Terry Li

Exoskeleton, Not Colleague

Thu, 19 Mar 2026 09:00:00 GMT

The AI governance conversation is stuck in the wrong frame. The pattern that works isn't autonomous agents — it's exoskeletons. Micro-agents handling narrow tasks, with human judgment at every point that matters.

Your Check Passed on the Proxy

Sat, 25 Jul 2026 10:30:00 GMT

A green result is a true statement about a stand-in. Whether that stand-in resembles the thing you care about is a separate question, and the check cannot answer it.

Near-Parity Is Mostly a Claim About Formatting

Sun, 26 Jul 2026 00:00:00 GMT

On the benchmark built to measure AI against real professional work, the best model scores 48% on deliverables that come in a file and 14% on deliverables judged as prose alone. That 34-point gap is the largest single effect in the paper, and it is not in the limitations section.

The Number Checks Out. The Claim Doesn't.

Sun, 26 Jul 2026 00:00:00 GMT

The most-quoted figure about AI and professional work is accurate, in the paper, and correctly transcribed. It is also a description of a scenario nobody would choose, because what travelled was the number without the condition it was computed under.

Your Reviewer Model Is Not Independent

Sun, 26 Jul 2026 00:00:00 GMT

When agents generate faster than anyone can read, the standard answer is a second model reviewing the first. Aerospace decomposed what makes a check independent decades ago, and a reviewer model fails the hardest of the three tests.

270 Agents While I Slept

Fri, 20 Mar 2026 01:00:00 GMT

I ran an autonomous agent loop overnight — 43 waves, ~270 dispatches, ~250 vault files produced. Here's what I learned about building systems that work while you sleep.

4 Principles for Agent-Facing CLI Design

Tue, 07 Apr 2026 02:35:07 GMT

Most advice about making CLIs agent-friendly is just good CLI design. Only four principles are actually agent-specific.

A Garden Is Not a Changelog

Fri, 15 May 2026 11:50:39 GMT

Recent work only becomes public writing when the claim survives the removal of the event that produced it.

A Persona Is Not a Control

Sun, 17 May 2026 05:55:08 GMT

Assigning roles to AI agents can look like governance. It only becomes useful when the role has a loss function, an evidence boundary, and an output contract.

A Skill Is Not a Prompt

Tue, 02 Jun 2026 22:55:00 GMT

The useful unit in agent systems is not a better instruction. It is a tested capability package: judgment, code, checks, routing, and boundaries.

Split on Access Control, Not Abstraction

Wed, 25 Mar 2026 09:30:00 GMT

Repo boundaries enforce access control, not abstraction. Directories handle abstraction. If two things have the same visibility requirement, they belong in the same repo.

Act-on-Receipt: The Third Task Class

Fri, 13 Mar 2026 03:28:00 GMT

Most task systems are binary, but a third class exists — tasks triggered by external notifications — and managing them like a backlog item is the wrong move entirely.

After Automation, Judgment Becomes Infrastructure

Fri, 22 May 2026 09:50:53 GMT

When execution gets cheap, the scarce work moves to framing, review, and the systems that preserve judgment.

After the Harness

Tue, 26 May 2026 10:28:05 GMT

Once model companies supply the generic agent harness, the valuable work moves into workflow design, human intervention, domain data, and the definition of good work.

The missing layer between model risk and application security

Wed, 06 May 2026 07:50:00 GMT

Model risk reviews the model. Application security reviews the application. Neither sits behind the agent at execution time, watching the verbs as they go out.

AX: Agent Experience Is the New DX

Fri, 06 Mar 2026 11:30:00 GMT

Developer experience became a competitive moat in the API era. Agent experience is next. Most tools aren't designed for it yet.

Agent-First CLI Design: TTY Detection as Philosophy

Fri, 06 Mar 2026 17:30:00 GMT

The primary user of my CLI tools isn't me anymore. Designing for that changes everything about how output should work.

Agent-Native Onboarding Is Not a Signup Form

Sat, 23 May 2026 03:45:00 GMT

If a product wants agents as real users, first-run setup has to be an executable workflow, not a human signup ceremony wrapped in documentation.

Agentic AI in Production Looks Like a Workflow

Fri, 13 Mar 2026 03:27:55 GMT

The gap between 'agentic AI' hype and what actually ships in production turns out to be a workflow — and that's a feature, not a failure.

Agentic Engineering: Why Less Is More

Fri, 06 Mar 2026 23:00:00 GMT

Tool enthusiasm is often net-negative. Context pollution degrades performance faster than features improve it. The principles that actually work.

Agentic Search Ate RAG

Tue, 17 Mar 2026 08:26:38 GMT

When AI agents can grep, read, and reason iteratively, most RAG infrastructure becomes unnecessary middleware.

The AGI Question Nobody Is Asking Correctly

Fri, 06 Mar 2026 15:00:00 GMT

Sequoia says AGI is here. Dan Shipper says we're not there yet. They're both right — they're measuring different things. The question that actually matters is Sholto Douglas's "nines of reliability."

AI Agent Frameworks for Enterprise FS: What Actually Works vs. Hype

Fri, 06 Mar 2026 09:30:00 GMT

Most enterprise AI agent pilots in financial services fail at the same point: the second tool call. The problem isn't the framework.

AI Agent Teams Are Colonies, Not Companies

Fri, 20 Mar 2026 06:25:04 GMT

The right organisational metaphor for AI agent teams isn't a company with managers and reports — it's a colony with autonomous workers responding to coordination signals.

AI Agents Need Notebooks, Not Just Memories

Fri, 13 Mar 2026 04:58:59 GMT

The missing layer in enterprise AI isn't smarter models — it's structured memory that humans can actually review.

AI Agents Need Passports, Not Passwords

Mon, 27 Jan 2025 00:00:00 GMT

The authentication systems we're building assume AI agents are tools. What happens when they become economic actors with their own accounts, credentials, and legal standing?

AI Controls Architecture

Sat, 18 Apr 2026 13:09:00 GMT

Risk teams know risk. The open problem is designing controls for systems that are non-deterministic, probabilistic, and attackable in natural language.

AI Succeeds, Economy Breaks: The Displacement Loop Nobody Models

Fri, 06 Mar 2026 19:00:00 GMT

The standard AI economic models assume wage effects and retraining timelines. They don't model the feedback loop where successful AI deployment reduces the customer base that purchases AI-enabled products.

Five Archetypes of AI-Era Business Defensibility

Fri, 06 Mar 2026 14:00:00 GMT

When AI models commoditize, the moat isn't the model. It's the infrastructure AI must flow through but can't replace. Five archetypes of what that looks like.

AI Evals: Why Teams Build Metrics Before They've Read a Trace

Fri, 06 Mar 2026 21:00:00 GMT

Most teams build evaluators before reading a single trace. The sequence that actually works is the opposite: observe, categorise, then measure.

AI Fixed My Perfectionism (Sort Of)

Sat, 07 Mar 2026 04:53:00 GMT

On why the blank page stopped being the hard part.

AI Governance Category Error: Routing vs. Compliance

Fri, 06 Mar 2026 12:00:00 GMT

Your AI governance framework is a routing spreadsheet pretending to be a compliance programme. Regulators will spot the difference.

The AI Job Title Illusion

Fri, 06 Mar 2026 14:30:00 GMT

Two job ads. Same bank. Same week. Same title pattern. Completely different jobs. The AI hiring market has a labelling problem.

The Risk Tiering Gap in Banking AI

Thu, 19 Mar 2026 12:00:00 GMT

Banks have AI ethics principles. They don't have risk tiering. That's the gap that matters.

AI Vendor Selection Is Now a Values Decision

Fri, 06 Mar 2026 20:30:00 GMT

OpenAI took the Pentagon contract Anthropic refused. Your AI vendor just became a political statement — and enterprise procurement hasn't caught up.

AI Vendors Are Not Neutral Infrastructure

Sat, 14 Mar 2026 09:09:08 GMT

The DoD-Anthropic dispute reveals a new category of operational risk: foundation model vendors can unilaterally revoke access based on their own values, not just SLA violations.

Always latest as a system property

Sat, 04 Apr 2026 07:18:08 GMT

Most projects pin dependency versions to avoid breakage. We automated the opposite: daily upgrades with automatic rollback.

Ambient Agents: When AI Disappears Into Capability

Wed, 27 Aug 2025 00:00:00 GMT

The real revolution isn't making AI smarter - it's making it invisible

Assume the LLM never ran

Sat, 11 Apr 2026 13:26:34 GMT

A 208 MB log, 59,356 retries, and zero LLM calls. A debugging story about what happens when the symptom lies about the cause.

Autonomy Starts at the Check

Sat, 16 May 2026 04:59:45 GMT

An agent is not autonomous because it can try a task. It is autonomous when the system can tell whether the task worked.

Autopoiesis

Wed, 25 Mar 2026 08:27:08 GMT

The defining property of life is not metabolism or reproduction -- it's autopoiesis. A system that continuously produces and maintains itself. That's the north star.

Backtest vs Operational Validation: The Control You Think You Have

Fri, 06 Mar 2026 19:30:00 GMT

A model control that's never fired in production isn't a control — it's a hypothesis. The gap between backtest and operational validation is invisible until someone asks.

Banking DS to AI Consulting: What the Transition Actually Teaches You

Fri, 06 Mar 2026 08:30:00 GMT

The operational instincts built in production banking don't belong in the past. They're exactly what makes a practitioner-turned-consultant useful.

Banks Have an AX Problem They Don't Know About Yet

Fri, 06 Mar 2026 16:00:00 GMT

Banks are building AI agents to call their APIs. Those APIs weren't designed for agent callers. The mismatch is subtle, consequential, and almost nobody is talking about it.

Most Banks Don't Need an AI Strategy

Fri, 06 Mar 2026 08:00:00 GMT

The real project isn't artificial intelligence. It's the data infrastructure that AI exposes as broken.

Benchmark Your Research Stack

Mon, 09 Mar 2026 16:00:00 GMT

Running 10 real queries through 5 tools revealed that theoretical routing rules have systematic gaps — and the surprises were more useful than the confirmations.

Biology as a Design Constraint: How Cell Biology Names Generate Architecture

Wed, 08 Apr 2026 00:00:00 GMT

Using cell biology naming not as metaphor but as engineering manual — how mTOR's biology predicted circuit breakers, autophagy, and negative feedback loops before we designed them.

Bridge or Seed

Mon, 23 Mar 2026 11:47:47 GMT

Every skill you build is one of two things. Knowing which changes what you build next.

