the team. Side effects the founder didn’t anticipate. Backwards compatibility was overlooked. Data inconsistencies were introduced because it was quicker to create a new table than to extend the existing schema.

Then comes the question from the co-founder: “Why can I ship so fast, but the team struggles to deliver?”

AI is accelerating this dynamic. Need to refactor 5,000 lines? Just run it through Claude or GPT-4, and ship the output. However, the real work begins after the code is pushed. That's when the shortcuts break down.

If you want to contribute, follow the team’s process. Share your knowledge of the product. Instil the company’s vision. Encourage your team to experiment and leverage AI tools. That’s how you increase throughput beyond your keyboard.