It’s a Tuesday afternoon in my last quarter at Belkins and I’m watching an analyst on my team email the same investor update for the third week running. PDF attachment. The numbers in the PDF were true the second she exported it and started rotting on the way to the inbox. The investor never opened the previous two; she’s about to make the same try with a third copy of yesterday’s numbers stapled into a static frame nobody will scroll. I stop her mid-send.
The thing she should have sent — the thing every analyst, every CSM, every junior consultant, every founder is sending right now — is a link.
That’s the whole chapter. Stop sending dead files. Every deliverable — report, pitch, audit, deck, financial model, competitive teardown, contractor brief — ships as a live interactive HTML artifact in a private repo, deployed to a link. You don’t send the numbers. You send the link. The link is current because the repo is. Next week’s update is a commit, not a re-export and a prayer that they open the newest attachment instead of the one from two Tuesdays ago.
The change is the medium, not the content#
I rolled this across the portfolio and the surprise wasn’t “nicer reports.” It was retention and circulation. An interactive doc with the source attached gets opened, gets clicked into, gets forwarded. A PDF gets archived unread. The medium changes the behavior — you can put a slider in it, you can let the reader filter the catalog, you can make them move the assumptions and watch the number move. Nobody forwards a spreadsheet. People forward a thing they got to play with. That’s not aesthetics, that’s distribution, and distribution is the only thing that was ever scarce.
It’s also faster to make than the dead version. A deck is hand-built, slide by slide, by a human at midnight. An interactive HTML artifact is a single file an agent writes for you in the time it takes to argue about the title font. The cheaper, better, more current artifact also happens to be the one that takes less of your night. There is no tradeoff here. That’s rare enough that you should be suspicious you’re missing something — and you’re not.
And the medium itself is the signal. A live link says operator. A PDF named Deck_v7_FINAL_final.pdf says the opposite, before anyone reads a word. The recipient hasn’t opened either; they’ve already decided which one was made by someone serious.
The one-to-two gap#
The first deliverable you ship this way is a curiosity. The second one is the format you can’t go back from. I’ve watched this happen across five portfolio companies in the same shape every time. Pitches go first because no compliance team has to approve an investment doc that doesn’t exist yet. Audits go next because the methodology hits a wall the moment a client wants to “see what happens if we change one assumption.” Internal QBRs follow because the format is the answer to “the board keeps asking the same three what-if questions.” By the fourth or fifth deliverable in this shape, nobody on the team is exporting PDFs anymore. The format that won wasn’t decreed; it spread because the better way was obviously better.
That’s the Chapter 26 point pointed at a document instead of a tool. When something is genuinely better, you don’t roll it out. You ship one and the second one starts spinning up next to it without you telling anyone.
Why this is its own chapter#
It’s also the same idea as the living link from Chapter 19, taken to its conclusion. Chapter 19 is “build a product on Saturday.” This chapter is “every deliverable is a product.” The mechanic is identical — private repo, single HTML file (or set of files), deploy to a link, share. The only thing that changes is what’s in the file. A Next.js app, a one-page audit, a board memo, a brief, a deck, a model — same recipe, same pipeline, different content.
The Playbook itself is the maximal version. You’re inside the case study right now. Edition 6 of this book added three embedded case studies (the AFC pitch, the AFC robot stable, the Folderly audit), Edition 7 made the launch its own artifact at /launch, and Edition 8.1 made the launch week its own daily-updating artifact at /launch-week. The maximal version is also the minimal version. A Tuesday morning report is one HTML file. A 39-chapter book is many HTML files. Same pattern, both ends of one ladder.
The cost math nobody priced in#
A PDF deck assembled by a human takes three to five hours per quarter to refresh. The interactive equivalent takes me twelve minutes to commit and twenty seconds to deploy. Multiply across a portfolio with eight to ten recurring deliverables (board updates, investor newsletters, customer QBRs, monthly metric reports, the weekly KPI digest, the quarterly all-hands prep, the offsite memo, the sales-team enablement deck) and the time-savings is one full FTE of saved hours per year. That’s not a productivity story; that’s a roles-and-headcount story you can see in your finance pages.
The deeper saving is on the second-order work. A live artifact updates from a data source; the static deck doesn’t. The hours you used to spend “refreshing the deck for the board call” because the numbers shifted between when you exported it and when the call happened — those hours are gone. The artifact is already current. You spent zero time on the refresh because there was no refresh; the link is the live read.
If you want the cost-per-task math at the prompt-and-token layer, Chapter 29 has the four-cost picture and the cache hit rate that makes this run economically at scale.
Where the format fails#
I’d be the only person publishing a chapter on a format if I didn’t say where it doesn’t work. Four real edges.
Documents the receiver legally needs as a PDF. Court filings, regulatory submissions, signed contracts, tax filings, insurance claims. The format is the format. Don’t htmlize the LLC operating agreement. Reserve the link for the brief that decided you needed the contract.
Anything that has to be e-signed. DocuSign and the rest are doing legal work, not communication work. The artifact you ship is the one that explains why the contract — that’s the operator move. The contract itself is still PDF.
Recipients who don’t open links. Some clients won’t click. Send the file. Then htmlize the next one for the same client and put the link in the email body as “live version with the assumptions you can move” — by the third deliverable, half the holdouts are clicking.
Internal deliverables where the deploy overhead beats the value. A fifty-word Slack ask doesn’t need a deployed webpage. Reserve the format for what gets forwarded; if there’s no distribution problem, there’s no format problem to solve.
The longer field guide on this — including the recipe to run a deliverable through HTML-ization on a Saturday, the twelve applications that work, and the public examples in the wild from Stripe / Pudding / Ciechanowski / AI 2027 / Anthropic / NYT / Linear — lives at /html-first.
What it asks of you#
The change is small and the discipline matters. Three things have to be true for an operator to actually live this way.
One — you write your deliverables in a format that deploys. Not Word. Not Pages. Not PowerPoint. Markdown becomes HTML. HTML is the universal target. The drafting environment becomes a text editor and a chat with Claude — that’s it. Anything else is friction between “the thought” and “the link.”
Two — you have a place to deploy. A private GitHub repo is the cheapest, most durable, most operator-respectable home that exists. The /github-playbook page is the path in if that part of the stack is new to you — five things you can do on GH without writing a line of code, the eight gh commands you actually need, what you can safely ignore.
Three — you build muscle on the format. The first artifact takes an evening. The second takes a morning. The fifth takes ten minutes. By the twentieth, you’ve forgotten the format you used to use. The skill compounds against the deliverable type — once you’ve shipped one QBR as an artifact, the next QBR is faster than the deck version was, every quarter, forever.
Do this Monday#
Pick the deliverable you’ve sent most recently — the board update, the customer QBR, the strategy memo, the contractor brief. Don’t send the next one as an attachment. Spend forty-five minutes and ship it as a link.
The mechanic from Chapter 19’s Hour 2 pointed at a document instead of a product: gh repo create --private, drop the HTML file in, vercel deploy (or GitHub Pages), share the link. Updated by commit, current forever, forwardable.
The first one is the curiosity. The second one is the format you can’t go back from. The thirteenth is the year your finance pages started showing the time-savings the rest of the team noticed before the spreadsheet did.
The deep version of everything in this chapter — the recipe, the prompt template, the twelve applications gallery, the four anti-cases, and the twelve public examples in the wild — is at /html-first. The story is here; the playbook is there. The two together is the whole move.