Via theverge.com
The company is pushing engineers to feed its coding agent with real-world fixes as part of a broader plan to have AI generate the vast majority of its code by 2026
Meta is requiring its engineers to submit at least one AI-assisted code fix per week, a move designed to improve the company’s internal coding agent through real-world feedback.
The mandate fits into a much larger, more aggressive AI integration strategy at Meta. The company is targeting 65% of its engineers to write over 75% of their code using AI by the first half of 2026.
The feedback loop strategy #
The company has built a system called MetaMateCR, which automates code review fixes. It has already been tested on tens of thousands of engineers internally. The weekly code fix requirement ensures that engineers are constantly interacting with, correcting, and refining the AI’s output.
The company has also piloted AI-assisted coding interviews, which began in October 2025, with a broader rollout expected sometime in 2026.
The bigger picture: AI replacing engineers, sort of #
Meta’s internal discussions have referenced the possibility of generating up to 95% of its new code via AI.
Hundreds of engineers were laid off between early and mid-2026, with the reductions explicitly tied to AI adoption strategies.
What this means for the crypto and tech investment landscape #
Meta’s AI ambitions require enormous computing power. The company has been one of the largest buyers of AI chips globally, and its push to have AI write most of its code will only accelerate demand for compute resources. That’s relevant for crypto mining operations and decentralized compute networks like Render and Akash, which are positioning themselves as alternatives to centralized cloud providers.
Open-source AI models, which much of the crypto ecosystem relies on, don’t have access to the kind of structured, high-quality training data that comes from tens of thousands of professional engineers correcting code output every single week. This could widen the gap between proprietary and open-source AI capabilities over time, which has implications for any decentralized AI project betting on community-driven model improvement.
Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our