Chaos Monkeys

An insider memoir covering roughly 2008 to 2014: a physics PhD working on Wall Street pricing derivatives, moving to an advertising startup, founding his own, going through an accelerator, selling to one company while negotiating with another, and then working inside a very large one.

Its value is specificity about processes usually described in general terms — how an acquisition negotiation actually proceeds, what an accelerator does, how advertising targeting is built.

Its value is specificity about processes that are usually described only in general terms.

Devmark’s reading — not a quotation from the book

The most genuinely useful content. Martínez explains real-time bidding, exchanges, audience targeting, attribution and the data infrastructure behind them with a clarity that the industry's own literature avoids.

His account of what data is actually used, how identity is resolved across services, and why the economics push toward more invasive targeting is more concrete than most critical writing on the subject and considerably more credible for coming from a participant.

His account of founding a company is deliberately anti-mythological: funding is a status game, accelerators are partly filtering mechanisms, term sheets contain provisions founders do not read, and acquisitions are frequently about the team rather than the product.

His description of running two acquisition negotiations simultaneously, and the ethics of it, is the book's most-discussed passage and he does not present himself well in it, which is to his credit.

His chapters on working at scale describe the political reality of a product organisation: competing internal teams, the difficulty of shipping anything, and metrics that shape behaviour in unintended directions.

His account of how advertising product decisions were actually made — the tension between user experience, revenue and the stated mission — is the material that gives the book its documentary value.

Chaos monkeys refers to the software practice of deliberately introducing random failures into a production system to test resilience.

His extension of the metaphor to founders — as agents of deliberate disruption released into functioning industries — is doing more rhetorical work than analytical, but it is the organising image and it explains his own self-presentation.

Founders as deliberately released failures, testing whether an industry is resilient. Rhetoric rather than analysis.

Devmark’s reading — not a quotation from the book

It must be addressed because it is the main reason people stop reading. The book is crude, frequently misogynistic in its descriptions of women, and the narrator is deliberately unlikeable.

Some of this is self-aware performance and some is not, and the distinction is not always available to the reader. Several passages describing colleagues and former partners were widely criticised as gratuitous, and that criticism is fair.

It is one participant's account, written for effect, about named people and companies. Several individuals disputed characterisations, and the author was reportedly subject to legal pressure over the book's contents.

The technical explanations of how advertising systems work are the most trustworthy content; the accounts of specific meetings and motivations are the least.

Its lasting value is as a primary source on a particular period — the years when social advertising infrastructure was built, before the public argument about it began in earnest.

It records what the people building it understood themselves to be doing, which is difficult to reconstruct from outside and which later accounts tend to smooth.

Against Zuboff it is the view from inside the machine: less analytical, much more specific, and useful precisely because it lacks her critical frame.

Reading them together is instructive — Zuboff explains what the system is for, and Martínez explains how it was actually assembled and by people with what intentions.

The advertising technology explanation, which is the clearest available to a non-specialist and is the foundation for understanding almost any argument about data and platforms.

And the unromanticised account of startup mechanics, which is a useful corrective for anyone whose understanding of the industry comes from its own publicity.