The Mythical
Token‑Month
Essays on Building Software with Machines That Write Software
Fred Brooks's The Mythical Man-Month for the era of AI coding agents: which of his insights still hold, which need updating, and which the new tools have quietly made obsolete for engineers, architects, PMs, and the clients who now have to trust all three.
The discipline survives. The instruments got sharper.
Half a century after Brooks named the tar pit, software teams have a new kind of contributor: tireless, fluent, and judgment-free. This book argues that the discipline he described (holding a system's conceptual integrity in one mind, agreeing on what to build, shipping something maintainable) hasn't gone away. It's gotten harder, because the accidental difficulty Brooks spent a career fighting has partly been automated away, exposing how much of the job was always the essential difficulty underneath it.
Written for practicing senior engineers, not executives skimming for a keynote soundbite. The book revisits Brooks's central propositions one at a time (Brooks's Law, the surgical team, the second-system effect, no silver bullet) and asks what each one becomes when one of the “engineers” doesn't sleep, doesn't forget, and bills by the token.
- No hype. Every capability claim gets a mechanism, a date, or an explicit “this is speculative.”
- No fabricated benchmarks or invented statistics: uncertain claims are flagged, not smoothed over.
- Written by a practicing engineer who has shipped regulated ML systems, not a commentator narrating the space from outside it.
- Short-form, essay-driven: closer in length to Brooks's original than to a 300-page business title.
Verbatim, from the manuscript
“Tokens are to software as bricks are to architecture: a necessary input, an inadequate measure, and when produced faster than they can be inspected, a hazard.”Ch. 1 · The Mythical Token-Month
“The natural tendency of generative tools is to produce locally consistent and globally incoherent code. Defending coherence is now an active, continuous, costly activity. It is worth the cost.”Ch. 12 · Proposition 3
“The system has no surgeon, only orderlies, and the patient gets sicker without anyone being able to say exactly why.”Ch. 2 · Teams and the Tireless Resident
“The instruments have changed; the work has not.”Ch. 12 · Closing line
Who this book is for
They rarely sit in the same room, which is part of the problem this book describes.
The Project Manager
Used to estimate schedules by asking engineers and doubling it. That still works, except now one line item is “let the model have a go at it,” and nobody can say with confidence whether that means four hours or four days.
The Architect
Did not apply for the job of AI architect but has it now regardless. Where a model call sits, what happens when it fails silently, what it costs at ten times volume: these are architecture decisions nobody handed them a new title for.
The Developer
Two people, one job description: the senior deciding how much instinct to lend a fluent, confident, occasionally broken tool; and the junior acquiring judgment about code by a route nobody has fully mapped yet.
The Client
Just told, in a proposal or renewal meeting, that the team is now “AI-augmented.” Needs to know what a reasonable supplier can credibly promise, and what they're quietly still guessing at.
Inside the book
Twelve chapters, each opening with a concrete failure mode, not an abstract thesis.
Ch. 01The Tar Pit and Its New Stickiness
Ch. 02Teams and the Tireless Resident
Ch. 03The Attention Economy of the Architect
Ch. 04The System's Soul
“AI components should be islands in a sea of code, not the other way around.”
Ch. 05Estimation and Reality
Ch. 06The Compilation of Prompts
Ch. 07Software as Assembly, Not Authorship
Ch. 08Production as a Place to Live
Ch. 09The Junior Engineer Problem
Ch. 10Tools, Trust, and Business
“Confidence is built slowly and lost quickly.”
Ch. 11The Dissolution of “the System”
Ch. 12Closing: Propositions
“The skills of the senior engineer of 1995 are the skills of the senior engineer of 2026, in greater demand, with greater leverage, and with greater responsibility.”
Read the opening chapter
The tar pit, redrawn for the era of gradient descent, and the argument the whole book is named for.
Continue reading Chapter 1I have a colleague who runs a small consultancy that does data engineering for pharmaceutical companies. She told me, six months into adopting an agentic coding tool, that her team's velocity had roughly doubled. I was about to congratulate her when she added: “and our review queue has roughly tripled.”
Code that used to be read once, by the author, before being submitted, was now being generated by a machine, skimmed by the author, submitted, reviewed by a peer, returned with comments, regenerated by the machine, re-skimmed, re-submitted, and re-reviewed. The number of lines committed had risen. The number of lines understood had not, and may have fallen.
Brooks opened his book with the image of the tar pits of the Pleistocene. Great beasts thrashed there; the more they thrashed, the more deeply they sank…
Francesco Gadaleta, PhD
Francesco Gadaleta is a computer engineer and entrepreneur working at the intersection of Artificial Intelligence and the Internet of Things. He holds an M.Sc. in Computer Engineering from Politecnico di Milano and a Ph.D. in Information Technology from KU Leuven.
He is the founder of Amethix Technologies, a company specializing in AI-driven solutions for industrial and embedded systems, and the host of Data Science at Home, a podcast on AI, machine learning, and data science with an audience of practitioners worldwide.
Over the course of his career he has designed and deployed machine learning systems across healthcare, manufacturing, and connected devices, from research prototypes through to production at scale. This book draws directly from that experience.
Get the book
The ebook is available now, direct from the author. Kindle and paperback arrive on Amazon September 10, 2026.
Ebook bundle
- PDF & EPUB, no DRM
- Delivered instantly by email
- Free updates if the text is revised
- Highest margin goes to the author, not a platform
Kindle edition
- Reads on any Kindle app or device
- Syncs across devices automatically
- One-click checkout with your Amazon account
Paperback
- Printed and shipped by Amazon
- Eligible for Prime shipping
- The version that survives a power outage
International buyer? The direct ebook checkout handles VAT/sales tax automatically at checkout. Prices shown at checkout are final.
Before you buy
What formats do I get with the direct ebook?
Both PDF and EPUB, DRM-free, delivered to your email immediately after checkout. Read it on a laptop, a tablet, or any e-reader that accepts EPUB. Kindle uses its own format, which is why a separate Kindle edition exists on Amazon.
Is this the same text on Amazon and on the direct store?
Yes. Same manuscript, same edition. The only difference is delivery mechanism and, for the direct purchase, format flexibility (PDF + EPUB instead of a single Kindle-locked format).
Was any of this book written with AI?
No. Every sentence was written by the author, one word at a time, in the usual way, then reviewed by native English speakers and copyeditors: the traditional way. No prompt was submitted and no draft was generated for a human to skim and approve: a distinction the book itself argues is not incidental. See the “On Authorship” note in the front matter.
Do you offer review or desk copies?
For press, podcast, or course-adoption inquiries, reach out via Amethix Technologies or the Data Science at Home contact page.