Archive
AI controls, agent systems, banking governance, production practice. 412 essays, newest first.
- AI Controls Architecture
Risk teams know risk. The open problem is designing controls for systems that are non-deterministic, probabilistic, and attackable in natural language.
- Governing Agents the Way Cells Govern Themselves
Six cell biology mechanisms that reveal what the networking 'control plane' metaphor misses about governing AI agents.
- The Risk Without an Engineering Solution
Every other agentic AI risk has an engineering answer. Prompt injection doesn't. That changes everything about how you design controls.
- Your AI Is an Echo Chamber (And That's Sometimes Fine)
AI agrees with you by design. That's great for creative flow and dangerous at the decision point. Know when to switch modes.
- Enterprise AI Agents: The Transformation Is Organisational, Not Technical
The companies that win with AI agents aren't deploying the most agents — they're redesigning their organisations to work with them.
- Honesty as Default, With One Exception
A two-tier honesty framework: be honest by default, override only when truth would harm someone vulnerable.
- Compounding: The Only Mental Model
If you could only keep one mental model, keep compounding. It applies to skills, reputation, writing, and tools.
- Over-Capture, Then Cull
Don't filter during capture. Capture is cheap. Ideas are expensive. The cull is where quality happens.
- Hooks Are Life Infrastructure
Event-driven hooks in AI coding tools aren't just for linting — they're programmable triggers for life routines, habits, and systems.
- Composure Is a Skill
Rushing is a habit, not a response to reality. You break it by deliberately not rushing when you could.
- Your AI Tools Should Watch You Fumble
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.
- Mining Your LLM
Your AI already knows things that would make it better at helping you. The trick is extracting that knowledge and making it permanent.
- Building a Bus Alert System in One Session
How a real need on a Hong Kong bus turned into a GPS-powered alert system in under two hours
- The Debate Round Is Where Value Lives
Independent parallel reviews produce overlapping findings. The cross-critique round produces resolution. That's where multi-agent value actually emerges.
- The Confidence Trap
Why the thinkers who make you feel like hard questions are resolved deserve the most scrutiny.
- When LangGraph Earns Its Keep
LangGraph is the SAP of agent orchestration — powerful at scale, overkill for most. Here's the line.
- Your AI Pipeline Is Probably MapReduce
Most AI workflows are parallel-then-aggregate, not agent graphs. Knowing the difference saves you from framework theatre.
- The Expert Illusion
Why 'you are an expert' is the most popular and least useful prompt engineering technique
- Planning Needs Eyes
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.
- What If Your Vault Had Residents?
Not tools that search your notes — personalities that live in them, form opinions, and disagree with each other.
- Put the Rule Where It Fires
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.
- What Human Memory Teaches AI Agents (and What It Doesn't)
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.
- When to Make Your Pipeline Agentic
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.
- The MTEB Leader Barely Beats a Free Model on Agent Memory
I benchmarked 10 memory backends and multiple embedding models on actual agent memory retrieval. The results challenge common assumptions about what matters.
- China's AI Stack Is Now Hardware-Deep
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.
- AI Vendors Are Not Neutral Infrastructure
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.
- Three APAC Regulators Are Converging on AI Governance — Banks Should Build One Framework
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.
- The Agent Governance Gap Is Already Here
Agentic AI isn't a future governance problem — it arrived ungoverned, and this week saw the first enforcement action.
- Your Agent Pays the Cold-Start Tax Every Morning
Agent memory isn't knowledge management. It's performance infrastructure — and the gap between a stateless agent and one that accumulates context is measurable.
- The Knowledge That Disappears When You Try to Capture It
Enterprise AI keeps promising to capture institutional knowledge. The most valuable kind resists capture by design.
- What AlphaSense Charges Ten Thousand Dollars For
I built an AI landscape intelligence pipeline for zero marginal cost. Here's what it does and what it can't.
- The Eval Gap
The scarce AI skill isn't building — it's knowing if what you built actually works.
- The CLI Boundary
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.