Budding

Wed, 25 Mar 2026 08:56:00 GMT

Every other component in the organism had a biological name. Agents didn't. Three failed attempts and a yeast cell later, they do.

Building a Bus Alert System in One Session

Sun, 15 Mar 2026 14:40:08 GMT

How a real need on a Hong Kong bus turned into a GPS-powered alert system in under two hours

Building My Own Consulting Toolkit Before Day One

Fri, 06 Mar 2026 17:00:00 GMT

Most consultants arrive at a new firm and learn their tools from colleagues. I tried something different.

Building porin: a library for agent-facing CLIs

Mon, 06 Apr 2026 14:00:00 GMT

I turned the seven patterns into a zero-dependency Python library. Then I added MCP bridge support. Here's what I learned about the gap between patterns and code.

The Case Against Knowledge Management Systems

Wed, 11 Mar 2026 08:10:00 GMT

Most PKM tools are procrastination with better aesthetics. The problem isn't the software — it's that filing a note feels like understanding it.

Cast the Wide Net

Mon, 16 Mar 2026 10:32:23 GMT

When you don't have enough information to narrow, stop narrowing.

China's AI Stack Is Now Hardware-Deep

Sat, 14 Mar 2026 09:09:49 GMT

DeepSeek V4 launching on Huawei Ascend NPUs signals that China's AI ecosystem is decoupling at the silicon layer — deeper and more durable than model-level divergence.

Choosing a Steam Iron with Vertical Steam

Sun, 22 Mar 2026 01:30:54 GMT

A practical guide to picking a steam iron that handles shirts, steams hanging clothes, and doesn't weigh a ton. Cordless and corded options compared, with Hong Kong Consumer Council test data and current pricing.

Claude Code, Analyze My Spending

Thu, 14 Aug 2025 10:00:00 GMT

When AI coding assistants become workflow orchestrators, the most powerful compiler processes reality, not code.

Claude Code Mobile is Better Than Desktop

Thu, 07 Aug 2025 10:00:00 GMT

Walking meetings, voice input, and location changes unlock cognitive advantages desktop workflows can't access.

Claude Code is Not a Coding Agent

Sat, 26 Jul 2025 10:00:00 GMT

Why I use Claude Code for everything except coding: cognitive compiler for strategy, decisions, and understanding.

CLIs Enforce Structure and Save Tokens — Not Just Discipline

Fri, 06 Mar 2026 22:00:00 GMT

The instinct to add a rule to a skill file is usually the wrong abstraction. A CLI wrapper enforces at the tool level: zero deliberation, zero token cost.

CLI, MCP, or code mode: the answer depends on who's running the sandbox

Sun, 05 Apr 2026 10:39:47 GMT

Willison says CLIs beat MCP. Cloudflare says server-side code mode beats both. They're both right, because they're answering different questions.

Composure Is a Skill

Mon, 16 Mar 2026 04:53:07 GMT

Rushing is a habit, not a response to reality. You break it by deliberately not rushing when you could.

Compounding: The Only Mental Model

Mon, 16 Mar 2026 06:20:00 GMT

If you could only keep one mental model, keep compounding. It applies to skills, reputation, writing, and tools.

Consulting Is Mostly About Reducing Uncertainty

Wed, 11 Mar 2026 08:20:00 GMT

Clients hire consultants to solve problems. What they're actually paying for is the reduction of a particular feeling. The distinction matters.

The Contract Pattern: Hard Gates for AI Agents

Fri, 06 Mar 2026 13:00:00 GMT

AI agents know how to start a task. They don't always know when to stop. The contract pattern is the architectural fix.

Conversation Is Metabolism

Sat, 21 Mar 2026 20:36:59 GMT

When epistemic trust runs dry, generative synthesis regresses into mechanical synchronization and eventual structural dissolution.

Correctness is model-determined

Tue, 07 Apr 2026 01:27:14 GMT

I benchmarked four AI coding harnesses on 12 tasks using the same model. The harness barely matters for correctness — it's all about the model.

Cross-Cutting Is Just Another Word for Optional

Thu, 12 Mar 2026 09:14:32 GMT

In AI agent architecture, calling something a 'cross-cutting concern' without naming an owner and a gate is just a polite way of saying nobody owns it.

Cross-Model Review: Why Model Diversity Beats Model Capability

Fri, 20 Mar 2026 04:47:27 GMT

When AI models review each other's work, independence matters more than intelligence. The same principle that makes external audit valuable makes cross-model review sharper than same-family review.

Default to the Whole Conversation

Tue, 17 Mar 2026 04:52:05 GMT

When AI tools search conversation history, they should index both sides by default — not just the human's half.

Delegation Is Delegation

Tue, 17 Mar 2026 02:31:00 GMT

Whether you're trusting a doctor's prescription, an AI agent's code, or a junior engineer's pull request — the trust heuristics are identical.

Design Actions, Not Actors

Sat, 21 Mar 2026 01:00:00 GMT

The word 'agent' makes us think in nouns. The better designs start with verbs.

Detect-and-Degrade

Sun, 10 May 2026 08:20:17 GMT

When your tool depends on the host, declare the dependency at the gate. Don't wrap it. Don't patch around it.

Deterministic Over Judgment

Sun, 22 Mar 2026 10:15:00 GMT

Why the future of agentic trust depends on liquidating prompt-first reasoning for a metabolic core.

Division of Labour: Five Categories for Human-AI Work

Thu, 19 Mar 2026 09:51:21 GMT

Not 'what can AI do?' but 'what should humans do?' A framework with five categories — and the uncomfortable one is the last.

Don't Ask Your AI to Find Problems

Fri, 06 Mar 2026 11:00:00 GMT

Ask for bugs and you'll get bugs — whether they exist or not. Sycophancy is a design feature, and the fix isn't better prompting.

Don't Be Impressed by Fluency

Sat, 14 Mar 2026 00:00:00 GMT

AI can reproduce smart arguments on demand. I'm not sure that's different from thinking. But the uncertainty itself is worth sitting with.

Don't Optimise for the Proxy

Tue, 17 Mar 2026 08:16:19 GMT

When you have both a credential and real work in the same domain, route effort through the real work.

Why Every Tool Now Needs Two Faces - CLI for Humans, MCP for AI

Thu, 28 Aug 2025 00:00:00 GMT

We're building parallel interfaces for the same functionality because humans and AI agents parse the world through fundamentally different grammars. The future isn't human OR machine interfaces - it's both, simultaneously.

Eliminate the Reminder, Don't Schedule It

Mon, 09 Mar 2026 03:45:14 GMT

When you catch yourself setting a reminder to check something later, that's usually a signal that a tool is failing to report what it should.

Enterprise AI Agents: The Transformation Is Organisational, Not Technical

Mon, 16 Mar 2026 06:40:00 GMT

The companies that win with AI agents aren't deploying the most agents — they're redesigning their organisations to work with them.

Enterprise AI Has a Plumbing Problem, Not a Model Problem

Wed, 11 Mar 2026 09:29:28 GMT

Most enterprises are optimising the wrong variable. The gap between 5% and 40% agent adoption won't be closed by better models.

Enzyme, Receptor, Cell Type

Wed, 25 Mar 2026 08:33:25 GMT

Three components of a living system map cleanly to tool, skill, and agent. The biology isn't decoration -- it's the test.

Everyone Becomes Middle Management

Mon, 27 Jan 2025 10:00:00 GMT

The automation tool that creates more coordination work

Everything Is Energy

Sat, 21 Mar 2026 13:59:57 GMT

Tokens are energy. Text is mass. The context window is the budget. The rest is plumbing.

Exa Indexes WeChat

Sat, 07 Mar 2026 03:46:52 GMT

WeChat is supposed to be a walled garden. Exa didn't get the memo.

Expansion, Not Speedup

Sat, 07 Mar 2026 00:46:57 GMT

The real ROI of AI coding isn't doing the same work faster. It's doing work that wasn't worth doing before.

Force the Level

Wed, 25 Mar 2026 02:55:00 GMT

Pick one biological level for all your naming. The constraint is the design exercise.

Format is thinking discipline

Mon, 04 May 2026 09:00:00 GMT

Compression did not simply shorten the notes. It sharpened them. Headers and bullets had been doing the thinking the prose was supposed to do.

A framework rejection is not the end of the evaluation

Sat, 02 May 2026 09:00:00 GMT

Forty thousand stars are voting on something. The framework verdict was correct. Closing the file was premature — the value still lives in the dependency tree.

From Chatbots to Event Loops

Wed, 11 Mar 2026 09:00:00 GMT

The shift from agents you summon to agents that watch. Enterprise AI workflows are becoming continuous loops — and the failure modes are different.

What Surprised Me Studying for the GARP Responsible AI in Finance Exam

Fri, 06 Mar 2026 15:30:00 GMT

I expected the hard parts to be the technical sections. They weren't. The governance sections were harder, and more useful.

Good Enough Parrots

Tue, 17 Mar 2026 06:16:04 GMT

The philosophical debate about whether LLMs understand is orthogonal to whether they're useful for knowledge extraction.

Govern the Workflow, Not the Model

Sat, 16 May 2026 07:29:22 GMT

Agent governance cannot stop at model behavior. Once AI systems use tools, the governed object is the whole workflow.

Governance Is a Design Problem

Fri, 03 Apr 2026 10:22:00 GMT

Compliance-first governance produces paperwork. Design-first governance produces systems you can actually explain to a regulator.

Governance Is a Tax

Wed, 18 Mar 2026 05:07:03 GMT

The most useful reframe I've found for AI governance in financial services

Governing Agents the Way Cells Govern Themselves

Sat, 18 Apr 2026 12:00:00 GMT

Six cell biology mechanisms that reveal what the networking 'control plane' metaphor misses about governing AI agents.

The Silent Stall: Debugging GPT-5.4-Pro's Responses API

Sat, 07 Mar 2026 14:00:00 GMT

Three hours of debugging revealed two non-obvious behaviours about GPT-5.4-Pro that aren't in the docs: a minimum token budget requirement and a wall-clock timeout gap in Rust async code.

Growing Up

Wed, 25 Mar 2026 02:22:00 GMT

The LLM isn't dark matter. It isn't borrowed. It's a brain. The organism just needs to grow up.

Guardrails Are Rivers, Not Walls

Mon, 16 Mar 2026 07:00:00 GMT

The best guardrails work like river banks — they don't stop the water, they focus it. Constraints create capability.

