My book has a standing rule: no name and no number reaches the page without a primary source behind it. Transcripts count as raw input, not evidence. Every proper noun gets checked against an authoritative source before it prints. I wrote that rule into the project before I wrote the first chapter, because the book is partly about a discipline: AI systems fabricate confidently, and the engineers who deploy them earn trust by verifying everything.
For version 1.5, I pointed that rule at my own book. Every statistic, every date, every attribution, re-checked against primary sources: press releases, SEC filings, the original posts. Four verification sheets now sit in the book’s research notes.
Here is what the audit caught.
The book’s most-quoted fact was framed wrong. I had written that OpenAI “capitalized its vehicle at $10 billion” on May 4, the same day Anthropic announced its services venture. The primary record says otherwise. Ten billion is the venture’s valuation; the committed capital is more than $4 billion. And the same-day story is only half true: Anthropic announced on Monday, May 4, while OpenAI’s finalization leaked that same Monday, with the formal launch following a week later. Still a remarkable Monday. Not the one I had printed.
Worse: for three editions, I attributed a widely shared essay on becoming an FDE to a person who does not exist. The essay is by Vasuman Moza, founder and CEO of Varick Agents. The name I printed was wrong, and it survived versions 1.2, 1.3, and 1.4 because I verified the essay’s content and never re-verified its author. That is exactly the class of error the book warns about. It is fixed, and the correction is logged in the changelog where anyone can read it.
I am telling you this because the alternative is pretending it did not happen, and because the lesson generalizes: verification is not a phase, it is a habit, and the moment you exempt your own work is the moment it fails.
While I was checking, the field moved. Between early May and July, AWS committed $1 billion to a Forward Deployed Engineering organization. Microsoft announced Frontier Company: $2.5 billion, six thousand embedded engineers. OpenAI raised more than $4 billion for its Deployment Company. Anthropic launched a $1.5 billion venture with Blackstone, Hellman & Friedman, and Goldman Sachs. Salesforce committed to a thousand FDE hires. Indeed listings for the role grew eight-fold in a year. Over $9 billion, committed in one summer, to the job this book describes.
Version 1.5 catches the book up and deepens it. Twenty-nine new sources, most from the Forward Deployed Engineering track at the AI Engineer World’s Fair, where practitioners from Anthropic, Ramp, Sierra, Decagon, Factory, Cognition, and Kepler said in public what this work actually looks like. The new material includes the discovery war stories (a 47-page dashboard spec that became a Slack alert built in 4 hours, once someone watched the work instead of asking about it), the four models of who pays for deployment, the agent-readiness argument, and a market chapter on what I call the manufactured-FDE wave: three different jobs, one title, and how to tell which one you are being offered.
Here is the part I care about most. All nine billion of those dollars serve enterprises. Pods of five or six engineers on 45-day cycles, sold through channels a Fortune 500 procurement office understands. None of it shows up for a founder-led B2B SaaS with a team of 5 to 50, real customers, and no procurement department. That segment has the same gap between what AI could do in their product and what actually ships. It just cannot buy the enterprise version of the fix.
That is the corner I work in, and it is why the book exists alongside a service. The AI Feature Sprint takes the forward-deployed playbook and scales it down: one AI workflow, shipped into your product in 10 business days, with an evaluation rubric, telemetry, and a handoff your team owns. It starts with a $500 Discovery Day, credited toward the Sprint. The first three teams get beta pricing at $5,000; after that it moves to $7,500 to $10,000. If that is your company, the details are at ghanemzadeh.com/ai-feature-sprint.
The book is on Amazon in paperback and Kindle: amazon.com/dp/B0H3VKKG38. The first two chapters are free at ghanemzadeh.com/books. And if you want the ongoing version of both the book and this letter, you are already in the right place. Subscribe, and I will see you next week. [Amazon Link]
— Nasser