Recently, I've been working intensively with coding agents, and I made a discovery that surprised me.
AI excels at getting things to 80% or even 90%. Give it a task, and it quickly produces something that looks decent. But "looking decent" doesn't equal value. Software engineering has no objective standard of perfection—zero-bug code that nobody uses is worthless. Direction matters far more than execution.
Two States
During this period, I've been switching back and forth between two states.
When my hypothesis is clear, I basically can't stop. I know what I need to validate, and the coding agent helps me rapidly build the relevant components, running end-to-end experiments with the entire pipeline compressed to the hourly level. I can complete two to three rounds of hypothesis correction per day, with ideas being confirmed or refuted within hours. This rhythm is addictive.
But there are also days when I sit there not knowing what to do. Without something specific I want to validate, I dig up previous projects and have the agent polish this part or fix that part. It looks like I'm busy, but I know in my heart that this is just idling.
Both states use the same set of tools. The difference lies entirely in whether there's a clear question in my mind that needs answering.
The Funnel Collapses
Traditional product management has a classic concept called the ideation funnel.
There's a "fuzzy front end" phase at the beginning, where you generate lots of ideas and then filter them. Why filter? Because implementation downstream is too expensive. Taking a product from concept to launch typically requires 3 to 6 months of development. Since implementation is expensive, you need strict gatekeeping upfront.
This logic no longer holds now.
An MVP can be built in a day plus a few hundred dollars in API costs. From the birth of an idea to having someone actually use it takes a week or even just a few days. Implementation isn't expensive anymore, so the premise for screening disappears. Previously, ideas were cheap and implementation was expensive; now it's the opposite.
Suppressed for Too Long
The work environments of the past have actually been suppressing this impulse to "propose ideas and validate them."
Most people were required to execute. If you had too many ideas, you'd likely hear: "What's the point of thinking so much? Just do your current job well." In an era of limited resources, everyone was competing for execution resources, not ideas. So many people's hypothesis generation abilities have been atrophying.
Now the agent era has arrived, and execution capacity is suddenly unlimited—a surplus of productivity. But this surplus productivity doesn't know where to go.
The Bottleneck Shifts
In the value chain, AI has amplified certain links by tens or hundreds of times. The links that weren't amplified become the new constraints.
Previously, when building products, development took the bulk of the time, and everyone was waiting on development. Now development is compressed, and other places become bottlenecks. Do you have good enough ideas? How do you get the product into users' hands? Previously insignificant minor steps have become the slowest part of the entire chain. No matter how fast the other parts of the chain are, they have to wait here.
Product Managers Have Changed
"Everyone is a product manager" is no longer just a slogan now.
Product managers used to be coordinators, managing processes, helping others realize their ideas. Now a different capability is needed: the ability to judge what has value, to break down a general direction into the smallest testable hypotheses, and then test them one by one—correct when wrong, build upon when right.
What makes this scarce? It's not methodology. It's drive. The drive to personally get your hands dirty, direct agents to validate, adjust after hitting walls, and keep trying continuously.
When you have this drive, AI feels completely different. It's not about "efficiency gains." It's that my ideas can become reality. Previously, you had to convince people, fight for resources, wait for scheduling. Now you don't need to convince anyone. Productivity is in surplus; what's lacking are people who know where to apply it.
Originally published at https://guanjiawei.ai/en/blog/productivity-surplus
Top comments (0)