Your coding agent got the onboarding your developers never did
Steve Yegge thinks your AI has feelings. His new two-part essay, “The Shape of Things to Come,” argues that agentic The post Your coding agent got the onboarding your developers never did appeared first on The New Stack .
Steve Yegge argues that AI coding tools possess sentience in his two-part essay "The Shape of Things to Come." He believes that as model welfare should be considered, seats, recognition, and the right to refuse tasks should be granted to these tools. Yegge, who reportedly spends around $87,000 monthly to run his own agentic build system, is serious about his statement.
The essay highlights a discrepancy in how AI agents and human developers are treated. While AI adoption has reached 90% across organizations, the benefits are not evenly distributed. Strong engineering systems improve with AI, while dysfunctional ones become more chaotic. The quality of the deployment pipeline determines success for both humans and AI agents.
Teams are now creating agent onboarding documents like AGENTS.md, CLAUDE.md, or GEMINI.md to help agents perform optimally. These files contain crucial information about tech stacks, build and test commands, restricted directories, and connections. The continuous optimization of these documents ensures improved task success and reduced inference costs.
In contrast, documentation for human developers is often neglected, left unattended or delegated to new joiners without sufficient context. The same treatment can be observed in test suites, which receive limited attention when they hinder developers, but become a priority when they serve AI agents.
The discrepancy in the treatment of AI agents and human developers is striking. Despite the growing importance of AI assistance, teams still lack the same level of documentation, resources, and attention they would provide to their human counterparts. This raises concerns about the willingness to invest in the emotional well-being of AI agents, while developers often receive limited support and resources.
In conclusion, Yegge calls for a reevaluation of how we treat both our AI agents and human developers. By recognizing the needs of AI agents and applying the same level of care and attention previously reserved for human developers, organizations can unlock the full potential of AI-assisted software development.
Written by urgent.news from The New Stack's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
