Proton's AI manifesto was published in full on its official blog on November 1, 2026. Its author is chief technology officer Bart Butler, and the memo was written about a month earlier. Its judgment is direct: software engineering has changed irreversibly, engineers no longer have hand-writing code as their core work, and they must instead lead models and agents to write code better and faster. This is not a product announcement, but a public statement from a company known for privacy services about how it is reorganizing engineering work internally.
The engineer's new job: team lead for AI
Butler quotes a senior engineer as saying, "No one writes code anymore." He describes writing great code by hand as "a core skill whose time has passed." The new job is to help large models write better code, faster: every engineer is now a team lead and a manager, and their direct reports are the models and agents they use, which they must let work autonomously; when a model cannot solve a problem on its own, they must not only solve it this time, but also leave behind repository-level context, shared skills and prompts, so that repeated failures trigger investigation and fixes. The manifesto stresses that this is not the end of expertise: models are fast, while people are slow but bring experience and judgment, are less easily misled by hallucinations, and remain accountable for results, while documentation and testing remain essential.
The bottleneck is not writing code, but what comes after
The manifesto locates the problem in "phase 1": merely speeding up existing workflows, so the bottleneck moves to continuous integration, design, approvals, specifications and code review. Butler writes that the company needs to reach, and will reach, a point where most human code review ends. When Claude Code Review became generally available, reviewing a PR took just 6 to 10 minutes, but who is accountable for the result remains an open question. Trade-offs in testing must also be recalculated: as testing becomes cheaper, old limits on excessive end-to-end testing may move. Collective action is another obstacle: it is easy for an individual to use a large model well, but rebuilding a whole company's processes takes organization and resources, which Proton has underinvested in, to the point where it does not even know where the bottlenecks are. This reorganization is "phase 2", and the manifesto demands entering it now.
What the manifesto does not answer is exactly where the dispute lies
The manifesto first draws its own boundary: it does not discuss the details of security, reliability and infrastructure, and only points to a direction. It does not say how security review should change or who is accountable for it, nor does it offer a way to measure the effectiveness of AI code review. On October 29, 2026, 404 Media reported a dispute inside Proton over its AI push, citing employee interviews and internal records, and saying that chief executive Andy Yen had called an employee request for "a few days of penetration testing" "bullshit and loser thinking", and had written that 2026 would be the last year of hiring engineers; the report also said Butler had said that 90% of October's code was written by AI, but gave no common definition and the figure had not been independently verified. 404 Media said Proton had not disputed the report; the claims above come from investigative reporting, not from the manifesto itself, and the reporting did not say that a product data breach had occurred. Proton said the concrete steps for advancing "phase 2" would be announced in the coming days and weeks, and only then will it be possible to judge whether this manifesto changes how work is done, or remains only words.
