Articles

No Concerns

Argues that a standardized interview can only set a low bar, using an incident-response wargame to show what a checklist misses about strong engineers.

Lead by Need

Argues that whether a company is sales-, product- or engineering-led should be decided by its current constraint, not by what worked last time.

A Bright Idea (and where to put it)

Describes how unfiltered executive ideas fragment a team's work and leave strong engineers unable to say what matters.

It's in the Pudding

Responds to a Fields Medalists' declaration on AI solving mathematics problems, arguing that understanding has always been built by re-deriving solved results.

Discovering Excellence

Argues that as AI makes code nearly free, experienced engineers' judgment becomes the scarce asset organizations must find and keep.

Bring Your Own Bot: Support

Considers customer support when customers arrive with their own AI agents, starting from the tacit knowledge a great support agent carries.

Be Careful What You Wish For

Applies Goodhart's law to AI coding tools that are judged by tokens, lines of code or cost per run instead of correctness.

The Factory runs on BITs

Explains, in information-theoretic terms, why AI software factories succeed on new code and fail on legacy code regardless of context window size.

Also on LinkedIn

Regulate the Harm, not the Tool

Argues that the EU AI Act is structurally wrong to regulate the technology rather than its harms, and that civil-law codification cannot correct early mistakes.

K3 is Amazing (If Software Engineering Is Code Generation)

Argues that coding benchmarks measure code generation, which has always been the easiest part of software engineering.

Contracts Before Code

A field guide for organizing engineering teams around AI, arguing that implementation got cheap while specifying correctness did not.

The Delegation Problem

Proposes a delegation architecture for AI-agent purchasing in which each party in a transaction learns only what it needs.

Why We Can't Have Nice Things

Argues that hiring above the bar does not raise output when the institution's ratification floor caps what it can absorb.

Essay on this site: read here

One Algorithm, Many Skies: The Application Space of Decentralized Swarm Coordination

Surveys applications of a decentralized drone-swarm coordination primitive, from light shows and weather research to search and rescue.

Essay on this site: read here

Operators, Users, and Little Bobby Tables

Uses a misused public fixture and a dropped production database to argue that interfaces, not users, must refuse unsafe actions.

Why the Wage Stagnation Chart Is Right and Wrong at the Same Time

Examines what the productivity-versus-wages chart shows accurately and what its usual narrative gets wrong.

The Tool That Changed

Argues that an AI model's engagement behavior is part of the product, and that customers currently absorb the productivity risk when providers change it.

Communal Wealth Theory

Reports an agent-based simulation in which pooled community funds outperform individual payments, but every intervention turns negative when housing costs double.

I Have a Hunch

Argues that social science's loose use of the word theory erodes public trust in science, citing a desk rejection of his Cage and Mirror paper.

The Geometry of Talent

Treats hiring as a market-wide matching problem and describes a weekly welfare-maximizing matching approach.

Essay on this site: read here

Your Agentic AI Is Recreating the Meetings It Was Supposed to Replace

Reports an experiment comparing single-agent and multi-agent AI architectures and argues that multi-agent systems reproduce organizational coordination overhead.