Why Is Writing Code Almost Solved?

Aravind Srinivas says the rise of NEO cloud shows the demand for tokens and agents is there, and he calls the demand a good problem. If everybody tries to build another NEO cloud, he says, there is not actually any differentiation. He says maybe some NEO clouds differentiate each other by their own power or by software they can run on top. Even so, he says it is just yet another thing. While everybody searches the same market, he says they are trying to build their own market, which is why he likes to think they have no competitor.

Number one, they are not going after coding just like everybody. Aravind Srinivas says the bottleneck is not in producing more code. He thinks most of the best engineers, asked if they are bottlenecked on writing more code, would say no. They would say they are really bottlenecked on problem solving and do not have enough time to problem solve. Once the problem, the specs, the system design, and the constraints are clear, writing code is dead simple with AIs today. He says it is almost already a solved problem. Figuring out the constraints, and what problem to even solve, is the hard problem.

He does not want an AI that just spits out slop and makes a vibe-coded clone. He wants the AI to say it is stupid of you to even go and build the clone, then ask what the business model is, why you even have a right to build it, and how it fits into your strategy.