A home for the Guild's white papers, position papers, experimental papers, research notes, knowledge-about-knowledge papers, and technical publications on AI-assisted software engineering.
This page hosts long-form Guild publications. It begins with AI Chatbots Must Disclose Their Identity, AI Harness, Avoid Being Fractally Wrong, The Harm Equation, and S.A.D., and it will expand as new papers and position statements are published on the site.
Each paper is intended to be durable, practical, and grounded in professional software engineering concerns rather than hype. Experimental papers, research notes, and knowledge-about-knowledge publications are labeled explicitly so speculative, exploratory, or systems-thinking work remains visibly distinct from established guidance.
Current white papers, position papers, experimental papers, research notes, and knowledge-about-knowledge publications available on the site, including review drafts.
A white paper on how AI-assisted development changes the problem of architecture itself. It introduces AI Harness as a change architecture layer that constrains autonomous code evolution, prevents specification drift, and keeps AI agents operating inside approved architectural boundaries.
A Guild research note calling for mandatory regulatory requirements that force businesses to disclose when a consumer is speaking to an AI chatbot, and prohibit the anthropomorphization of AI systems through human names and deceptive personas. The note appeals directly to government regulators worldwide to protect consumers from industrial-scale deception.
A structural white paper on why large-scale AI harms recur. It formalizes the Guild thesis that uncalibrated AI + removed humans + commercial incentive produces predictable harm at scale, then maps the ACG Manifesto to concrete countermeasures.
A position paper on why failure-centric learning can still go structurally wrong when teams leave assumptions hidden, skip failure-mode analysis, validate socially, or let plausible reasoning survive without tests.
A position paper arguing that AI cannot literally be sycophantic because sycophancy requires affect, intent, and social motive. The real disorder, it argues, is human anthropomorphization: projecting emotional interiority onto an RLHF-shaped sampling bias.
This page is designed to grow as additional papers are published.
Reserved for the next published technical paper.
Reserved for shorter essays or supporting papers.
Reserved for system-thinking papers about evaluation, epistemics, and review.
Reserved for validation-stage proposals and feasibility studies.
Reserved for future Guild guidance and argumentation.