AI Gives Everyone the Same Answer. Novelty Is the Only Moat.

The core argument. After two years of building over a dozen AI-assisted applications, Hari Prasad Sudharshan landed on a sobering observation: most software built with AI tools lands in the same zone. He calls it the median. Whether you use Claude Code, Cursor, or Codex, the outputs converge. Models and harnesses are rented, the problems are well-known, and the solutions look alike. The industry is producing enormous volume, but similarity, not freshness.

The build-first trap. Sudharshan describes a reflex he calls 'build first.' When confronted with a problem, the natural instinct now is to open an agent and start scaffolding. Two hours later, you have a demo. Four weeks later, a shippable product. But that product is likely the 400th RAG system or the 200th cost calculator. The dopamine of rapid creation replaces the hard work of deciding whether to build at all. He admits falling for it repeatedly.

The median is free, novelty is scarce. The real insight is economic. Before 2023, a working product required a team, time, and a full stack. Today, a solved problem is free. And free for everyone. The consequence is a flood of similar products where distribution, not product quality, becomes the moat. Sudharshan argues that the only competitive lever left is what he calls a novelty budget: at least one unit of true differentiation, the way Page Rank was for Google.

A concrete example. Sudharshan built a calculator to compare the cost of running open-weight models inside an enterprise versus consuming tokens via APIs. He found 244 similar calculators already on the market. None addressed the specific operational question enterprises face about inference cost. By focusing on that single novel angle, his calculator found traction. The novelty was forced; the rest was median.

Organizational implications. For engineering leaders, the video shifts the conversation from 'which model' to 'what is the one thing we own that nobody else does.' The model is rented. The harness is commodity. The team's unique judgment, opinion, and domain expertise are the only assets that can break the median. Sudharshan recommends three questions for any project: who has already done this, what do we have that others don't, and where will this live?

The verdict on AI slop. Sudharshan pushes back on the term 'AI slop.' He sees millions of people building software for the first time, meeting their own success criteria. That is not slop; it is a demographic shift. The problem arises when teams confuse median output with competitive advantage. Internal tools, learning projects, and hobby apps are fine. But products launched into open competition with nothing but median solutions will be indistinguishable from the crowd.