I Couldn't Sit Through My Own Video — So I Distilled Nine Books for a Few Bucks

Let me start with something embarrassing: I rewatched one of my own videos recently and couldn't finish it. Two and a half minutes of slow, filler-heavy talk. My own content.

We treat content as a product, so our first video is still publicly up. This piece covers what we learned iterating on content — and how that led us to something much bigger: FDE (Forward Deployed Engineering).

Five content lessons, learned the hard way

1. Cut the filler. I'm naturally wordy, and AI-assisted editing made it worse. I couldn't bring myself to delete my own words, so I added a "filler inspector" role to our crp-content skill and let AI do the cutting.

2. Thumbnails need one selling point, not ten. V1 was handmade by me — "good enough, ship it." V2 was GPT-generated: prettier, richer. Then I realized rich is bad. Thumbnails live in tiny recommendation windows; nobody reads small text. V3's rule: the headline is as big as possible. That's it.

3. Build content like a product. I don't chase perfection. Ship first, publish first — but when you get positive signals, reflect and polish. Content formats are iterated, not designed.

4. Tune speaking pace with data. My old audience was seniors, so slow pacing never surfaced as a problem. Once I started talking about AI, it was painfully slow. Don't trust your own filter — clip a one-minute video from a benchmark account, calculate their words-per-minute, compare it to yours. Can't change your natural pace? Change the playback speed. I landed on 1.1x.

5. Stop lecturing. My biggest flaw. Nobody opens an entertainment app to be lectured, and lecturing produces zero revenue. Cut filler, cut preaching.

Content is practice. FDE is the real game.

In September we did something tedious: mapping out where a company actually spends money on external service providers today. After many rounds of discussion, enterprise AI transformation — FDE — landed in front of us.

Our hand isn't bad: we're an AI-native company, everything is self-built, AI capability packaging is already done, and I've touched most parts of the business personally. Plus, AI transformation projects generate long- and short-term outsourcing demand, which feeds back into our talent platform.

What we lack is methodology for doing it for someone else — and there's no Chinese-language FDE course worth taking. So September's plan has four tracks:

1. Build the knowledge base. Map the full FDE lifecycle, distill the untranslated foreign material into skills — like hiring nine FDE mentors.
2. Build a diagnostic tool. Package pieces of the skills into an enterprise AI-readiness diagnostic. Don't sell transformation; sell diagnosis first.
3. Pre-launch content. Document the whole process — including this article — so the dev cycle isn't dead time.
4. Build a case study. Pick an industry with budget, take one company's public materials, and do an anonymized AI-transformation demo.

Nine books, a few dollars, one battle manual

The most counterintuitive thing about FDE: the core skill isn't writing code. It's walking into an unfamiliar company and understanding its real workflows, systems, data, and pain points within two weeks.

So we did something dumb and simple: across six dimensions — FDE methodology, business fundamentals, operations, AI engineering, enterprise tech environments, and organizational change — we picked nine books and distilled each one into a skill using book-to-skill.

The cost is absurdly low: each book takes roughly 500k–1M tokens, a few dollars. Nine books in, what we got wasn't nine "I've read this" memories — it was a ready-to-use battle manual: 7 sub-skills covering the full lifecycle of an enterprise AI project:

1. Discovery — the client's real processes, systems, data, pain points, stakeholders
2. Opportunity design — AI vs. no-AI, value, feasibility, risk, prioritization
3. Business case — aligning ROI and resource commitments, go/no-go
4. Solution architecture — enterprise data, systems, security boundaries, deployment constraints
5. Production ops — readiness review, monitoring, incident response, rollback
6. Capability transfer — getting the client's team to actually run it, knowing when FDE exits
7. Field learning — failure patterns, reusable capabilities, expansion candidates

The full version involves book copyright and compliance, so we may not be able to open-source it all. But a stripped-down version is coming — without the nine books underneath, but with a skeleton you can use immediately. Want to replicate the full thing? The recipe is this article: a few dollars per book, plus taste in book selection and patience in distillation.

The first customer of this skill set is ourselves: we're using it to reverse-analyze our own workflows and turn the front-loadable parts into a diagnostic tool on our site. Two birds — validating the skills and grounding the product roadmap.

From cutting filler out of my own videos to distilling nine books into methodology, the underlying logic is the same: build content like a product, and stockpile capability like content.

One last word — business owners, look this way. The theoretical ammunition is loaded, our own projects are fully AI-native. More technical than sales, more product-minded than engineers, more operational than PMs. Barring surprises, we'll be the top FDE platform you come across.