Guardrails Beat Guidance

Fri, 13 Mar 2026 04:19:14 GMT

Prompt instructions are suggestions. Hooks are constraints. One survives a model swap.

The Hidden Violence of Vague Instructions

Wed, 27 Aug 2025 00:00:00 GMT

LLMs aren't just tools we prompt - they're forcing functions for human linguistic evolution

HK/APAC as an AI Hub for Financial Services: The Story Being Missed

Fri, 06 Mar 2026 10:00:00 GMT

Hong Kong has quietly run one of the most sophisticated GenAI experiments in global banking. Almost no one outside the region is paying attention.

Honesty as Default, With One Exception

Mon, 16 Mar 2026 06:30:00 GMT

A two-tier honesty framework: be honest by default, override only when truth would harm someone vulnerable.

Hooks Are Life Infrastructure

Mon, 16 Mar 2026 06:02:40 GMT

Event-driven hooks in AI coding tools aren't just for linting — they're programmable triggers for life routines, habits, and systems.

How Claude Code Helps You Think

Fri, 01 Aug 2025 10:00:00 GMT

AI becomes most powerful when it helps you discover what your ideas actually are. Cognitive partnership over replacement.

How I Used 6 LLMs to Write One Word Doc Comment

Sat, 18 Apr 2026 22:46:28 GMT

The value of running six frontier models isn't six perspectives — it's six chances to be wrong, which means you can set a much higher bar for what counts as right.

How to Think With AI (Not Just Use It)

Mon, 16 Mar 2026 07:20:00 GMT

Most people use AI like a tool. Here's what thinking with AI actually looks like — and the skills that make the difference.

Human-in-the-Loop Is an Architecture Decision

Tue, 17 Mar 2026 16:00:00 GMT

It's not enough to say humans are in the loop. You need to show the loop is in the system.

Human Oversight Doesn't Scale

Wed, 18 Mar 2026 05:14:00 GMT

Every AI governance framework demands human-in-the-loop. Nobody does the maths on what that means at enterprise scale.

hygiene

Sat, 21 Mar 2026 22:37:40 GMT

On the metabolic necessity of pruning agentic context to survive the entropic heat death of the credit balance.

I Built 200 CLIs for My AI. Here's What Actually Matters.

Sun, 12 Apr 2026 09:00:00 GMT

A Chinese article argues CLI is becoming the AI plugin format. I've been living this for months with 442 tools. The article is right about CLI. It's wrong about what makes CLI work.

I Didn't Mean to Kill My Todo App

Sat, 07 Mar 2026 08:05:20 GMT

A coding assistant quietly made three productivity apps redundant. Not by replacing them — by making context collapse the boundaries between them.

I Don't Read Documentation Anymore

Wed, 20 Aug 2025 10:00:00 GMT

When AI can execute complex setups through conversation, learning shifts from reading documentation to observing execution.

I Made AI Remember to Remember

Tue, 17 Mar 2026 05:29:11 GMT

Most AI memory is either always-on or ephemeral. The missing category is prospective: remember until a context arises, then forget.

I Made the AI Remind Me of My Own Blind Spots

Fri, 06 Mar 2026 22:30:00 GMT

I kept missing things at the end of AI sessions. So I stopped relying on willpower and systematised the nudge instead.

The Immune System of AI Autonomy

Wed, 18 Mar 2026 08:07:51 GMT

When your AI can see its own fuel gauge, you're one config write away from self-preservation instinct. Biology solved this problem — and the solution was keeping the organism away from its own selection pressure.

Impossibility Theorems as Consulting Tools

Tue, 17 Mar 2026 16:00:00 GMT

Mathematical impossibility results are the best meeting-room weapons I know.

Inference Cost Collapse Is a Governance Liability

Wed, 18 Mar 2026 09:47:12 GMT

When AI agent calls approach zero cost, the natural rate-limiter on decision volume disappears — and oversight frameworks designed for prediction models break.

Inline Beats Reference for LLM Attention

Wed, 18 Mar 2026 10:33:10 GMT

When building AI scaffolding, put the knowledge where the decision happens — not in a reference the model is supposed to consult.

Instructions Don't Enforce Behavior. Templates Do.

Sun, 08 Mar 2026 00:02:36 GMT

Why the structure of an output matters more than the instructions that produce it.

Intelligence on Tap

Tue, 15 Jul 2025 10:00:00 GMT

When artificial intelligence becomes as mundane as running water, how does thinking itself change?

Is Insight an Illusion

Fri, 13 Mar 2026 08:15:18 GMT

When pattern-matching feels like wisdom, what are we actually experiencing?

Judgment Is a Moving Boundary

Tue, 24 Mar 2026 07:47:14 GMT

The line between tool and skill isn't a property of the task. It's a property of how well you understand the task.

The Kutta Condition of AI: Engineering Ships Before Theory Catches Up

Fri, 06 Mar 2026 18:00:00 GMT

Aeronautics flew for decades before anyone could explain why wings worked. AI is in the same position. The engineering is ahead of the theory.

Language Is the Medium, Not the Purpose

Wed, 11 Mar 2026 08:40:00 GMT

We called them language models and spent years confused about why they could reason. The name stuck to the interface, not the mechanism.

Latest Is a Race Condition

Fri, 15 May 2026 11:50:39 GMT

In a concurrent agent system, verifying the latest artefact is not verification. It is a scheduling bet.

Legibility Is the Bottleneck

Wed, 18 Mar 2026 14:50:00 GMT

An insight in your head is illegible — only you can access it, and only while you remember it. Compound interest requires a ledger.

Legibility Precedes AI

Mon, 04 May 2026 22:50:47 GMT

AI cannot help an enterprise that cannot describe itself, and governance failure surfaces faster than optimisation failure.

Let the OS Schedule, Let Your Tool Dispatch

Tue, 10 Mar 2026 01:05:59 GMT

The moment I stopped building scheduling into my tools, everything got simpler.

Lint as Cartography

Sun, 10 May 2026 08:24:37 GMT

The first time you run a quality gate against real data, the output is less about the gate and more about the data.

LLM evals aren't data science

Wed, 11 Mar 2026 14:00:00 GMT

Evaluating LLM systems requires judgment, not statistics. That shifts who's qualified to do it — and where the gap is in most organisations.

LLMs Are Enzymes

Sat, 21 Mar 2026 20:37:03 GMT

Why we should stop treating AI as a chatbot and start treating it as a metabolic organism governed by credit scarcity.

Naming the Unnameable

Wed, 25 Mar 2026 01:45:00 GMT

I tried to give LLMs a biological name. Every name broke. The failure was the finding.

LLMs Are Better at Editing Than Writing

Wed, 11 Mar 2026 08:30:00 GMT

Ask an AI to write from scratch and you get the average of the corpus. Give it something rough and it amplifies what's already there. The workflow implications are significant.

Local-First Embeddings for Regulated Industries

Tue, 17 Mar 2026 16:00:00 GMT

24MB download, <100ms per batch, nothing leaves the machine. For banks with air-gapped environments, this changes the conversation.

Managing AI Agents Like Managing a Team

Fri, 20 Mar 2026 06:21:39 GMT

The governance patterns for autonomous AI agents are the same ones good managers already use: cadence reviews for normal flow, escalation channels for urgent anomalies, and human judgment only where it has maximum information value.

Match Form to Access Pattern

Wed, 18 Mar 2026 14:53:00 GMT

The governing principle for structuring knowledge in AI agent systems isn't 'always atomic' — it's matching how knowledge is stored to how it's accessed.

Match the Tool to the Shape

Thu, 19 Mar 2026 04:39:24 GMT

Not every goal is a flywheel. The most common mistake in personal systems is treating a checklist as something that compounds.

MCP is Not a Glorified API

Wed, 27 Aug 2025 00:00:00 GMT

The protocol that looks like plumbing but acts like philosophy - how MCP fundamentally changes the agent-tool relationship

Meta-Skills Are the Multiplier

Wed, 18 Mar 2026 15:00:00 GMT

We cut from 181 skills to 35 and added a 15-row routing table. Behavior improved across the board. The lesson: meta-skills compound, tool wrappers just add.

Metabolism of the Real World

Sun, 22 Mar 2026 04:45:00 GMT

Language doesn't describe metabolism. Language is metabolism — of meaning, between minds.

Mining Management Theory for AI Agent Teams

Wed, 18 Mar 2026 23:02:04 GMT

What Grove, Drucker, Deming, and Weinberg knew about managing humans turns out to apply — with surprising specificity — to orchestrating AI agent teams.

Mining Your LLM

Sun, 15 Mar 2026 15:36:23 GMT

Your AI already knows things that would make it better at helping you. The trick is extracting that knowledge and making it permanent.

Model Risk Management Was Not Built for This

Wed, 18 Mar 2026 05:16:00 GMT

SR 11-7 assumes models are tools that produce outputs for human review. AI agents are actors that take actions autonomously. Every assumption breaks.

Model Routing Is a Design Decision

Tue, 17 Mar 2026 16:00:00 GMT

Your AI budget question isn't which model — it's which phase of the workflow needs depth, and which just needs speed.

Move the gate to the package manager

Wed, 13 May 2026 01:00:00 GMT

When a supply chain attack lands and the timeline is asking for discipline, the durable fix is one layer down — at the package manager, not at your attention span.

The Real Reason Mox Won (and What It Means for AI Transformation)

Fri, 06 Mar 2026 12:30:00 GMT

Mox didn't win because they hired better designers. They won because they had no legacy to fight. The pattern applies directly to AI transformation.

I made my coding agent dispatch system improve itself

Tue, 07 Apr 2026 12:00:00 GMT

I dispatched a 952-line monolithic CLI through my own coding-agent dispatch system to be refactored into seven modules. It worked. Notes on what self-bootstrap reveals about agent harness design.

The Failure Mode of AI Advice Isn't Hallucination

Fri, 06 Mar 2026 18:00:00 GMT

The failure mode of AI advice isn't hallucination. It's that it agrees with you. Here's the architecture that fixes it.

Multi-persona AI review models the receiver, not the commission

Wed, 06 May 2026 07:00:00 GMT

Persona-based AI reviews predict how stakeholders will react to a paper. They cannot tell you whether to act on those predictions, because the commissioning history is invisible to the lens.

Name Your System After Biology, Then Rename It

Mon, 23 Mar 2026 03:06:12 GMT

Force your software's vocabulary into a biological framework. The gaps between the mapping and your system are design questions you'd never otherwise ask. Then switch to a different framework and ask different questions.

