By Gergely Orosz, the author of The Pragmatic Engineer Newsletter and Building Mobile Apps at Scale
Navigating senior, tech lead, staff and principal positions at tech companies and startups. An Amazon #1 Best Seller. New: the hardcover is out! As is the audibook. Now available in 6 languages.
There were hints of field use. The log’s operator codes matched names in the personnel database: contractors and a handful of government engineers whose last recorded assignments involved moving legacy infrastructure off support lifecycles. One entry, dated three years prior, listed an operator as “OBS1” and the outcome as “observed.” In the margins of the PDF, beside the min_install() function, a final note read: “Observation protocol: record anomalies; do not attempt rollback. Inform Registry JUR immediately if state persists.”
When Lena mounted the drive, the directory structure was sparse and purposeful. A lone PDF, a script, and a short log file. The PDF’s first page bore a stamp: JUR Department — Confidential. The header read “ENGSUB — Conversion Protocol v0.20006.” Below it, a terse sentence: “Minimum install required for legacy conversion.” The rest was a marriage of technical precision and bureaucratic omission: diagrams of connector pins annotated with shorthand, code snippets in a language that slotted somewhere between an embedded assembler and a markup dialect, and a checklist that moved from “verify power rail (3.3V nominal)” to a single ambiguous line: “Observe: convert020006.” jur153engsub convert020006 min install
And yet the warnings persisted. An engineer’s scrawl had become a warning: “Do not run without observe flag.” Someone had learned the hard way. The registry, in this telling, was not only an archive but a safeguard: ensuring that devices could testify to the exact process that brought them into a new operational state. Without that testimony, machines could drift into behaviors that mimicked deliberate action while being byproducts of earlier, undocumented conversions. There were hints of field use
She located an archive entry referencing “jur153” in a decommissioned internal wiki. The entry was sanitized, stripped of the most sensitive diagrams, but the redactions only widened the mystery. In a comment thread, an engineer months earlier had posted one line: “We tried the minimal path, but conversion 020006 introduces ghost states in legacy controllers. Observers required.” The post had been closed by an administrator with the single-note rationale: “See protocol.” Inform Registry JUR immediately if state persists
She copied the files to a secure archive and wrote a short report: the protocol worked; observe changed outcomes; registry connectivity mattered. But the report was clinical; it didn’t capture the small, uncanny moments when a machine’s logs answered like an echo. In the margins of her notes she wrote what the engineer’s scrawl already had: “If you must run it, watch closely. The machine will remember you back.”
The book is separated into six standalone parts, each part covering several chapters:
Parts 1 and 6 apply to all engineering levels: from entry-level software developers to principal or above engineers. Parts 2, 3, 4 and 5 cover increasingly senior engineering levels. These four parts group topics in chapters – such as ones on software engineering, collaboration, getting things done, and so on.
This book is more of a reference book that you can refer back to, as you grow in your career. I suggest skimming over the career levels and chapters that you are familiar with, and focus reading on topics you struggle with, or career levels where you are aiming to get to. Keep in mind that expectations can vary greatly between companies.
In this book, I’ve aimed to align the topics and leveling definitions closer to what is typical at Big Tech and scaleups: but you might find some of the topics relevant for lower career levels in later chapters. For example, we cover logging, montiroing and oncall in Part 5: “Reliable software systems” in-depth: but it’s useful – and oftentimes necessary! – to know about these practices below the staff engineer levels.
The Software Engineer's Guidebook is available in multiple languages:
You should now be able to ask your local book shops to order the book for you via Ingram Spark Print-on-demand - using the ISBN code 9789083381824. I'm also working on making the paperback more accessible in additional regions, including translated versions. Please share details here if you're unable to get the book in your country and I'll aim to remedy the situation.
I'd like to think so! The book can help you get ideas on how to help software engineers on your team grow. And if you are a hands-on engineering manager (which I hope you might be!) then you can apply the topics yourself! I wrote more about staying hands-on as an engineering manager or lead in The Pragmatic Engineer Newsletter.
I've gotten this variation of a question from Data Engineers, ML Engineers, designers and SREs. See the more detailed table of contents and the "Look inside" sample to get a better idea of the contents of the book. I have written this book with software engineers as the target group, and the bulk of the book applies for them. Part 1 is more generally applicable career advice: but that's still smaller subset of the book.