Korva is a scroll of the internet’s best ideas.
Design can't be sped up like code
Hand a coding agent a binary request and it usually gets it right. Design is messy and fluid. An idea looks great, then fails when it is tried, and the next attempt fails too. Then it goes to users or to colleagues, and the feedback says it is wrong.
AI can get many more ideas out fast, but the feedback loop is still there. At a bigger company, getting many people aligned adds overhead on top of that.
Designers do get more done. It just isn't the same direct output as engineers multiplying theirs.
Why do coding agents handle binary requests well?
A coding task often has a checkable outcome: the tests pass or they fail, and the program runs or it crashes. That gives an agent a clear signal it can use to correct itself over many rounds. Design has no equivalent pass-or-fail check, because whether an interface feels right depends on people. Automated tests have been standard in software engineering for decades, which is why coding tools could build on them.
Why does design need a feedback loop?
Design theorist Donald Schön described design as a reflective conversation with the situation in his 1983 book The Reflective Practitioner. A designer makes a move, sees how the material responds, and adjusts. He called the cycle seeing, drawing, and seeing again. The conversation can speed up with better tools, but each move still needs a response before the next one.
Why does alignment slow large companies?
Fred Brooks argued in The Mythical Man-Month, published in 1975, that adding people to a software project adds communication overhead. Each new person must be briefed, and every decision involves more people to inform. Brooks used this to explain why adding staff to a late project tends to make it later. Overhead of this kind grows with the number of people who have to agree.
More from Lenny's Podcast: Product | Growth | Career
