Overview Page

COAST & SINK: A Methodology for High-Stakes Knowledge Work The limiting factor in the age of AI is no longer access to information. It is whether you can trust the thinking that produced your conclusionsand whether others can trust what you share. When explanation, critique, and synthesis are available on demand, speed becomes dangerous. The new risk is not ignorance, but epistemic capture: the slow surrender of your judgment to tools that agree with you too smoothly, and the flattening of complex ideas into persuasive performance. COAST and SINK are two complementary practices designed for this environment. They form a complete protocol for knowledge work: one governing how ideas are produced, the other governing how they are shared without loss of rigor or agency.

The Problem: Speed Without Accountability

Most approaches to “working with AI” focus on tactics: better prompts, faster outputs, automation of tasks. These methods tend to collapse thinking into generation and evaluation into trust. The result is fluency without understanding, and consensus without verification. When the stakes are highpolicy decisions, strategic analysis, complex system designthis collapse is catastrophic. You need a method that preserves intellectual sovereignty: the ability to stand behind your conclusions because you have seen how they were formed, tested, and stressed. COAST and SINK begin from a single premise: AI should extend human thinking, not replace it. That requires two disciplines: a way to think with AI without surrendering authorship, and a way to share ideas in an AI-mediated world without turning them into persuasion or noise.

COAST: The Production Discipline

COAST is a five-rule framework for producing ideas with AI while retaining accountability. It treats AI as a disciplined tool for synthesis and adversarial stress-testing, not as a co-author or oracle. C Clarity Before Proliferation Use AI only when it reduces cognitive load. If a query doesn’t sharpen your thinking, it is noise. O One Primary, Many Adversaries Designate one AI instance for continuity (your “Synthesizer”). Use “foreign” AIscleared sessions, different models, instances with no contextstrictly to attack your conclusions. A Adversarial Engagement is Discipline When a conclusion feels too easy or too perfect, submit it to hostile review. Instruct the foreign AI to find the fatal flaw, not to polish your prose. The critique is data, not verdictbut you must face it before you finalize your position. S Save Transcripts Log every interaction that materially influenced your thinking. These are not chats; they are audit trails. When a director asks, “Why did we decide this?” the transcript is your firewall against blame. T You Are Still the Thinker Own the framing, the conclusions, and the accountability. AI extends reasoning; it does not replace responsibility. COAST is slow by design. It introduces friction where fluency would otherwise mask error. The outcome is not a polished answer, but a conclusion you can genuinely defendbecause you have witnessed its formation under pressure.

SINK: The Distribution Protocol

SINK (Self-Interrogation of Networked Knowledge) is a method for sharing complex ideas once they exist. It assumes that your readers now approach material with AI at their side, and that this changes how knowledge must be packaged. Instead of asking readers to absorb long papers linearly, SINK distributes ideas as structured, interrogable systems. These summaries function like source code: bounded, readable, and designed to be executed through questioning rather than absorbed passively. Narrative Scaffolding SINK documents use clear prose for human readers and semantic anchors (declarative headers, consistent terminology, explicit relationships) for AI indexing. The AI can map the logic; the human retains the causality. The Bundle & Map Protocol Under SINK, users provide their AI with the entire corpus of related summaries at once. The START_HERE document functions as a semantic kernela map that teaches the AI how to triangulate answers across the full bundle, rather than summarizing documents in isolation. Interrogation Over Reading Readers are encouraged to question assumptions, surface tensions, and explore consequences immediately. Because the AI holds the full context, interrogation moves laterally across documents. Understanding emerges through stress-testing, not skimming. Knowledge Sovereignty The AI explains; the reader decides. No conclusion is authoritative by default. SINK makes disagreement safe and productive by shifting control to the reader, who retains full ownership of what they accept, reject, or revise.

Why They Are Designed as Complements

COAST and SINK address different phases of the same problem.

• COAST governs creation: how ideas are generated, tested, and stabilized in the presence of AI. It minimizes error and self-deception during production.
• SINK governs distribution: how ideas move into other minds without being flattened, distorted, or weaponized. It minimizes entropy during transmission.

Used together, they form a closed loop: COAST produces material worth interrogating. SINK ensures that interrogation remains possible for others without recreating the entire process. Neither method requires the other to function, but each becomes more powerful when paired. The result is a practice that favors understanding over speed and interrogation over persuasion.

What This Isand Is Not

COAST and SINK are intentionally modest in scope.

They are not:

• A productivity system or automation framework
• A collaboration platform or governance model
• A replacement for primary research or domain expertise
• A guarantee of consensus or correctness

They are a way of working that assumes curiosity, good faith, and personal responsibility. They do not prevent misuse, but they offer a better default: one that preserves the integrity of thinking in an environment optimized for consensus and fluency.

Who This Is For

These methodologies are designed for high-stakes knowledge work: strategists, policy analysts, systems architects, researchers, and anyone who must reason under uncertainty and defend their conclusions to others. If you have found that AI makes decisions feel easier but less robust, or that your written work is increasingly met with passive agreement rather than substantive engagement, these practices are built for you.

The Central Claim

As AI becomes ubiquitous, the quality of our thinking will depend less on what tools we use and more on the structures we adopt around them.

COAST and SINK propose such a structure. They do not tell you what to think. They give you a way to thinkand a way to let others think with youwithout losing yourself in the process.

Begin with the bundle, or begin with the transcript. But begin with accountability.