Software ephemerality

The idea that AI-generated software should be treated as disposable and replaceable, in contrast to the business context and institutional knowledge that underlies it, which constitute the genuinely durable assets of an organization.

Core concept

Tom Blomfield articulates the principle directly: the software itself is not the valuable artifact — rather, "the business context and skills are the valuable part. I think the software on top of it is ephemeral." His practical implication follows naturally: organizations should "treat the software as ephemeral. You can generate it. You can regenerate it." How to Build a Self-Improving Company with AI, 11:34

This inverts a conventional assumption in software development — that codebases represent significant accumulated value and should be carefully maintained and extended. Under the ephemerality framing, the code produced by AI is closer to a temporary artifact: worth generating on demand, but not worth preserving at the expense of clarity or correctness, since it can simply be regenerated.

Implications

The locus of competitive advantage shifts, under this view, away from the code itself and toward the organizational knowledge required to specify, evaluate, and deploy that code effectively. A company's ability to articulate its domain precisely — its requirements, constraints, and judgment about what "good" looks like — becomes the scarce resource that AI cannot simply regenerate. The software is the output; the context is the asset.