On 1 November Proton's chief technology officer, Bart Butler, published an internal engineering memo saying that developers' work has changed: they should direct AI models and agents instead of writing code by hand. It is a statement of direction, not a report of results. The memo gives no productivity figures, no budget and no timeline. Watch: use it as a list of questions about your own pipeline, not as evidence that the approach works.
What it asks of engineers
The memo's own words are blunt: "Your job is to help the LLM write better code faster." Engineers are asked to own how the work gets done, not only the result. The memo names three habits: repository-level context files, shared skills and prompts, and a habit of making repeated failures trigger an investigation and a fix. Engineers are also asked to flag and escalate improvements they cannot make alone. The memo adds that this is not the death of expertise, and that the company will "organize and resource" the investments that are needed.
The part that is more than a slogan
The memo describes two phases. Phase 1 is using AI models to speed up tasks inside the existing workflow, and it says Proton has been "almost completely stuck" there. Phase 2 is rebuilding the workflow itself. The reasoning is that writing code is not the only step in building software: if you make one step much faster, the waiting moves somewhere else. The memo lists where it might move: the speed and capacity of continuous integration (the automated build and test system), design, approvals, specifications and code review. It even says that most human code review will end, with no date.
Picture a kitchen where the cooks suddenly work twice as fast. Unless the pass and the waiters change too, the dishes just pile up at the window. That is the memo's argument for phase 2.
What it does not say
The memo asks why productivity has not risen dramatically if everyone already uses AI models, but it does not measure it. It names no tools, vendors or models, sets no rules for code review or testing, and says nothing about hiring, training or job security. It promises more detail "in the coming days and weeks". Until that arrives, this is one company's intent, written by its chief technology officer.
What to do
Watch. If your team already uses AI to write code, ask the memo's question of your own pipeline: where does finished work wait now, in build capacity, in review, in approvals? Revisit when Proton publishes the promised details or any numbers.