The biggest challenge in the next 12 months: what NOT to build.
This is the paradox that keeps me up at night. AI made building so easy that the constraint has flipped completely. The hard part isn't creating software anymore — it's deciding which software is worth creating.
I see it in my own work. Every gap in time, every moment waiting on a customer, every idle cycle between requirements — I fill it by building another feature. The system grows. The scope expands. The complexity accumulates.
But should I be building all of this?
Traditional development naturally prevented feature creep. Building was expensive. Each feature had a cost that forced prioritization. You had to ask: "Is this worth the engineering effort?"
With AI, that question becomes invisible. Everything is easy. So everything gets built.
The real work isn't technical. It's discipline. It's learning to sit with the temptation to ship and asking instead: "Does this actually serve the user? Does it solve a real problem? Or am I just building because I can?"
This is the unglamorous part of the AI revolution. Not the speed of shipping. But the responsibility of choosing what to ship.
Part of the #100DaysToOffload documenting agentic development in 2026