← Carlos Cardona

The other axis

I met a guy who automated most of his life. His to-do list runs itself: reminders, errands, follow-ups, scheduling. He has more free time than almost anyone I know, and he uses it well. This is not a story about a fool. It works. It is what most of us are chasing when we reach for AI, and he actually got there.

But the story kept working on me afterwards. Everything in it happens on one axis: speed. The same tasks, done sooner. The same week, compressed. Almost all the tooling is built for that axis, almost all the discourse celebrates it, and when we say someone is "good at AI" we usually mean fast.

There is a second axis, and almost nobody is working on it: quality. Not doing the same things sooner, but deciding better. When I look at where things actually go wrong for me, it is almost never because I was slow. It is because I decided with too little in view. The bottleneck on a decision was never time. It was context, how much you can actually hold in your head when you choose.

A small example. I recently asked my assistant for an exercise plan. Nothing exotic, I just want to get back in shape. But it knows I have a herniated disc, because it knows almost everything about me, so it checked the plan against that. Ask anyone which swimming stroke is the gentle one and they will say breaststroke; it is the one every cautious beginner picks. For my specific injury it came back flagged as one of the risky ones, because the head-up breathing arches exactly the part of my back that is damaged. The plan was not produced faster. It came out different. The obvious safe choice was the dangerous one, and I had no way to see it on my own.

That is the whole idea in miniature. Context did not save me time. It saved me from a plausible mistake.

I have been building toward this on purpose. My assistant works from a knowledge base of my life: projects, health, finances, people, principles, decisions and why I made them. Building it costs me time instead of saving it, which feels strange to say out loud, because it runs against everything the productivity story taught us. I am slower this month so that every decision from now on gets made against everything I know, instead of against whatever I happen to remember in the moment. Saved minutes do not compound. Better decisions do.

Now take the same idea to product development, because that is where I live. The AI story in product right now is a speed story: features that took weeks ship in days. Real, and I use it every day. But the expensive mistakes in product were never typing mistakes. They were decision mistakes. Building what nobody asked for. Misreading why users leave. Rerunning an experiment that already failed two years ago because nobody in the room remembers running it. That last one is the breaststroke moment at company scale: the roadmap item everyone agrees is obviously right, while the archive knows something the room does not.

A CTO who runs a seventeen-developer team told me that when his people pick up a feature they look at the code, but the code is not enough. The real problem is that nobody shares the same context, developers and product people least of all. He was describing the bottleneck without naming it. His company does not lack speed. It lacks a shared memory to decide against. Imagine the same decisions made against every user interview, every support ticket, every churn reason, every experiment and why it failed, instead of against whatever survived into the meeting. Not made faster. Made better.

None of this is an argument against automation. I automate plenty, and the point is not to be slow; nobody wants slow. The two axes are complementary. Keep pointing AI at the mechanical work, that trade is good. Just notice that "how fast can this be done" is only one of the two questions you can ask it. The other is "how well can this be decided", and it is quieter, harder to demo, and mostly unexplored. That is exactly why I would look there.