Software Ecology
The holistic study of social and technical systems that produce software.
Software Ecology (Adam Bender, Google I/O 2026) is the study of socio-technical ecosystems — the humans, tools, processes, incentives, and cultural norms that together determine what gets built and how.
The framework asks two questions of any team or process:
- WHY? — Why do we test this way? Why this language, this review process, this deploy cadence?
- WHAT IF? — What if AI wrote all the code? What if we removed this review layer entirely?
It explains why identical AI tools produce wildly different results across teams: the tools are the same; the ecology differs. At 10× AI velocity, five second-order effects dominate before benefits:
- More code → more liability, not just more productivity
- 10× code → potentially 100–1000× test cases (quadratic growth)
- Code review becomes a bottleneck — tech leads can't review for 10 AI developers
- Internal APIs become de-facto public — agents find and call them without negotiation
- Agentic edit wars — one agent changes, another reverts; you pay tokens for both
In plain terms
A forest ecologist doesn't study trees in isolation — she studies soil chemistry, species interactions, water cycles. Software ecology says code needs more than smart developers: it needs the right culture, process clarity, and incentive alignment.