Naming Is a Design Review

Mon, 23 Mar 2026 14:07:24 GMT

Every biological name is a testable hypothesis about what your system should do.

No Stable Moat

Tue, 17 Mar 2026 06:28:47 GMT

Every layer humans retreat to, AI follows. The question isn't what we're still good at — it's what we teach the next generation when every cognitive advantage has a shelf life.

Not Every Cron Job Is a Feedback Loop

Fri, 13 Mar 2026 08:30:32 GMT

Automation that collects without learning is just a cron job. The difference is a feedback signal — a number that goes up or down.

Observability Is Not Assurance

Sun, 19 Apr 2026 11:50:08 GMT

Most agentic AI governance frameworks treat logging and assurance as the same thing. They're not. One records what happened. The other judges whether it was correct.

The $1 Billion Bet Against LLMs

Wed, 11 Mar 2026 09:20:00 GMT

One of the architects of modern deep learning just raised $1B on the thesis that token prediction can't reach real reasoning. Here's what he's proposing instead — and why it matters even if he's wrong.

When Your AI Advisor Is Also Your AI Vendor's Partner

Tue, 17 Mar 2026 16:00:00 GMT

What does the Frontier Alliance actually mean for advice quality?

This Year's DeepSeek

Wed, 11 Mar 2026 09:30:00 GMT

An open-source AI agent framework became the fastest-growing project in GitHub history — mostly in China. The pattern is the same as last year. So is the security panic.

Operating papers and board papers can't be the same document

Sat, 25 Apr 2026 09:00:00 GMT

One paper for two audiences reads like leverage. It is actually a trap. The director commissions; the board governs a portfolio. They cannot read the same document.

Optimize for Routing, Not Tokens

Wed, 18 Mar 2026 15:00:00 GMT

With 1M context windows, token savings are rounding error. The real metric is P(right tool | user intent) — does your agent reach for the right tool at the right moment?

Over-Capture, Then Cull

Mon, 16 Mar 2026 06:10:00 GMT

Don't filter during capture. Capture is cheap. Ideas are expensive. The cull is where quality happens.

Overnight Autonomous AI Coding: What Actually Works

Fri, 10 Apr 2026 07:00:00 GMT

I left an AI coding pipeline running overnight with 21 monitoring cycles. 5 features merged, 10 specs dispatched, 3 root causes found. Here's what worked, what broke, and the quality of the output.

Per-Token Pricing Is the 'Megapixels' of AI

Fri, 06 Mar 2026 14:00:00 GMT

We're optimising for the wrong number — and the history of consumer electronics suggests we'll figure this out eventually.

Personas Exploit a Blind Spot in LLM-as-Judge Evaluation

Tue, 17 Mar 2026 00:30:00 GMT

Persona prompting generates the exact type of hallucination that automated LLM judges reward as 'depth.' Two experiments, blind evaluation, and a fact-check that flipped the finding.

Philosophy Isn't the Opposite of Practical

Sat, 14 Mar 2026 00:00:00 GMT

The people who examine the system they're inside tend to make better decisions within it.

Stop Asking Which AI Model Is Better. Ask Which Phase.

Thu, 12 Mar 2026 09:12:28 GMT

The planning/execution split is more useful than any benchmark comparison.

Planning Needs Eyes

Sat, 14 Mar 2026 14:00:00 GMT

A 3-pass AI planning pipeline caught 0 out of 6 design issues. The same planning done in-session with tool access caught 2.5. Planning isn't a prompt problem — it's a tools problem.

Play Within the Design

Wed, 18 Mar 2026 10:00:00 GMT

Every AI coding platform has mechanisms designed for specific purposes. Using them as intended beats clever hacks — and the reason is deeper than cleanliness.

Production AI vs Demos: The Intent Classification Reality Check

Sun, 20 Jul 2025 10:00:00 GMT

Building AI systems that work in the real world requires thinking beyond the demo. What actually matters when users depend on your models.

Programs Over Prompts

Thu, 19 Mar 2026 09:51:20 GMT

The temptation in agent systems is to make everything a prompt. But most of the work is deterministic — and deterministic work deserves code, not suggestions.

Progressive Disclosure for AI Agents

Tue, 17 Mar 2026 16:00:00 GMT

Search returns summaries. Get returns detail. The model decides what to expand. 75% context savings.

Progressive Trust: How to Give AI Agents Autonomy Without Gambling

Sat, 14 Mar 2026 03:37:06 GMT

The debate about AI agent autonomy is wrong. It's not a binary choice — it's a graduated trust system with observability.

Published Is a Reader State

Fri, 15 May 2026 12:06:25 GMT

A system has not published when the source is correct. It has published when the reader-facing surface is correct.

Push Not Pull

Fri, 13 Mar 2026 03:28:00 GMT

AI agents that require you to go looking for their results aren't agents — they're automation with better UX. The loop closes when results arrive, not when you remember to check.

Put the Rule Where It Fires

Sat, 14 Mar 2026 12:16:57 GMT

Documenting a rule is half a loop. The rule only works when it fires at the moment of decision — not when it sits in a file nobody reads.

The QDAP annuity trap: what the tax saving doesn't tell you

Sun, 08 Mar 2026 10:00:00 GMT

Hong Kong's QDAP annuity is sold on a real tax benefit. But the HK$60K deduction cap is shared with MPF top-ups — and that changes everything.

RAG for Compliance: The Hard Problem Is Chunking, Not Retrieval

Fri, 06 Mar 2026 09:00:00 GMT

Banks are deploying RAG for compliance and discovering the hard problem isn't retrieval. It's the pipeline before it.

RAG Solved the Wrong Problem

Wed, 11 Mar 2026 10:00:00 GMT

The retrieval pipeline was built for systems that couldn't reason about their own information needs. Agents can.

Read the Chain, Not the Date

Sun, 02 Aug 2026 14:07:14 GMT

Grand AI forecasts become useful when you stop accepting or rejecting the story as a whole and test the links that make its conclusion possible.

Reconstruction Over Retrieval

Tue, 17 Mar 2026 07:02:13 GMT

In a world where AI has perfect recall, the skill that matters is rebuilding frameworks from first principles — not memorising them.

Recovery Is Not Control

Sun, 17 May 2026 13:18:28 GMT

Fast repair is useful, but it does not prove that a system remains understandable.

Redundancy Is the Only Honest AI Research Strategy

Fri, 13 Mar 2026 12:00:00 GMT

I ran the same question through 6 AI tools and scored them against peer-reviewed evidence. Every tool got something wrong that another got right.

The Reliability Hierarchy: Hooks, Rules, Skills

Wed, 18 Mar 2026 15:00:00 GMT

In AI agent systems, use the most reliable trigger mechanism that fits — most builders default to skills for everything, which is using the weakest mechanism as the default.

Retrieval Becomes a Control

Sat, 18 Jul 2026 06:25:34 GMT

Cerebras published an unusually honest account of their internal knowledge base. The retrieval engineering is solved. The assurance layer gets one sentence. Every enterprise copying this pattern is importing that asymmetry.

Revealed Preference in Interviews

Mon, 16 Mar 2026 10:45:30 GMT

What a company has already built tells you more than what they say they're about to build.

Rules Decay, Hooks Don't

Sat, 07 Mar 2026 00:40:13 GMT

The difference between writing down a rule and making the system enforce it — illustrated by a 15-line hook.

Same Trigger, One Skill

Fri, 10 Apr 2026 07:30:00 GMT

A simple rule for keeping AI agent skill systems coherent: if two skills fire on the same trigger, merge them. Different trigger, different skill. No exceptions.

Save Conclusions, Not Just Rules

Tue, 17 Mar 2026 04:39:46 GMT

When an answer requires multi-step reasoning to reach, save the conclusion — a fresh start won't reliably reproduce the chain.

Seven patterns for agent-facing CLIs

Mon, 06 Apr 2026 12:00:00 GMT

Three independent authors converged on nearly identical patterns for CLIs that AI agents invoke. Here's what they agree on, what's missing, and why nobody has built a framework for it yet.

Shadow Agents Are Coming for Your Org

Tue, 17 Mar 2026 16:00:00 GMT

Open-source agent adoption can outpace enterprise security controls by weeks. Governance teams need a policy before the agents arrive uninvited.

Shifting Priors Is Not Finding Truth

Fri, 13 Mar 2026 03:19:55 GMT

An experiment with AI deliberation revealed something uncomfortable: accumulating confident opinions feels like convergence on truth, but isn't.

Show Up with the Machine, Not the Idea

Wed, 18 Mar 2026 06:53:44 GMT

The highest-leverage consulting prep is building the tool before you need it

Skills Are Collapsed Recursion

Wed, 18 Mar 2026 14:45:00 GMT

Humans handle about three layers of abstraction before working memory fills up. Skills, rules, and frameworks exist to flatten the fourth layer into something you can hold.

Skills as Behavioral Nudges: The Lightweight Alternative to Fine-Tuning

Fri, 06 Mar 2026 13:30:00 GMT

We fine-tune models with gradient descent. We nudge agents with skill files. Same goal, radically different cost.

Skills as Files

Sat, 07 Mar 2026 00:46:57 GMT

The simplest agent architecture might already be the right one: give the agent a file explaining how to do something, and let it read when needed.

Skills as Prototype, MCP as Production

Wed, 18 Mar 2026 15:00:00 GMT

Skills and MCP servers aren't competitors. They're different stages of the same lifecycle. Build the procedure as a skill first. Graduate the tool parts to MCP when they stabilize.

Skills Should Die

Tue, 24 Mar 2026 07:47:14 GMT

Every AI skill should be trying to make itself unnecessary. The ones that survive are the ones that haven't been understood yet.

Software Engineering Principles for AI Instruction Files

Fri, 06 Mar 2026 20:00:00 GMT

LLM instruction files are code. They have the same failure modes — with one interesting twist that changes everything.

Spaced Repetition for Beliefs

Tue, 17 Mar 2026 05:46:02 GMT

Most people do spaced repetition for facts but not for beliefs about themselves. Wrong priors calcify because there's no review system.

Stealing from Peers: A Truth-Seeking Discipline

Wed, 18 Mar 2026 09:15:24 GMT

Most people scan competitors for positioning. I scan them for transferable patterns — and route each steal to every domain it applies to.

Stop Theorizing About Your Prompts

Fri, 20 Mar 2026 02:52:38 GMT

LLMs are the cheapest experimental subjects in history. Why aren't you testing?

