New code, trusted code, proven code

AI tooling is not impacting the software industry in a uniform way. The YC startup bubble gets most of the attention and that distorts the picture. Move fast and break things is simply not viable in big parts of the industry.

Take heavily regulated industries like finance. It’s easy to mischaracterize these companies as slow-moving or even lazy, when in reality they’re solving a harder problem. Generating plausible code faster is clearly useful. But a bank is not being asked to ship plausible products. It’s being asked to produce changes that can be explained, reviewed, approved, operated and defended later.

In an AI agent world, we have to pay more attention to the maturity level of the code we write and ship.

• New code is cheap to produce. It may be clean, well-structured and even well-tested. It may also be wrong, weakly understood or poorly owned.

• Trusted code is code that has been reviewed, exercised and understood well enough that people are willing to stand behind it.

• Proven code goes one step further. It has survived contact with reality. It has been lived with.

In the old world, this distinction was blurred, since manually tested, properly structured, handcrafted code had obvious value. With AI agents, proof of use is much rarer and more valuable than just the code itself.

Used well, AI tools can help explore implementation options before humans commit to one. They are also particularly useful for test generation, refactoring, summarisation, documentation, change explanations and evidence drafting.

Used badly, they can flood an organisation with polished but weakly understood output. The danger isn’t just incorrect code. It’s opaque systems that nobody can properly explain later, plus a false sense of progress because the artifacts look finished.

That is also why older codebases are not automatically a disadvantage anymore. In regulated environments, code that has already been exercised, reviewed, and lived with is often a stronger foundation than fresh output, however elegant that fresh output looks. Proven code carries operational memory.

In the AI era, proof of work gets cheaper.

Proof of use matters more.