Eating Your Veggies in the Age of AI

I’ve always loved eating my vegetables, in spite of the common opinion among children that they are not appetizing. I have childhood memories of sneaking into the kitchen refrigerator, popping open the vegetable drawer, and snacking on the tomatoes that my mother had planned to use for dinner. Spinach, broccoli, bell peppers… none were safe when I was snacking.

The term “eat your vegetables” can also be used as a metaphor for doing something that you don’t necessarily want to do, with the added benefit of having a positive result down the line.

Knock back that spinach and you’ll grow up to become big and strong. Suck it up, deal with the unappealing task on your plate, and reap the rewards later.

Yet, the tech industry is currently experiencing a collective childhood tantrum, trying to sweep those unappealing tasks under the rug. Everywhere we look, some AI evangelist or CEO is seeking to change the foundations of the tech industry:

• Engineering jobs are done.

• Vibe coding is the future.

• Replace all of your workflows with AI.

It’s exhausting and flawed. Sure, AI is an accelerator and may be here to stay, but the LLMs of today will never be able to replace the abilities of a software engineer who has spent their career building the capabilities necessary to deliver real value. Of course, that’s only true if we have strong foundations.

Just like the greens that people love to dump into the garbage, many are quick to shirk the tedious chores of yesterday. Difficult tasks like writing thorough unit tests, debugging legacy code manually, thinking deeply about architectural trade-offs, or parsing raw documentation and comment history are passed off to AI.

Business mandates or market conditions may require teams to use AI so that they can output at a breakneck pace, but engineers should always ask whether they understand the problem and how they would solve it without assistance. If the answer is not clear, then using AI should be seriously reconsidered, or at least downgraded to a learning assistant.

Over time, without accrued human experience to help shape and guide AI, output will always tend toward slop. So, we had better eat our veggies before we make a mess of our future.