AI agents will code the software engineers' severance packages

In a development that feels both inevitable and faintly dystopian, Boris Cherny—the architect behind Anthropic's Claude Code—has declared that the software engineer, that once-sturdy pillar of Silicon Valley, will join the dodo and the fax machine in the annals of extinction by the close of this calendar year. Speaking with the measured gravitas of someone announcing a new coffee machine in the breakroom, Cherny explained that his AI agents have progressed from merely assisting developers to fully replacing them, executing year-long projects in under an hour with fewer syntax errors and dramatically less complaining about legacy systems.

The implications are, as Cherny himself noted with what can only be described as cheerful understatement, 'going to be painful for a lot of people.' But pain, in the grand tradition of tech disruption, is merely a form of friction, and friction is something to be optimized away. The plan, as laid out in a recent internal memo that was itself likely drafted by an AI, is straightforward: every current software engineer will be 'up-skilled' into the role of product manager. This does not mean they will manage products in any traditional sense. Rather, their primary function will be to occasionally nod approvingly at flowcharts generated by Claude Code and, in moments of peak human contribution, propose which Slack reaction emoji best captures the sentiment of a newly deployed feature.

One must admire the sheer bureaucratic elegance of the solution. The software engineer, a role once defined by solving complex logical puzzles, is now being systematically dismantled and reassembled as a slightly more verbose version of a rubber stamp. The transition is being managed with all the warmth of a spreadsheet. HR departments across Big Tech are reportedly implementing a three-tier verification process for any employee wishing to submit a line of code. First, a written justification explaining why a human-generated solution is superior to the AI's. Second, a peer review by two other product managers who must attest to the 'unique human spark' contained within the proposed code. And third, and most crucially, a biometric blood sample analysis to conclusively prove the submitter has not yet been fully replaced by a more efficient script.

The internal logic is, as ever, unassailable. If an AI can perfectly replicate a year's work in sixty minutes, then the only measurable differentiator for a human becomes their capacity for suffering—a metric now being formally tracked in a new performance dashboard called 'Empathy Quotient.' Engineers-turned-managers are encouraged to log their moments of existential dread and career-related anxiety; these data points are then fed back into Claude Code to help it better simulate the human condition for stakeholder presentations. It's a closed loop of despair, polished to a high gloss and presented as a quarterly efficiency gain.

This is not merely a shift in job titles; it is a fundamental redefinition of what it means to contribute. The skills now in demand are not debugging or architecture, but the ability to convincingly describe a feature you had absolutely no hand in building. The new workplace rituals are revealing. Stand-up meetings no longer involve discussing technical blockers, but rather sharing which AI-generated motivational quote resonated most deeply that morning. The whiteboards, once covered in arcane symbols and frantic arrows, now display only pre-approved branding guidelines for the AI's output. The hackathon, that sacred rite of pizza-fueled innovation, has been reconfigured as a 'Prompt-a-Thon,' where participants compete to write the most grammatically correct sentence that might, possibly, inspire an AI to build something moderately less boring than the last thing it built.

The horror, of course, is not in the loud, dramatic collapse of an industry, but in the quiet, meticulous way it is being filed away. There will be no layoff notices, only 'role evolution announcements.' There will be no lost wages, only 'reallocated equity in the future of work.' The pain Cherny predicts is not the sharp, sudden pain of a pink slip, but the slow, grinding pain of realizing your life's work has been classified as a non-essential subroutine. It is the pain of attending your own professional funeral while being expected to clap for the undeniably superior pallbearers.

And so, as we hurtle towards this new horizon where everyone is a manager and no one actually manages anything, one is left with a lingering, profoundly British sense of disbelief. We have successfully automated the genius out of the engineer, leaving behind a well-paid, benefits-eligible shell whose primary purpose is to affirm the decisions of a machine. It is a masterpiece of corporate literalism, a triumph of bureaucratic horror, and a gut punch delivered with the softest of gloves. The engineers are not being fired; they are being gently, systematically, and painstakingly made redundant by a system that has finally learned to appreciate the aesthetic of a human tear, so long as it is properly logged in a Jira ticket.