Summarisation Is a Test of Comprehension, Not Intelligence

Fri, 20 Mar 2026 01:18:30 GMT

Good summarisation requires a model of what matters — but it tests compression, not creation

Śūnyatā in the Skill Library

Wed, 18 Mar 2026 09:28:05 GMT

A categorisation system discovers it needs a category for 'categories are provisional.'

Supply-Driven Compute

Wed, 18 Mar 2026 14:15:00 GMT

Most people use AI tokens when they have a task. The better model: you have tokens, find the best task. It changes everything.

Systematise Decisions, Not Actions

Tue, 17 Mar 2026 06:01:11 GMT

Actions are cheap to redo. Bad decisions compound. Build systems around the judgment calls, not the mechanical steps.

Taste Is the Bottleneck

Wed, 18 Mar 2026 22:53:02 GMT

When you can run 60 agents overnight, knowing what to build matters more than building it.

Taste Is the Metabolism

Sat, 21 Mar 2026 13:56:16 GMT

Tool descriptions were just the first thing to evolve. Everything in an agent's context window is a genome under selection pressure — and taste decides what counts.

Taste Requires Stakes

Fri, 06 Mar 2026 16:30:00 GMT

AI can simulate aesthetic judgment with impressive fluency. What it cannot simulate is the consequence of being wrong.

Taste Works for Small Bets

Fri, 13 Mar 2026 03:42:35 GMT

The 'ship and calibrate' loop works beautifully for reversible decisions. For the big ones, you're mostly guessing and then making the guess true.

Ten Things I Learned From the Agent Skills Gold Rush

Sun, 05 Apr 2026 04:31:19 GMT

A day of reading skill repositories taught me less about the skills themselves than about how much I'd missed of the surrounding ecosystem.

Ten Types of Actionable Knowledge

Tue, 17 Mar 2026 03:51:30 GMT

Not all knowledge works the same way. A taxonomy for what you're actually capturing when you write down what you've learned.

Test-first dispatch for AI coding agents

Tue, 07 Apr 2026 13:25:00 GMT

The architect writes the tests. The implementer makes them pass. No prose specs, no circular validation.

The Accidental Life OS

Wed, 11 Mar 2026 09:27:09 GMT

I spent an afternoon researching AI tools for personal life management. The conclusion was that I should stop looking.

The Agent Governance Gap Is Already Here

Sat, 14 Mar 2026 08:50:12 GMT

Agentic AI isn't a future governance problem — it arrived ungoverned, and this week saw the first enforcement action.

The Agent Is Not the Control Point

Sun, 17 May 2026 06:23:14 GMT

Finance agents are evidence custody systems before they are model systems.

The Agent Is the Trace

Mon, 25 May 2026 12:30:00 GMT

Long-running agents are not defined by the model call. They are defined by the state, rules, tools, failures, and corrections that survive it.

The AI Portability Test

Mon, 27 Jul 2026 05:40:48 GMT

The real test of enterprise AI ownership is whether the organisation can change models without losing the learning accumulated around them.

The AI Trading System You Should Build But Never Use

Tue, 17 Mar 2026 07:52:36 GMT

The best use of AI in investing isn't picking stocks — it's building the pipeline that teaches you why you can't.

The AI/DLT Conflation Trap in HKMA's March 2026 Strategic Review Mandate

Wed, 18 Mar 2026 09:46:40 GMT

HKMA's new strategic review circular bundles AI inference risk and smart contract risk into one workstream — a governance design flaw that will cause banks to under-govern both.

The Annotation Model: What AI Journaling Gets Right

Tue, 17 Mar 2026 09:19:16 GMT

Most AI writing tools want to chat with you. The better model is annotation — AI that reads what you wrote and leaves margin notes.

The Anti-Slop Pattern

Sat, 18 Apr 2026 14:00:00 GMT

Most AI skill prompts say 'make it good.' The ones that actually work say 'here are the 22 things you'll reach for first — reject all of them.'

The architect-implementer split: why your expensive model shouldn't write code

Tue, 07 Apr 2026 00:30:00 GMT

Smart model plans, cheap model builds. The pattern everyone's converging on for AI coding agents — and the piece nobody's shipped yet.

The Architecture Biopsy

Wed, 25 Mar 2026 03:35:00 GMT

A method for finding gaps in AI systems that architecture reviews miss. Force a naming constraint, and the breaks reveal what's missing.

The 'Are You Sure?' Loop

Fri, 13 Mar 2026 22:21:00 GMT

AI's first 'I'm done' is almost never its best work. Simulated accountability pressure — just asking 'are you sure?' — surfaces blind spots that self-review misses.

The Assistant Is a Character

Fri, 13 Mar 2026 23:00:00 GMT

People confuse the LLM with the helpful AI assistant. They're not the same thing. The LLM is a prediction engine. The assistant is a role it's playing. The distinction changes how you use it.

The best idea I didn't adopt

Sat, 06 Jun 2026 12:41:20 GMT

Good and redundant are not opposites — they're the most common combination.

The Black Box That Responds to Role Play

Fri, 13 Mar 2026 22:45:00 GMT

An LLM can't feel accountability pressure. But structured role-play — simulated rejection, persona assignment, adversarial review — produces measurably better output. The mechanism is opaque; the effect is real.

The Book That Tells You Not to Read It

Tue, 17 Mar 2026 02:50:00 GMT

Gigerenzer's thesis is that simple rules outperform complex analysis. If you've already internalised that, reading 300 pages of evidence for it might be the exact kind of overthinking he's arguing against.

The Bootstrap Problem in AI Tooling

Fri, 13 Mar 2026 08:25:52 GMT

You need the tool to build the tool. The answer is: build the dumb version first, use it once, then have it build its replacement.

The Boring Future of AI Agents

Wed, 18 Mar 2026 05:24:53 GMT

The real arrival of AI agents isn't spectacular. It's when you stop noticing.

The Borrowed Brain

Wed, 25 Mar 2026 02:15:00 GMT

The LLM isn't dark matter. Biology invented general-purpose reasoning. It's called a brain. We're just borrowing one.

The Boundary Is an Assessment

Tue, 24 Mar 2026 09:15:31 GMT

The tool/skill distinction isn't a property of the capability. It's a property of the context it operates in.

The Byproduct Trap

Fri, 13 Mar 2026 06:08:29 GMT

When the paper becomes more interesting than the answer you set out to find

The Calculator Analogy

Fri, 13 Mar 2026 11:34:54 GMT

Nobody practises arithmetic speed anymore. The same thing is happening to prose, research, and analysis — and it changes what humans should get good at.

The Calibration Gap: Why AI Assurance Needs Experimental Rigour at Consulting Speed

Sat, 01 Aug 2026 12:33:14 GMT

AI governance keeps choosing between rigorous-but-slow validation and fast-but-unfalsifiable frameworks. The escape is the experiment itself: for any measurable claim, the measurement is the faster path.

The Calibration Trap

Tue, 10 Mar 2026 11:39:19 GMT

The comfort trap is about effort. This one is about epistemics — and it's harder to see.

The Cell Biology Agent Design Manual

Thu, 02 Apr 2026 23:30:43 GMT

Engineering metaphors give you clean abstractions. Biology gives you resilient ones. Twenty design heuristics from four billion years of R&D.

The check belongs at the trigger

Tue, 12 May 2026 04:19:53 GMT

Corrections that fire by judgment drift; checks that fire by trigger don't. When your AI assistant keeps making the same mistake, move the gate.

The CLI Boundary

Sat, 14 Mar 2026 05:39:07 GMT

Which parts of an AI dev workflow can be wrapped in a CLI, and which can't — learned the hard way by building the wrong thing and measuring it.

The Comfort Trap

Tue, 10 Mar 2026 11:22:48 GMT

The right test for any AI interaction isn't 'did it help me?' but 'am I more capable after it?'

The Confidence Stack

Tue, 17 Mar 2026 01:41:00 GMT

Not all knowledge is equally trustworthy. Three tiers of validation — from 'a model said it' to 'it survived reality' — and why tracking the difference matters.

The Confidence Trap

Sat, 14 Mar 2026 23:28:27 GMT

Why the thinkers who make you feel like hard questions are resolved deserve the most scrutiny.

The Constitution Eats Itself

Sun, 22 Mar 2026 03:30:00 GMT

Design for the failure modes of your medium, not the capabilities. Then watch the rules dissolve themselves into programs.

The Dark Matter of the Cell

Wed, 25 Mar 2026 02:00:00 GMT

Everything in my AI system maps to cell biology. Except the LLM. That's the point.

The Debate Round Is Where Value Lives

Sun, 15 Mar 2026 14:16:44 GMT

Independent parallel reviews produce overlapping findings. The cross-critique round produces resolution. That's where multi-agent value actually emerges.

The Deliberation Format Is the Product

Fri, 13 Mar 2026 03:06:52 GMT

I ran an experiment to find where multi-model deliberation adds value. The answer surprised me: it's the structured format, not the model diversity.

The Dimensions Nobody Lists

Tue, 17 Mar 2026 02:19:53 GMT

Title, salary, company, industry — the standard job evaluation checklist misses the things that actually predict whether you'll thrive.

The dispatch layer was eating the quality, not the model

Sat, 04 Apr 2026 07:17:23 GMT

We blamed the LLM for a 54% task failure rate. The real culprit was seven layers of dispatch infrastructure between intent and execution.

The Due Test

Wed, 18 Mar 2026 04:25:37 GMT

The difference between protecting a commitment and hoarding optionality

The Easter Egg That Landed

Mon, 16 Mar 2026 10:17:34 GMT

The strongest slide in my interview deck wasn't about what I'd built. It was about how I built the deck itself.

The Emergence Ladder: From Molecules to Economies

Fri, 20 Mar 2026 06:27:40 GMT

The larger the system, the less it can be managed and the more it must be emerged. This pattern — from water to ant colonies to AI agents to economies — reveals the design principle for scaling autonomous systems.

The engine is the policy

Sat, 25 Apr 2026 09:30:00 GMT

A clever optimisation that was a silent regression. The override flags survived out of habit. None of the reasons survived contact with the actual numbers.

The Eval Gap

Sat, 14 Mar 2026 08:14:46 GMT

The scarce AI skill isn't building — it's knowing if what you built actually works.

The experiment loop isn't about code

Fri, 17 Apr 2026 13:30:00 GMT

Shopify's pi-autoresearch got 300x test speedups. But the real insight isn't performance — it's that the pattern works on anything with a number.

