As LLMs penetrate more and more areas of work, I feel compelled to warn against something that those of us in the trenches (software engineers) are battling. I call it "AI Productivity Theater."

Hopefully the phrase "Productivity theater" evokes something for you. It's the feeling of updating ticket statuses, or sending followup notes, or writing a really tight meeting agenda. I don't mean to say that these actions aren't useful. Rather that they're merely actions taken on the way to a higher goal - without that goal, they become theater. It's very possible to fill your days with a high volume of this kind of work, while avoiding the more important questions, which often require deep thought and reflection.

LLMs, for all of their benefits, do something dangerous - they make it possible to engage in more traditionally "expensive" areas of work in a way that is essentially productivity theater.

Take software engineering as an example, as this is what I know best. Because designing and building software is an expensive and complex task, it has become the specialized domain of software engineers, who are intentionally walled off from the daily noise of business operations. Their work is fed through roadmapping and refinement, by which market feedback and technical reality are reconciled into a design that can be implemented without additional context.

Because the work of implementation is getting faster and cheaper, everyone involved in this process is naturally trying to speed up the rest of the cycle - the ideating, the designing, the refinement. The problem is that LLMs are a generative tool, not an intentional tool. They fill out your prompt - they do not attempt to understand your market or product. The result is high velocity in a vacuum of understanding.

To make it more concrete, this looks like:

• Shipping random features, just because you can
• Pointless refactors
• Outsourcing product and design decisions to LLMs
• Summarizing other peoples product and design docs with LLMs, instead of reading them

The danger here is that you can go on for a long while (I think my personal journey through this lasted a couple months at least) without knowing that you're in it. It's especially dangerous as a group dynamic - you'll all enter a kind of LLM psychosis, slinging tokens at warpspeed while your collective understanding of what you're actually doing decays.

This is not to say that LLMs have no role in these areas of work. But they must be applied carefully.

So, if productivity theater is to be avoided, what are we supposed to do with all our newfound time?

For software engineers specifically, I see a few options. All involve expanding your scope:

• Propose bigger technical projects. Go through product docs from the last couple years and try to find some that were great ideas but too expensive to build.
• Get involved in marketing. This might mean reviewing technical content, writing it, or helping plan conferences. This works especially well if you have domain experience in the market your business serves
• Do more product design. Talk to customers, think about the problem space, and propose product enhancements
• Hammock driven development