Proton's AI manifesto for software engineering

Proton’s engineering manifesto argues that AI has changed software development so fundamentally that manually writing code is no longer an engineer’s central job. Language models can take on boilerplate, documentation, testing, and parts of design, reducing work that has traditionally consumed time and been neglected. Engineers should instead direct AI agents, expand what they can handle independently, and improve shared resources - such as repository guidance and reusable prompts - so repeated mistakes are investigated and prevented. This shift does not make human expertise or sound engineering obsolete: people still provide judgment, catch misleading outputs, and take responsibility. Documentation, maintainable architecture, and tests remain essential, though practices may change as AI makes some tasks cheaper.

The manifesto says Proton has yet to realize AI’s full productivity gains because it is mostly using models to accelerate existing tasks. The next step is to redesign the software development lifecycle as a whole: as coding speeds up, constraints may move to CI capacity, design, approvals, specifications, or code review. AI could also make previously unaffordable automation practical, but progress requires identifying bottlenecks and rethinking workflows rather than optimizing isolated tasks. That transformation is a company-wide coordination challenge, not something individual engineers can accomplish alone; Proton must provide organization, resources, and measurement tools. The urgency is strategic: companies that adapt slowly risk falling behind, while engineers who engage with the transition can help shape it.