The Working Library
Every book here is on the shelf because it gets used — in architecture reviews, governance debates, pricing arguments, or essays. Each carries two lines: the one idea, and where that idea earns its place when you're building systems institutions can trust.
Don't read the library. Read the path that matches the problem on your desk.
Capability is the easy third. The hard two-thirds are operating discipline and knowing which problems deserve you.
Read capability, then institutional drag, then capital cycles. Most bad AI takes skip the second book.
Judgment is the scarce resource when routine cognition gets cheap. This stack is judgment maintenance.
Organised by the job they do, not the section of the bookstore they came from.
The books open on the desk when designing platforms, patterns, and the gates between them.
A practical map for moving from chatbots to agents that plan, act, and improve workflows.
The shared vocabulary for executive conversations about what an agent actually is — before the architecture debate starts.
Simple operating systems prevent failure in complex, high-stakes work — surgery, aviation, finance.
The strongest argument for policy gates ever written. An approval gate is a checklist the system cannot skip.
Taste, courage, and asking the questions that make important work possible.
Read before every platform decision that will outlive you. "What are the important problems in your field?" applies to roadmaps too.
A compact reference of design patterns and cognitive principles.
Agent UX is still UX. Most "model failures" in production are affordance failures in disguise.
An enterprise playbook for turning digital and AI ambition into operating change.
The honest McKinsey answer to why pilots die: the operating model, not the technology, is the bottleneck.
Companies designed around data, prediction, and machine intelligence from day one.
The lens for spotting which vendors are AI-native and which are AI-wrapped — it shows in the data architecture every time.
Capability claims without the conference-stage adrenaline.
Frontier labs, scaling laws, and the economics of intelligence — from the people inside.
The calibration source for timeline talk. When a vendor's roadmap contradicts what lab insiders say out loud, believe the insiders.
The classic argument for taking advanced AI capability, control, and governance seriously.
The control problem at civilisational scale is the control plane problem at enterprise scale. Same shape, different blast radius.
Fast-moving technologies structurally outrun institutions and markets.
Names the gap the Situation Room tracks daily: the distance between what technology permits and what institutions can govern.
Progress comes from good explanations, criticism, and the unbounded growth of knowledge.
The philosophical case for why the knowledge layer compounds — and why explanations, not data volume, are the asset.
When routine cognition gets cheap, judgment is what's left to be good at.
Reusable thinking tools for seeing problems from multiple angles.
The pre-mortem kit for architecture decisions. Inversion alone has killed three bad designs this year.
Mental models, incentives, and the discipline of avoiding obvious stupidity.
"Show me the incentive and I will show you the outcome" explains more vendor behaviour than any analyst report.
What numbers can and cannot tell us about evidence, uncertainty, and risk.
The vaccination against benchmark theatre. Every model evaluation claim should pass through this book first.
Recurring patterns in civilisation, power, economics, and human nature.
Perspective at fifty pages. Useful the night before any meeting that feels existential.
The major philosophical systems and what each means for living well.
The Ledger is philosophical on purpose: you cannot design what an institution should trust without a theory of what trust is.
Eighteen years in banking says the constant isn't the instrument — it's the human holding it.
Money outcomes are shaped by behaviour and time more than intelligence.
Also true of AI budgets. Institutions don't have model problems; they have patience problems.
A macro-history of debt, empires, and the cycles that reshape global power.
The frame for reading sovereign AI investment — compute buildouts are reserve-currency behaviour by another name.
Wealth, leverage, judgment, and independent thinking, distilled.
"Code and media are permissionless leverage" is the entire thesis of writing The Ledger in public.
Durable lessons on saving, discipline, and financial independence.
The book that proves good financial systems are old. Babylonian record-keeping was a ledger with governance. So is ours.
The hidden forces that shape buying decisions.
Enterprise AI procurement is not rational either. Knowing which forces are operating in the room is an unfair advantage.
AI will expose weak institutions before it transforms strong ones. These books explain what institutions actually are.
Shared myths, institutions, and technologies shaped human history.
A bank is a shared story with an audit trail. So, increasingly, is an agent registry. Trust infrastructure is the oldest technology.
Humanity's rise through cooperation, imagination, and power, told for young readers.
On the shelf because the best test of understanding is explaining it to a twelve-year-old. (Field-tested at home.)
Inequality, hierarchy, and why societies become unfair.
The fairness questions children ask are the fairness questions AI governance committees ask. Children ask them better.
Informal power and the lived reality behind urban statistics.
A reminder that every org chart hides a real operating system. Map the informal one before deploying anything.
Identity, prejudice, and the cost of seeing people through categories.
The literary case against classification error — worth holding in mind while building systems that classify people at scale.
These books are load-bearing in the essays — the moat-is-the-knowledge-layer argument, the governance-in-runtime thesis, the 80/6 problem.
Read the essaysThe Ledger occasionally ships a single operating note on one book — what it changed in practice, not a summary.
Join the ledger