Laura (@poteto), a core team engineer for React Compiler who previously worked at Cursor, Meta, and Netflix and now leads the Grok Bot team at xAI/SpaceX, published a manifesto on October 30, 2026. She argues that with the proliferation of AI agents, code will arrive in a “continuous stream,” requiring most software to be rebuilt to adapt to a new, faster pace. The post was a response to a 404 Media investigation into Proton’s internal policies and garnered approximately 287,000 views within 24 hours.
What happened
On October 30, 2026, Laura (@poteto) published a manifesto post on X. Its core argument: due to AI agents, code will arrive in a “continuous stream,” like a raging river, meaning most software will need to be rebuilt and rewritten to adapt to a new, faster pace. According to her description, the codebase is becoming an “organic architecture”: agent contributions grow around constraints and guardrails designed by humans. This leads to her main call to action — “make the easy path the right path”: designing systems, frameworks, and languages that tame the chaos of agent-generated code. The catalyst was a 404 Media investigation into Proton’s internal policies, published the day before. The author later noted separately that she received many responses regarding maintainability and clean code.
Context
The manifesto emerged against the backdrop of Joseph Cox’s investigation in 404 Media, “Manifestos, Fear, and 'Loser Thinking': Inside Proton's AI Reckoning,” published on October 29, 2026. According to the publication, Proton CEO Andy Yang banned engineers from “loving and hand-writing code,” declared the current year the last for hiring traditional engineers, and in an internal security Slack channel, the call to conduct pentesting before deployment was labeled “loser thinking.” The manifesto’s author is not a random commentator: she works on the core React Compiler team, previously at Cursor, Meta, and Netflix, and now leads the Grok Bot team at xAI/SpaceX, so her text is read as the opinion of a practitioner from the heart of agent-based development. Her principle of “making the easy path the right path” essentially returns to classic production engineering: reliability in a continuous stream of code is achieved through designed constraints, not heroic review of every change.
Why this matters for the industry
For the industry, the manifesto marks a shift in engineering paradigm: when agents generate code in a continuous stream, competitive advantage shifts from the “cleanliness” of individual functions to designing constraints, guardrails, and frameworks in which agents — including those with minimal context — do not break the system. Writing code becomes cheap, while verification and environments where the agent stream does not destroy the product become the scarce resources. The principle “make the easy path the right path” reads as a ready-made product roadmap: packaging rules, boundaries, and automatic checks into infrastructure — agent-oriented APIs, skills, and “agent-friendly” codebases. If the trend of agent generation persists, de facto standards for “agent-friendly” repository structures are likely to emerge, along with increased demand for policy-as-code tools, automatic review, test generation, and pre-deployment security scans — as a separate category alongside AI coding assistants. The resonance is already affecting practices: teams actively using agents are re-evaluating repository structures and verification pipelines, although this has not yet created paying demand for “new platforms” — the shift is happening in practices and hiring, not budgets.
Why this matters for users
The practical takeaway for readers working with code-writing agents: design your project structure in advance so that the “easy path” for the agent aligns with the right one. This includes clear boundaries and guardrails, documentation for agents, and automatic checks — tests, CI, and evals on agent diffs; project rules are best formalized in documentation that the agent reads. The second takeaway is a warning from the Proton case: pressure for speed for revenue can suppress security checks, so when evaluating AI products, it is worth asking how the team verifies code, and in your own process, do not cancel security stages before deployment even under deadline pressure.
What is still unknown / limitations
The thesis that “we will need to rewrite most software” is a hypothesis from one practitioner, not a research conclusion: the manifesto lacks methodology, measurements, and benchmarks, and contains no data on the share of agent code, regression frequency, or support costs — only analogies with large tech companies that have already managed code avalanches due to team growth. The author’s experience in React Compiler and leadership of the Grok Bot team add weight to the opinion but allow for observer bias. There is also a technical caveat: large teams produced redundant code with independent errors from different people, while continuous agent generation may produce correlated systematic errors, inheriting biases from the model and training data — the difference in the nature and cost of such failures has not yet been measured. The Proton case is known from the 404 Media investigation as an organizational decision without published quality metrics, so there is currently no basis to verify its effectiveness. Finally, the manifesto is not a tool but an editorial text: it contains no API, pricing, or latency data, and forecasts about “agent-friendly” infrastructure standards are interpretations valid only if the trend of agent generation persists.
Sources
- lauren (@poteto) on X: “we will need to rebuild and rewrite much of software to adapt to a new, faster pace”
- Joseph Cox, 404 Media: “Manifestos, Fear, and 'Loser Thinking': Inside Proton's AI Reckoning”
- lauren (@poteto) on X: on the wave of responses about maintainability and clean code
Author
Look at AI, editorial team