Speedrunning your career

The slack messages flow by.

It seems impossible to get a grip. Every day, every hour there seems to be a new pattern, strategy, idea that feels pressing to understand and remain relevant. The experience of working in software can feel like that hamster wheel, continually rotating, never getting anywhere. The experience in the last 24 months has felt even more existential than that.

Seeing colleagues (or maybe competitors) tackling these major feats; coordinating agents, pursuing orchestrated flows, writing features in their sleep. It can feel overwhelming to stay current, keep relevant, keep up.

I likely contribute to this frenzy in my circle of influence. I am producing more at a baseline, maybe 3 to 4x my best output pre LLM. My work showcasing AI leverage is frequently referenced within our strategy and mission setting conversations. I’m also an optimist; I believe in the good of technology and am likely biased against the evil that lurks beneath the surface. I remember feeling so optimistic about Twitter in its early days. Faithfully certain that democratization of information would create the world I hoped to grow old in. And here we are.

I’m here to remind us that:

Every line of code we write is a liability.

My aim as software engineer is not to produce more code, it’s to produce better outcomes. I believe we are in service of humans, solving human problems. The purpose of AI is not to create more debt. I recently was linked an article that wrote:

I have a thousand+ hours of experience with vibe coding and one of the more interesting and unexpected things that I regularly encounter is this:

“after spending many hours with a swarm building an application with a bewildering infrastructure of services, API calls and complicated front ends, I often reach a point where I realise the problem can be described in a form that can be solved far more quickly in a spreadsheet.”

To me, this reads like waste and the epitome of a hammer seeking a nail. Start with the spreadsheet, and then figure out if you actually need a web site. AI is overly eager to produce interfaces and solutions that don’t leverage AI but, unsurprisingly, are built for an era of interfaces and solutions that just repeat our past.

Bad UI, complex overly engineered SaaS solutions; this is after all, the training set from which they are born. Why would they not recreate it?

AI by default won’t try to solve it in the narrowest slice, it’s too eager to please. Resist. Be insistent. Reframe.

/clear

So much of the leverage of AI is not in production code

Most of how I use AI is to explore the edge cases and the ‘whatifs’. Yesterday, at the whim of intuition, I was able to reduce the total cost of storage by ~30% on a client project with a few minutes of exploration. I landed on this optimization based on experience, but mainly inquiry. Questioning how this particular database stores data led me to realize the culprit wasn’t what I expected. I leveraged this insight to create a more optimized schema to reduce storage cost. Use AI to explore the multiple universes you used to ignore because of the time constraints. Go wide, review, then go deep.

Use cycle time wisely

Let AI write the code, use your time to learn, understand, go deep on the domain or the tech. This is the core value of AI in my mind; it frees you up intellectually to double down on intellectual pursuits. Writing code takes a lot of focus, motivation, creativity, and time. The beauty and the grief of code authorship are the same — there’s never enough time. What you’re doing when the LLM is working for you, well, that is exactly the point.