Proton CTO's Internal AI Manifesto: The Old Ship Is Gone

Proton CTO Bart's AI manifesto was obtained in multiple copies by 404 Media's Joseph Cox. Matt Stockton reposted it on X and recommended that anyone leading an engineering organization read it in full. Proton makes end-to-end encrypted email and VPNs; this internal text is about how software engineering itself is changing, and no single feature is the focus of the discussion.

One line in the manifesto is very blunt: the era of manually writing high-quality code is over. LLMs and accompanying tools have changed the way code is written. Repetitive work is reduced, efficiency increases, and new questions arise around how teams adapt and how to put the tools to use. The engineer's role is pushed into a leadership and management position: guiding LLMs to work efficiently while continuing to learn and improve their skills.

Matt Stockton's judgment is even more direct: the old ship is gone. Many people still believe it exists emotionally and in terms of their interests. Organizations need to change as a whole; this is a human alignment problem. A leader's task is to get the team onto the new ship and sail together. The sooner this is understood, the less painful it will be in the long run; if you don't act, your organization will be disrupted by organizations that do.

The manifesto's conclusion is also direct: programming is dead, engineering is not dead, and expertise is not dead. Old jobs disappear, and new jobs let builders do more and have greater impact. Proton will provide support, with more actions in the coming weeks. The experimental period continues, but the time to act is now.

There are disagreements in the discussion. Some say the memo is reasonable and forward-looking; others ask whether a privacy company now wants people to trust a bunch of vibe-coding hackers. Some question why it took the CTO four years to realize something obvious; others think this isn't a manifesto at all, just an acknowledgment of reality. One engineer mentioned that faster coding is only the first step; every part of the SDLC needs to change, and end-to-end testing matters more, because when you don't read or write unit tests, you need higher-level tests to verify behavior. Another commenter said that software engineers in the new era are essentially yesterday's engineering managers: observing, guiding, and questioning processes and results.

Marcus Gröber's challenge is worth including: if the direction is right, why haven't we seen those disruptive organizations appear in the wild? Many people claim disruption, but the externally visible results are few. This counter-question is more useful than slogans.

Matt Stockton's tweet: https://x.com/mstockton/status/2116548727473148203

404 Media original article: https://www.404media.co/manifestos-fear-and-loser-thinking-inside-protons-ai-reckoning/

Joseph Cox's quoted tweet: https://x.com/josephfcox/status/2115868430897385497