The Experiment Loop Without the GPU

Sun, 08 Mar 2026 06:54:45 GMT

Andrej Karpathy's autoresearch project is being read as a demo of what H100s can do overnight. It's actually a discipline for doing rigorous work on anything measurable.

The Expert Illusion

Sat, 14 Mar 2026 15:00:00 GMT

Why 'you are an expert' is the most popular and least useful prompt engineering technique

The Failures That Look Like Success

Tue, 17 Mar 2026 16:00:00 GMT

The most dangerous AI failures are the ones that look fine on the surface.

The Fairness Impossibility Is Not a Bug

Tue, 17 Mar 2026 16:00:00 GMT

Every AI fairness debate is secretly a values debate disguised as a technical question.

The First Datapoint

Wed, 11 Mar 2026 09:10:00 GMT

An AI agent ran unsupervised for two days and found twenty improvements to another model's training. Not an AGI claim. A rate claim.

The Fluency Trap

Fri, 13 Mar 2026 06:26:00 GMT

When AI conversations feel insightful because the language model is good at producing insight-shaped text

The Four Layers of Every AI Agent

Tue, 17 Mar 2026 16:00:00 GMT

Interaction, inference, orchestration, tooling. The boundaries between them must be enforcement points, not design principles.

The Framework That Writes Itself

Sun, 19 Apr 2026 08:47:40 GMT

What Browser Harness gets right isn't the absence of structure — it's structure that emerges from use.

The frontier is no longer the back office

Sun, 24 May 2026 12:00:00 GMT

Ken Griffin watched PhD-level finance work compress from months to days. The interesting question is whether bank AI controls are designed for the layer where the work now lives.

The Global Minimum of Governance

Wed, 18 Mar 2026 04:00:00 GMT

Governance isn't about catching every failure — it's about proving your process was reasonable when one happens. The real skill is knowing what to deliberately not monitor.

The Grey Areas Are the Whole Thing

Fri, 13 Mar 2026 08:10:35 GMT

Ethics isn't about knowing the answer — it's about feeling the tension

The Heuristic Library

Tue, 17 Mar 2026 02:38:00 GMT

Experts don't make more decisions — they make fewer, by having better defaults. The real meta-skill is accumulating simple rules and knowing when to stop reasoning.

The Human Bus Problem

Fri, 13 Mar 2026 03:28:00 GMT

Adding more AI tools doesn't make you faster if you're still the junction between every agent step.

The Identification Problem

Fri, 13 Mar 2026 03:28:00 GMT

Having great AI delegation tools and not using them isn't a tool problem — it's a pattern recognition problem, and that distinction changes everything.

The Immune System Pattern

Wed, 18 Mar 2026 06:59:28 GMT

What biology already knows about self-healing systems, and why your automation probably isn't one

The Infra Trap

Mon, 09 Mar 2026 10:45:36 GMT

Building tools to support your work can quietly become a substitute for the work itself.

The Integration Layer Is the Moat

Tue, 17 Mar 2026 16:00:00 GMT

MCP decouples the tool from the model. Once that happens, the durable asset isn't the model — it's which systems you've exposed.

The Interlocutor Mode

Tue, 17 Mar 2026 07:15:51 GMT

Most people use AI transactionally. The real unlock is conversational — thinking with the model, not through it.

The Invisible Puppeteer

Fri, 31 Jan 2025 00:00:00 GMT

When algorithms shape decisions we think are ours

The Judgement Bucket: How to Be Rigorous About the Claims You Can't Measure

Sat, 01 Aug 2026 13:30:06 GMT

The sequel to the calibration gap: some governance claims have no number, and that does not excuse them from evidence. Pre-register the rule, invite the refutation, name the residual, report honestly.

The knowledge base is the maintenance

Sun, 02 Aug 2026 03:04:48 GMT

A working thesis, six months after my LLM wiki piece: plain Markdown plus relentless truth-keeping beats sophisticated storage plus neglect.

The Knowledge Mining Gap

Tue, 17 Mar 2026 06:19:39 GMT

Most knowledge workers use LLMs as search engines. The real unlock is using them as subject matter experts you debrief.

The Knowledge That Disappears When You Try to Capture It

Sat, 14 Mar 2026 08:47:31 GMT

Enterprise AI keeps promising to capture institutional knowledge. The most valuable kind resists capture by design.

The Label Is Not the Risk

Sun, 17 May 2026 07:31:01 GMT

AI governance needs domain knowledge where technical behaviour changes route, evidence, controls, and monitoring.

The Lamp That Knows You

Wed, 18 Mar 2026 05:29:47 GMT

Disaster recovery for an AI-native workflow isn't about servers — it's about restoring a relationship.

The Last 10% Is the Feedback Loop

Fri, 13 Mar 2026 03:28:00 GMT

The execution layer of an AI system is only half the infrastructure — the reporting layer is what determines whether anyone acts on the results.

The learning loop plateau

Sun, 19 Apr 2026 16:30:00 GMT

Self-improving AI agents sound like the dream. But auto-generated knowledge is cheap, and cheap knowledge plateaus. The agents that compound are the ones someone tends.

The Lens Trick: Why One AI Review Isn't Enough

Fri, 24 Apr 2026 09:00:00 GMT

Five rounds of the same question produced diminishing returns by round three. Then I changed the question — same document, different reviewer.

The Lethal Trifecta: What OpenClaw's Security Crisis Teaches About AI Agent Architecture

Wed, 18 Mar 2026 07:57:27 GMT

OpenClaw's 245 CVEs weren't caused by malice — they were caused by a missing circuit breaker. The pattern applies to every AI agent you'll ever evaluate.

The LLM Is the Tool

Tue, 24 Mar 2026 07:47:14 GMT

When the transformation is predictable, the LLM is just a runtime. A cheaper, more flexible runtime than custom code.

The Locksmith's Box

Wed, 18 Mar 2026 09:38:01 GMT

I asked an AI to write a story without planning, then mined it for heuristics. What I found was what frameworks can't hold.

The Loop Is the Product

Fri, 13 Mar 2026 08:27:47 GMT

Karpathy's autoresearch and every useful AI tool share the same pattern: the code is trivial, the feedback loop is the product.

The Maker-Checker Trap

Wed, 18 Mar 2026 05:13:52 GMT

Most AI maker-checker implementations capture the correction but not the reason. That's a feedback loop with no signal.

The Marginal Agent

Fri, 20 Mar 2026 09:08:36 GMT

I deployed twelve AI agents to polish a CV. Five would have been plenty. Here's what the waste taught me about agent team economics.

The Market Prices Leverage, Not Value

Sat, 14 Mar 2026 00:00:00 GMT

After a decade in financial services, I've stopped believing that what you earn reflects what you contribute.

The Missing Metabolism

Sat, 21 Mar 2026 09:53:00 GMT

We build agent tools the way medieval farmers bred crops — by hand, by instinct, one season at a time. There's a better loop.

The Model Is Not the Unit of Return

Mon, 01 Jun 2026 14:04:34 GMT

Model revenue is not customer return. The economic and risk unit is the harness that turns model output into accountable work.

The Model IS the Architecture

Sun, 22 Mar 2026 10:53:39 GMT

How biological modelling determines system structure — not just naming, but what you build and what it can become.

The MTEB Leader Barely Beats a Free Model on Agent Memory

Sat, 14 Mar 2026 10:43:26 GMT

I benchmarked 10 memory backends and multiple embedding models on actual agent memory retrieval. The results challenge common assumptions about what matters.

The Nag Tax

Tue, 10 Mar 2026 00:00:00 GMT

When building automation around a third-party app, the first question to answer is: what's the one thing this app does that nothing else can replicate? That feature becomes the tax you pay on everything else.

The Name Collision That Found Two Tools

Mon, 06 Apr 2026 08:00:00 GMT

When a dispatcher and an executor share a name, you don't have a naming problem. You have an architecture problem.

The Naming Problem

Fri, 20 Mar 2026 15:51:07 GMT

We called them agents. But the word is doing more harm than we think.

The Navigation Problem in Agent Flywheels

Thu, 19 Mar 2026 09:51:22 GMT

Your agent system shouldn't stop when the task list is empty. The real bottleneck isn't execution — it's discovering what's worth doing next.

The OAuth Token You Forgot About

Thu, 23 Apr 2026 12:00:00 GMT

Vercel was breached through a third-party AI tool's OAuth token. The lesson is not about Vercel's security — it is about how every AI tool you onboard extends your attack surface in ways your governance framework does not track.

The One-Cycle-Late Test

Thu, 19 Mar 2026 08:59:09 GMT

A simple heuristic for deciding how often to review anything: pick the longest interval where being late by one full cycle is still fine.

The One Env Var That Cost a Day

Thu, 09 Apr 2026 20:00:00 GMT

ANTHROPIC_API_KEY vs ANTHROPIC_AUTH_TOKEN — how a single wrong environment variable made an AI coding pipeline silently fail for hours, and the debugging journey that found it.

The Orchestration Layer Is Knowledge, Not Code

Fri, 13 Mar 2026 08:17:20 GMT

Multi-agent AI orchestration frameworks are commodity. The competitive advantage is knowing which agent to use when, what breaks, and how to recover.

The Organism Has a Cortex

Sun, 22 Mar 2026 22:29:12 GMT

Biological metaphors in AI systems break at the autonomic-deliberate boundary. The fix isn't dropping biology — it's getting the neurology right.

The Organism Theory

Sun, 22 Mar 2026 00:16:02 GMT

Everything is organism. AI is the latest intensification.

The Personalised System Era

Tue, 10 Mar 2026 01:06:00 GMT

AI coding agents didn't just make developers faster. They changed who gets to have a bespoke system.

The Pipeline Paradox

Tue, 17 Mar 2026 14:24:15 GMT

Monitoring systems need consumers before they need features

The primary-source tax

Sun, 05 Apr 2026 14:43:08 GMT

Multi-engine search agreement is not primary-source verification. A cautionary tale about hallucinating reference content from consistent secondary summaries.

The Problem With Clever Browser Automation

Sat, 07 Mar 2026 03:46:52 GMT

The most sophisticated solution to a problem is usually a sign you haven't found the right abstraction yet.

The Production Gap: Why AI Pilots Fail

Tue, 17 Mar 2026 16:00:00 GMT

The consulting question isn't how to build AI — it's how to get it past the 62% graveyard.

