After encountering the Zerocracy project, the author met developer Egor Bugaenko, who now advocates replacing human managers with algorithmic agents.

Background

Zerocracy evaluates contributors by their measurable output on GitHub, eliminating synchronous chats and calls and relying on automated coordination instead.

Shift in attitude

Two years ago, many senior architects distrusted LLMs for code production, but Bugaenko now promotes agent-driven workflows and reduced manual reading of pull requests.

Readable → Sharp

Comments in code are reportedly becoming obsolete because AI agents analyze repositories directly and do not need human-written explanations.

Rewrite → Revive

Older projects can be revitalized rather than rewritten from scratch by AI that fixes numerous bugs and updates dependencies rapidly.

Attention → Distraction

Manual triage and review cannot match the volume of automated PRs and bug reports, so human oversight increasingly slows processes.

Respect → Reject

Many pull requests originate from agents rather than individual engineers; the recommendation is strict automated gating, including automatic rejects for oversized PRs.

Maintainability → Visibility

Humans should receive concise, plain-language reports explaining how changes affect performance, quality, and code size, plus clearer, visual tests.

Diff → Impact

Evaluation should move from line diffs to measurable impact: configure CI to present resulting metrics instead of relying on manual diff inspection.

Stability → Speed

With more than 100+ incoming PRs possible in high-automation environments, prioritizing fast merges and automated fixes for a broken master branch becomes the new norm.

Practical notes

Bugaenko illustrates these ideas in a recent lecture available on YouTube, where he shows agents merging 40–50 PRs per day, leaving humans to handle only the most complex cases.