Writing
Making AI Make Sense
A framework for deploying AI where it’s strong and routing around where it’s weak. These pieces build on each other.
Start here: Making AI Make Sense: The complete framework on one page.
Or explore the components:
| Essay | What it covers |
|---|---|
| Knowledge as Capability | Defining knowledge as capability to produce outcomes |
| Agent-Relative Tacitness | Why tacitness depends on the agent |
| Tacit Space Shrinkage | How AI makes previously inexpressible knowledge expressible |
| AI as Oracle vs. Assistant | Choosing between direct answers and verifiable execution |
| The Strong Oracle Trap | Why sophisticated dialogue doesn’t solve the verification problem |
| Executable Knowledge Architecture | The pattern for AI-augmented professional work |
| Capability Governance | The governance framework: stakeholder obligations for generated artifacts |
| Agentic AI as Universal Interface | How the pattern scales: AI removes barriers to direct system access |
| The Universal Interface Thesis | Beyond knowledge work: the pattern applies to any control surface |
| Automating Expertise Gets Easier and Easier | What faces pressure and what remains as more expertise becomes software |
| Ontology Generation | How organizational knowledge accumulates from universal interface logs |
| What Benchmarks Aren’t Measuring | Why current AI benchmarks test the wrong things for professional work |
| AI-First Software | Designing software around AI |
| The Return of the Assistant | AI delegation combines autonomy with additional execution capacity. |
| Update for February 2026 | Two years of advances support the distinction; agentic AI creates new governance gaps. |
| Answers to Critics | Direct responses to adversarial critique: what’s refined, what’s defended, what’s unchanged. |
Updates and Responses
The framework has been tested against critique and technological change. These essays document the evolution.
| Essay | What it covers |
|---|---|
| Update for February 2026 | Research validating the oracle/assistant distinction against 2024-2026 advances; the agentic erosion problem |
| Answers to Critics | Direct engagement with adversarial review: refinements, precision improvements, and defended positions |
Standalone Essays
Other writing, not part of the framework above.
| Essay | Topic |
|---|---|
| Parrots Are All You Need | Why “stochastic parrot” isn’t the insult people think it is |
| LLM Stochasticity and Determinism | Understanding the randomness in language models |
| The Consulting Threat and Opportunity | What AI means for professional services |
| Critical Thinking Rules | Principles for clear reasoning |
| The Support Team’s Real Job | Reducing overhead for professionals |
| Leadership Lessons from Science Fiction | What SF teaches about leading |
| The Remote Work Formula | Making distributed work work |
| AI Table Extraction Comparison | Testing AI on a specific task |
| The Office Availability Math | Why 80% in the office means 40% available |
Defining concepts and core distinctions
- Knowledge as Capability: A Better Foundation for Human-AI Cooperation An operational definition of knowledge as the ability to produce outcomes, and why that makes human-AI cooperation easier to reason about.
- AI as Oracle vs Assistant: The Choice That Determines Whether You Can Govern It The most important AI deployment choice is whether you ask for answers or execution. One creates verification problems; the other makes governance possible.
- Why 'Stochastic Parrots' Are More Powerful Than Critics Admit Why the 'stochastic parrot' critique misses the practical significance of generative AI for knowledge work.
The central frameworks
- The Universal Interface Thesis: Why AI May Replace Traditional Software Interfaces AI-generated code may become the universal interface for knowledge work, software, machines, and complex control systems.
- The Return of the Assistant: Why AI Beats DIY Knowledge Work AI assistants let knowledge workers delegate execution while retaining the autonomy they gained from doing the work themselves.
- Why Automating Expertise Gets Easier as AI Improves As AI lowers the barriers to implementation, more expertise becomes software. This changes where value holds and where it moves.
- Ontology Generation: How AI Logs Turn Work Into Organizational Knowledge When AI becomes the universal interface, interaction logs can reveal both what an organization does and the concepts it uses.
- Agentic AI as Universal Interface: Why Expertise Becomes Software Agentic AI removes two barriers that kept expertise from becoming software: coding skill and system access knowledge.
- Executable Knowledge Architecture for AI in Professional Work A framework for integrating AI into professional work while preserving expertise, improving reproducibility, and closing the governance gap around AI-generated output.
- What AI Benchmarks Miss About Real Professional Work Benchmarks like GDPval measure well-specified tasks with verifiable answers. They miss the ambiguous, judgment-intensive work where professional expertise still matters.
- Capability Governance: How to Govern AI-Generated Artifacts A practical framework for governing AI-generated artifacts through defensibility, exposed methodology, empirical track record, and clear stakeholder obligations.
- The Strong Oracle Trap: Why Better AI Dialogue Still Does Not Solve Verification Thorough AI dialogue can improve outputs, but it still does not solve the verification problem in professional work.
Extending the foundations
- Tacit Space Shrinkage: How AI Makes More Expertise Explicit As AI improves, more knowledge can be expressed, operationalized, and reused. That changes how organizations think about expertise and governance.
- Agent-Relative Tacitness: Why Some Knowledge Is Tacit for Humans but Not for AI Tacit knowledge depends on which agents can express and transfer it, and under what conditions.
Evidence and experiments
- The Office Availability Math: Why In-Person Work Helps Less Than You Think Two people in the same office are available to each other far less than most assume. The math explains why.
- Why LLMs Seem Random but Still Produce Reliable Results Results from 700 runs across seven models show why direct answers stay inconsistent while executable methods can still produce reliable outcomes.
- AI Table Extraction Comparison: Claude vs Docling vs ScaleDP A practical comparison of Claude, Docling, and ScaleDP for AI table extraction, with accuracy gaps that matter for document processing at scale.
Applying the framework
- AI-First Software: What Changes When Language Becomes the Interface Generative AI changes software design by making language a usable interface, shifting applications from fixed screens toward on-demand interaction.
- How AI Threatens Consulting and Creates New Opportunity AI-encoded expertise puts pressure on consulting's traditional business model, but also creates new opportunities for firms that adapt quickly.
Lessons from experience
- EKA in Practice: What Happened When We Deployed the Framework What we learned by deploying Executable Knowledge Architecture in practice, including where the framework held up and what changed.
- Practical Rules for Critical Thinking A practical set of rules for critical thinking across foundation, analysis, and engagement, written for non-academics.
- Leadership Lessons from Science Fiction Books The leadership lessons that most shaped how I lead came less from management books than from science fiction.
- The Remote Work Formula: Flexibility, Autonomy, and Team Responsibility A simple way to think about remote work that balances individual flexibility with responsibility to the group.
- What Support Teams Actually Do in Professional Services The real job of support teams is to protect the scarcest resource in professional services: professional attention and time.
Research and changes to the framework
Responses to criticism