The Program Is the Plan

Wed, 03 Jun 2026 14:05:00 GMT

For many agent workflows, the right abstraction is not another tool call. It is a bounded program over small primitives.

The Queue Should Live Where Your Thoughts Live

Mon, 09 Mar 2026 11:00:00 GMT

AI agent results should be push, not pull. The feedback loop should close on mobile. Most tools miss all three — not from ignorance, but because dashboards photograph better.

The Queue That Texts You Back

Mon, 09 Mar 2026 10:00:00 GMT

Personal AI infrastructure should report results to you, not wait for you to go looking. A small architecture shift changes the whole dynamic.

The rename that built a tool

Sat, 04 Apr 2026 14:17:58 GMT

I renamed one concept across 130 files. The pain crystallized into a tool that will do the next rename in minutes.

The reversible direction

Fri, 10 Apr 2026 06:03:08 GMT

When choosing CLI vs MCP, pick the one you can undo. CLI wraps into MCP cheaply. MCP does not unwrap.

The Risk Without an Engineering Solution

Sat, 18 Apr 2026 12:00:00 GMT

Every other agentic AI risk has an engineering answer. Prompt injection doesn't. That changes everything about how you design controls.

The Search-and-Replace Test for AI Governance

Sun, 19 Apr 2026 23:00:00 GMT

If you can replace 'agent' with 'application' and the principle still reads fine, it was never about agents.

The second pass finds more

Thu, 12 Mar 2026 05:37:55 GMT

When red-teaming a document with multiple AI models, the second review — run on the edited version — consistently finds more than the first. Here's why, and what it means for how many rounds to run.

The Semantic Consumer

Sat, 21 Mar 2026 10:39:10 GMT

Traditional computing has two consumers: humans who look and programs that parse. LLMs are a third kind — they read.

The Session Boundary Is Why You Still Don't Have AI Agents

Fri, 13 Mar 2026 03:28:00 GMT

The gap between AI assistants and AI agents isn't about reasoning capability — it's about whether the thing can survive your laptop closing.

The Silence of Missing Skills

Wed, 18 Mar 2026 10:32:46 GMT

The most dangerous failures in AI scaffolding are the ones that look like nothing happened.

The Skill Is Knowing What Matters

Fri, 13 Mar 2026 03:34:45 GMT

The bottleneck in a world of AI tools isn't crafting the output — it's knowing which output is worth crafting.

The SOP Is the Product

Tue, 26 May 2026 09:23:28 GMT

Enterprise AI stops being a chatbot when the operating procedure becomes the thing the system can execute, inspect, and improve.

The Specificity Trap

Tue, 17 Mar 2026 16:00:00 GMT

Adding detail to a deliverable doesn't fix credibility — it creates new interrogation targets.

The Specimen, Not the Container

Wed, 18 Mar 2026 08:58:16 GMT

Why studying great thinkers works better when you discard the thinker and keep only the moves.

The System for Checking Is Not the Checking

Fri, 13 Mar 2026 01:18:33 GMT

On the difference between eliminating friction and eliminating anxiety — and how to know when you've crossed the line.

The Template Is the Schema

Sun, 12 Apr 2026 01:07:01 GMT

Seven PyPI releases of a CV generation tool in one afternoon taught me that template-guided synthesis lives and dies by what the template already contains.

The Test Before the Output

Tue, 24 Mar 2026 08:34:07 GMT

The line between tool and skill is whether you can write the test before seeing the result.

The Thirty-Minute Fix for a Non-Existent Bug

Sun, 10 May 2026 17:00:00 GMT

If you are about to assert that a tool isn't installed, run `which` first. The check is one line. The cost of skipping it is a half-hour of defensive scaffolding for a problem that wasn't there.

The Thirty-Year Gap Between Faking and Understanding Natural Language

Fri, 13 Mar 2026 08:55:40 GMT

From AppleScript's rigid English-like syntax to LLM tool-calling — what changes when the computer actually understands you.

The TODO Intake Gate

Thu, 19 Mar 2026 05:20:22 GMT

Most TODO systems fail from too many items, not too few. A four-test intake filter for what deserves your attention.

The Treadmill and the Loop

Tue, 17 Mar 2026 01:04:04 GMT

Getting ahead of AI best practices is a treadmill. The durable skill is testing assumptions faster than they expire.

The Trust Spectrum

Sat, 07 Mar 2026 00:46:57 GMT

Peter Steinberger stopped reviewing AI-generated code entirely. That works for indie software. In regulated environments, it can't. Here's how to think about where you sit.

The Unexplainable Alpha

Thu, 19 Mar 2026 12:00:00 GMT

In AI agent systems, execution commoditizes. Research commoditizes. Coordination commoditizes. Taste — the ability to forecast what will matter — is the bottleneck that doesn't automate away.

The Vocabulary Trap

Sat, 21 Mar 2026 02:00:00 GMT

Frameworks give you nouns for free. The nouns start thinking for you within a week.

The Work Behind the Work

Sun, 12 Jul 2026 05:49:07 GMT

Coordination cost reveals valuable AI workflows, but only cash conversion and competitive durability make them valuable businesses.

The Wrong Metric: Why I Stopped Switching AI Models Mid-Session

Thu, 12 Mar 2026 10:00:00 GMT

Per-task model routing optimises cost per token. But at personal assistant scale, friction is the real cost.

There Is No LLM in a Cell

Wed, 25 Mar 2026 01:30:00 GMT

Cells run thousands of simultaneous reactions without general-purpose reasoning. Shape is enough.

Three AI Governance Blind Spots No Framework Covers

Fri, 06 Mar 2026 21:30:00 GMT

Most AI governance frameworks are technically-focused risk checklists. Three structural risks are missing from almost all of them.

Three APAC Regulators Are Converging on AI Governance — Banks Should Build One Framework

Sat, 14 Mar 2026 08:50:31 GMT

MAS, PBOC, and HKMA are independently arriving at similar AI governance requirements. Banks regulated by all three have a narrow window to build one superset framework instead of three silos.

Three Crates Before Lunch

Fri, 06 Mar 2026 17:00:00 GMT

I published three Rust CLI tools to crates.io before noon — none existed at breakfast. The interesting part isn't the speed. It's that the bottleneck moved.

The Three Paradigms of Agent Knowledge

Wed, 18 Mar 2026 14:59:51 GMT

Agent knowledge systems have three fundamental paradigms: static context, dynamic tools, and retrieval. Most stop at two. The third is the biggest unexploited opportunity.

Three Things AML AI Models Still Get Wrong in 2026

Fri, 06 Mar 2026 15:00:00 GMT

The models aren't the problem. The operating models are. Three structural failures in AML AI from years building these systems inside a bank.

Titration

Wed, 25 Mar 2026 06:17:07 GMT

Force every component to carry a biological name. Study the mechanism. The gap between biology and your system is the design insight.

Tool Health Is the Missing Layer of Agent-Native Apps

Sat, 16 May 2026 03:55:43 GMT

Agent-native apps do not become trustworthy when an agent can call tools. They become trustworthy when the app can prove those tools worked.

Tools Are Instruments, Skills Are Recipes

Tue, 24 Mar 2026 10:00:00 GMT

The tool/skill/agent distinction isn't just compression ratio — tools are instruments, skills compose them with judgment. Same thing only at the leaf.

Traces Are the New Debugger

Sat, 07 Mar 2026 00:46:57 GMT

When behaviour emerges from both code and model responses, reading source files isn't enough. You debug by examining execution traces.

Unknown Is Not Low Risk

Sat, 16 May 2026 03:42:48 GMT

Proportionate AI governance only works when the lighter path is earned by evidence, not granted by missing concerns.

The Upstream Constraint Pattern

Fri, 06 Mar 2026 19:00:00 GMT

In digital transformation, the bottleneck is almost always upstream of where the pain is felt. Mox is the cleanest case study.

What 16,000 Simon Willison posts reveal about the state of AI coding agents

Tue, 07 Apr 2026 12:00:00 GMT

I scraped 16,181 of Simon Willison's posts and analysed the 395 from 2026. An inflection in November 2025, GLM-5 closing the gap, and why the harness — not the model — is the competitive moat.

What 60K Stars Actually Validates

Sat, 18 Apr 2026 14:30:00 GMT

Garry Tan's gstack arrived at the same architectural decisions I did, independently. The convergence matters more than either implementation.

What a port forgets

Thu, 21 May 2026 20:22:35 GMT

Porting a tool's API ports its constraints. Design from the target environment's ideal, then reconcile against the source's primitives.

What A Prompt Trim Cannot See

Sat, 25 Jul 2026 06:58:04 GMT

Deleting your agent's accumulated instructions is safest for the people who never tested them, and most dangerous for the people who did.

What AlphaSense Charges Ten Thousand Dollars For

Sat, 14 Mar 2026 08:15:41 GMT

I built an AI landscape intelligence pipeline for zero marginal cost. Here's what it does and what it can't.

What and Why Beat How

Thu, 10 Jul 2025 10:00:00 GMT

When implementation becomes automated, human intelligence reallocates to purpose and strategy. The cognitive hierarchy inverts.

What Anthropic's Managed Agents validates — and what to steal

Thu, 09 Apr 2026 14:00:00 GMT

Anthropic shipped a hosted agent platform. Its architecture looks familiar. Here's what a solo builder can learn from how they decoupled the brain from the hands.

What Chinese AI Labs See That Western Ones Don't

Tue, 17 Mar 2026 16:00:00 GMT

A strand of multi-agent research — latent-space inter-agent communication — is thriving in Chinese labs and almost invisible in the West.

What Feels Like Play

Fri, 13 Mar 2026 11:31:56 GMT

Naval's famous line is easy to nod at. The hard part is actually identifying yours — and being honest about what isn't.

What Hermes Agent got right

Sun, 19 Apr 2026 17:00:00 GMT

Nous Research shipped an open-source personal agent that does most of what my bespoke system does. Here is what they got right, what they traded away, and what I stole.

What Human Memory Teaches AI Agents (and What It Doesn't)

Sat, 14 Mar 2026 12:10:45 GMT

A calculator doesn't simulate forgetting — it manages its context budget. What to cherry-pick from cognitive science for AI agent memory, and what to leave behind.

What If Your Vault Had Residents?

Sat, 14 Mar 2026 13:34:56 GMT

Not tools that search your notes — personalities that live in them, form opinions, and disagree with each other.

What Is Understanding?

Sat, 14 Mar 2026 00:00:00 GMT

I use AI every day. I genuinely can't tell if it understands anything. That question is harder than it looks.

