TE
Travis Eric
HomeNow
Work with meBook a 30-minute call
Travis Eric LogoTravis Eric

AI consulting and custom software, based in Fort Collins. Explore the work. Bring your next idea.

Travis Eric AI Consulting(970) 372-1973

Navigate

  • Home
  • Now
  • Projects
  • How I work

Deeper

  • Work with me
  • About
  • Share your experience
  • Case Studies
  • Resources
  • Impossible Vision
  • Technical Stack
  • Manifesto
  • Writings

Follow the work

New systems and field notes, sent when they ship.

Communities

AI Builder's LabView membershipTeneo Publishing CollectiveView details
@TravisEric_Travis Eric@TravisericLive Builds@travis_eric
PrivacyTerms© 2026 Travis Eric
/
/
Back to writings
Skills and playbooks

section-pass — one agent per section, the details put in

Travis EricSeptember 16, 20262 min read
Share this post:

When I ran the overnight build system using section-pass, I realized just how critical this tool is for refining long pages. The tool's genius lies in its methodical approach: after you've got your draft, you go back in with a laser-focus on each section. This involves assigning one agent per section, each armed with a fixed question set and an exhaustive proof inventory.

Here's how it played out. First, I assembled the necessary inputs before spawning any agents. I exported the page as text, sectioned it appropriately — in this case, using the markdown export /brand-pass/deep/<slug>/report.md for a DeepReport page. This gave me a clear structure to feed into the system.

The next step was identifying the reader. Rather than thinking in terms of demographics, I nailed down one specific individual: what they value, what they asked for, and what they will check first. This level of specificity is crucial, and it was documented as part of the input process.

The proof inventory was a beast, but an essential one. Each item had to be logged — its purpose, honest assessment, and, critically, a public link. This inventory was extensive, drawing from various sources like ProductOS/AppsOS/APP-PORTFOLIO-BOARD.md for apps, and the Live Integrations table in CLAUDE.md. Anything without a public link was off-limits for claims, ensuring accountability and transparency.

I then defined the rules for the page, which included avoiding certain claims or jargon the reader doesn’t use. These rules were guided by PITCH-PAGE-PRINCIPLES.md in docs/features/brand-pass/, which helped maintain consistency.

Now for the pass itself. One agent was spawned per section, each working in parallel yet blind to the others. This ensured focused rewrites without cross-contamination. Each agent received its section’s text, the entire page for context, the reader profile, the proof inventory, and the set questions.

One section came under scrutiny when the agent determined its purpose was merely "transition". It was flagged as a candidate for removal, highlighting how section-pass can trim the fat. Another agent identified a missing piece that the reader wanted, which was cross-referenced with the reader's profile to ensure relevance.

By morning, the page was leaner, more precise, and densely packed with value. The overnight run of section-pass transformed what was a sprawling draft into a tight, reader-focused narrative. If you're building an autonomous business and need your pages to speak directly to a specific audience, section-pass is an indispensable tool for that second, crucial pass.

Join Pattern Insights

Get weekly insights on AI architecture, pattern recognition, and building platforms without permission.

Tags:
build-in-public

Read next

September 12, 2026

The Brain and the Heart — why your AI needs both

AI is shaped by what people could measure. Taste is what they could not. Here is the structure I built to give a model both: a Brain that holds the facts and a...

Read it

September 11, 2026

Insights harvest — when the advice is already in your repo

A tool read months of my sessions and recommended a list of things to build. Most of them existed. The gap was never invention, it was enforcement.

Read it

September 3, 2026

Definition of Done — the run-truth ladder

Four rungs that say what done actually means — built, deployed, works, fed — and the seven ways I have watched a check lie about which rung it was on.

Read it

Join the Discussion

Have thoughts on this post? I'd love to hear them! Join the conversation on X where we can discuss AI architecture, pattern recognition, and building platforms.

Discuss on X

Or reach out directly at @TravisEric_