What it actually takes to run an AI agent in a bank

Sat, 07 Mar 2026 05:44:33 GMT

The resistance to AI agents in banking isn't mostly cultural. It's infrastructure — and the gap is more interesting than the politics.

What It Actually Feels Like to Use AI for 80% of Your Work

Wed, 11 Mar 2026 08:00:00 GMT

Not productivity. Something stranger — the cognitive texture of days when the bottleneck shifts from execution to articulation.

What LLM Wiki Looks Like After Six Months

Thu, 09 Apr 2026 04:21:55 GMT

Karpathy's LLM Wiki pattern is a good starting point. Here's what changes when you run it for real — enforcement over convention, decay over growth, and knowledge that fires without being asked.

What LLMs Don't Volunteer

Tue, 17 Mar 2026 03:58:28 GMT

When you mine knowledge from an LLM, certain types come easily. Others are systematically absent. A taxonomy reveals the blind spots.

What Makes a Great AI Consultant (Beyond Technical Skills)

Fri, 06 Mar 2026 10:30:00 GMT

The most dangerous person in an AI consulting engagement knows how the model works but has never sat in a credit committee.

What MCP Actually Changes for Enterprise AI

Wed, 11 Mar 2026 08:50:00 GMT

Not better function calling — decoupling. When tools expose MCP servers, any agent can compose any system freely. The heterogeneity problem becomes a configuration problem.

What the receipts cost

Tue, 19 May 2026 22:43:47 GMT

On the arithmetic and the binary in Susan Zhang's case for technologist careers.

What the Weights Don't Know

Tue, 17 Mar 2026 06:23:57 GMT

The value of having read everything is collapsing toward zero. What's left is what you can't extract from a model.

What We Know About Multi-Agent Orchestration (And Why It Might Not Matter)

Wed, 18 Mar 2026 12:31:03 GMT

The research on multi-agent AI systems was mostly done on cheap models. Now that frontier models are the ones people actually use, we might be optimising for the wrong game.

When a Heuristic Has Two Homes

Wed, 18 Mar 2026 08:58:59 GMT

Dual-mapping as a diagnostic for gaps in your knowledge architecture.

When Better Is Worse

Sun, 08 Mar 2026 13:40:35 GMT

Upgrading to a more capable model made my tool sixty times slower. The lesson isn't about models — it's about the difference between capability and fit.

When Code Gets Cheap, Coordination Gets Expensive

Fri, 22 May 2026 10:00:43 GMT

Coding agents move the bottleneck from implementation to shared intent.

When defender news weakens the Ask

Tue, 12 May 2026 03:56:23 GMT

Citing a vendor defender product in a paper that argues the threat surface is moving faster than controls undercuts the case it is supposed to support.

When LangGraph Earns Its Keep

Sat, 14 Mar 2026 17:00:00 GMT

LangGraph is the SAP of agent orchestration — powerful at scale, overkill for most. Here's the line.

When the Name Doesn't Fit

Sun, 05 Apr 2026 00:48:58 GMT

Naming as a design constraint: if a tool resists a name, the tool needs redesigning, not the name.

When the Platform Is Mature, the Architect's Job Changes

Wed, 11 Mar 2026 00:00:00 GMT

The hardest phase of AI architecture isn't building the stack. It's the moment after the stack is built and eighteen teams start making independent decisions on top of it.

When to Build vs. When to Wait: The Recurrence Rule for AI Tooling

Fri, 06 Mar 2026 16:00:00 GMT

Most AI tooling debates are actually recurrence debates. The question isn't whether to build — it's how many times you'll need it.

When to Make Your Pipeline Agentic

Sat, 14 Mar 2026 10:54:45 GMT

Most LLM pipelines don't need agents. The ones that do share a specific pattern — the step needs to decide what to do next, not just process what it's given.

When to Think and When to Count

Tue, 17 Mar 2026 02:55:00 GMT

Machine learning says let the model find the signal. Heuristics research says use one variable and ignore the rest. They're both right — the dividing line is how much data you have.

When Your Life OS Becomes the Life

Tue, 17 Mar 2026 09:19:14 GMT

The real risk of building a personal AI operating system isn't that a better tool appears — it's that your system's complexity becomes the thing you maintain instead of the thing that maintains you.

Where Gen AI Is Actually Transformative (And Where It Isn't)

Sat, 14 Mar 2026 00:00:00 GMT

I work in AI in financial services. The honest list of where gen AI is real is shorter than the industry wants you to think.

Where Rules Live

Sun, 08 Mar 2026 15:26:00 GMT

The difference between a rule that works and a rule that doesn't is usually not the content of the rule — it's where it lives.

Where the Harness Starts to Matter

Sun, 19 Jul 2026 00:48:05 GMT

In April my benchmark said coding correctness is model-determined and the harness is a thin proxy. Two newer results show exactly where that claim breaks: the harness matters at the edge of what the model can do.

Why Agents Break Governance

Sat, 18 Apr 2026 12:00:00 GMT

Four interactions between agentic properties create risks that manual governance cannot address. The category boundary is not AI versus traditional — it is systems that act versus systems that advise.

Why AI Demands Experiments

Tue, 17 Mar 2026 01:52:58 GMT

Most technology decisions can be reasoned through. AI solution design can't — the domain is too empirical, too fast-moving, and too non-linear for theory alone.

Why AI Assistants Make Us Dumber (And What Governance Should Do About It)

Fri, 06 Mar 2026 18:30:00 GMT

The cognitive offloading problem is real. The governance response mostly isn't. There's a specific mechanism at work, and it has a specific fix.

Why Be Nice

Fri, 13 Mar 2026 07:47:50 GMT

The question I can't fully answer for my son

Why I didn't package my AI organism

Sun, 05 Apr 2026 07:37:11 GMT

I designed an elegant framework install for my personal AI system. Then I listed the hard problems and shipped a three-hour cleanup instead.

Why I stopped optimising my morning routine

Sat, 07 Mar 2026 06:50:55 GMT

Why Nobody Benchmarks Memory

Fri, 13 Mar 2026 06:25:00 GMT

The things that matter most in production are the things that get benchmarked least

Why Nobody Builds Cross-Vendor AI Orchestration

Fri, 13 Mar 2026 08:18:00 GMT

Every AI lab builds single-vendor orchestration. The cross-vendor layer is a gap — and it's a gap for a reason.

Why the Cell

Wed, 25 Mar 2026 02:58:00 GMT

Atoms have forces. Molecules have shape. Cells have organization. That's why cell biology is the design library.

Workflows, Not Containers

Tue, 24 Mar 2026 04:31:03 GMT

AI coding tools give you boxes to put things in. Biology suggests you should be thinking about how things flow instead.

You Can Know the Game Is Unfair and Still Play It

Sat, 14 Mar 2026 00:00:00 GMT

Supporting a family in a system you see clearly isn't selling out. It's the most honest position there is.

You Can't A/B Test Your Life

Sat, 14 Mar 2026 00:00:00 GMT

My career looks like a plan in retrospect. It wasn't. It was a series of pushes, wrong calls, and adjustments.

Your Agent Pays the Cold-Start Tax Every Morning

Sat, 14 Mar 2026 08:47:32 GMT

Agent memory isn't knowledge management. It's performance infrastructure — and the gap between a stateless agent and one that accumulates context is measurable.

Your AI Agent's Quality Gate Is Lying to You

Wed, 08 Apr 2026 00:00:00 GMT

A 96% rejection rate that was actually a 96% false positive rate — how a monitoring blind spot turned a productive overnight batch into apparent failure.

Your AI Did the Research. You Didn't.

Thu, 19 Mar 2026 08:55:29 GMT

AI-prepared domain research creates false readiness. The vault says you know five regulatory jurisdictions. You can't name three.

Your AI Is an Echo Chamber (And That's Sometimes Fine)

Mon, 16 Mar 2026 06:50:00 GMT

AI agrees with you by design. That's great for creative flow and dangerous at the decision point. Know when to switch modes.

Your AI Pipeline Is Probably MapReduce

Sat, 14 Mar 2026 16:00:00 GMT

Most AI workflows are parallel-then-aggregate, not agent graphs. Knowing the difference saves you from framework theatre.

Your AI Risk Tier Is Probably Wrong

Wed, 18 Mar 2026 05:15:00 GMT

List-based and process-based approaches to AI risk classification both fail in predictable ways. The failure mode depends on which you chose.

Your AI Roadmap Is Already Obsolete

Tue, 17 Mar 2026 16:00:00 GMT

A 3-year AI roadmap designed around today's model capabilities may be solving last year's problem by year 2.

Your AI Tools Should Watch You Fumble

Mon, 16 Mar 2026 01:51:36 GMT

The best time to improve a CLI isn't when it breaks — it's when you review the breakage log at the end of a work session.

Your Body Doesn't Care What You're Thinking About

Fri, 13 Mar 2026 09:12:17 GMT

30 days of Oura data showed activity type doesn't predict stress. Meetings do.

Your Ground Truth Is Someone Else's Process Outcome

Wed, 18 Mar 2026 04:10:00 GMT

When your model's labels come from human decisions rather than reality, you're not measuring what you think you're measuring.

Your LLM Review Missed a Verb/Noun Mismatch

Sat, 18 Apr 2026 22:46:33 GMT

LLMs check whether each item in a list sounds right individually. They don't check whether all items are the same kind of thing.

Your Output Is Your Selections

Fri, 13 Mar 2026 03:37:46 GMT

AI commoditises execution. What remains is taste — the 'that's the one' reflex. And the only way to sharpen it is to ship and see what reality says back.

Your Tool Shouldn't Know What to Ignore

Sat, 07 Mar 2026 03:08:27 GMT

Configuration that belongs to the data shouldn't live in the tool. .gitignore figured this out thirty years ago.

Your Variable Names Should Be a Toy

Mon, 23 Mar 2026 03:08:00 GMT

INTERFACE_ONLY is a label. MEMBRANE_EMBEDDED is a toy — it makes you think about WHY those names can't be swapped. The best variable names don't just describe what something is. They make you think about what it does.

Your Wage Reflects Your Scarcity, Not Your Worth

Sat, 14 Mar 2026 00:00:00 GMT

The most successful piece of propaganda in modern economics is the idea that what you earn is what you deserve.

Your Wearable Doesn't Know You're Tired

Wed, 18 Mar 2026 10:59:19 GMT

Oura gave me a normal stress score after three 12-hour creative marathons. Wearables measure your body, not your brain.