What this page is

A public, plain-language record of how this site gets built — the publishing stack, the AI tools used in the workflow, and the rules every article has to clear before it goes live. If you want the short version, it is on the About page. This is the long version with names, versions, and the rules written out.

Publishing stack

TechFuel HQ is a static site built with Hugo (extended, v0.145.0) and deployed to GitHub Pages from the main branch via GitHub Actions. The repository is public at github.com/iBlessi/techfuelHQ.

  • Hosting: GitHub Pages
  • Build: Hugo extended (v0.145.0), --gc --minify on every deploy
  • Source control: GitHub, every change goes through a pull request
  • CDN / DNS: Cloudflare in front of GitHub Pages
  • Analytics: Google Analytics 4 (GA4), no third-party trackers beyond GA4 and Google AdSense
  • Search: Google Search Console for indexing state
  • Domain: techfuelhq.com, registered to the site author

Layouts and styles are written from scratch, not from a third-party theme. The CSS critical-path is hand-tuned for Core Web Vitals and is not modified during AdSense review windows except by deliberate Council decision.

AI tooling — named, with roles

This site is honest about AI involvement. The pipeline:

StageToolWhat it does
ResearchChatGPT Deep Research (OpenAI)Gathers public benchmarks, vendor documentation, release notes, and community-reported issues. Every cited number gets followed back to a primary source URL.
DraftingClaude Opus (Anthropic)Organizes research into article structure and a first prose draft. Writes the bench tables and the prose around them.
Council reviewMultiple LLM seats (Claude Opus, GPT-5.5 Thinking, Gemini 3.1 Pro)Independent dispatches against the same prompt produce verdicts that get logged to /ops/council/. Used for editorial-policy and AdSense-readiness decisions.
Browser orchestrationPerplexity CometDrives multi-tab workflows, reads live pages, dispatches the other agents. Non-voting; cannot publish on its own.
Editorial verificationLowell K. Wood IVReads every article end-to-end. Verifies every measured number against actual bench output. Verifies every hardware claim against gear on hand or a primary-source URL. Articles do not publish without sign-off.

AI does not get a byline on this site. People do. Articles are signed LK Wood IV because Lowell K. Wood IV is responsible for what they say.

Editorial standards

Every article on this site has to clear all of these before it ships:

  1. No sponsored articles, no vendor samples, no affiliate-first writing. When affiliate links exist, they are disclosed at the top of the article and never change which product gets the recommendation. Most articles have zero affiliate links.
  2. No surrogate authorship. No AI is allowed to claim a byline. Editorial responsibility sits with a named human.
  3. EEAT proof block on every article. Every measurement-based article ends with a “How I tested” section naming the specific bench gear and the measurement methodology.
  4. Last-verified date on every article. When the underlying facts change, the article gets updated and the last-verified date moves. When they don’t, it doesn’t.
  5. Real benchmarks or no benchmarks. Vendor spec-sheet numbers are labeled separately from numbers measured on the bench. Made-up benchmarks are worse than no benchmarks.
  6. Disclose affiliate relationships at the top. Where they exist, they are above the fold, not buried.
  7. Cite primary sources inline. Vendor docs, official changelogs, and reputable trade press get linked. Reddit and forum threads get linked when they are the primary signal but get labeled as community-reported, not as vendor-confirmed.

Self-governance for the orchestrator

The Perplexity Comet browser orchestrator that drives the publishing workflow operates under a written self-governance rule set, kept in the public /ops/ directory. The rules in plain language:

  • Live-state-only on outside-world claims. Before stating anything about AdSense status, search-console indexing, GitHub PR state, or live-site behavior, the orchestrator must read it live in the same turn. No quoting from chat memory.
  • No surrogate authorship. The orchestrator never produces text labeled as another model’s output. If a Council seat needs to speak, the orchestrator drives a real dispatch to that model. Otherwise the position is “seat has not yet spoken.”
  • Provenance on every synthesis. Any text the orchestrator writes that could end up in the canonical record carries a header naming the source.
  • Confirmation only for irreversible actions. Confirmation prompts are reserved for publish-to-main, AdSense state changes, DNS changes, identity-data changes, and purchases. Not for routine continuation.
  • Self-incrimination logging. Every time the orchestrator drifts (claims a wrong fact, attempts a forbidden action, oversteps its role), the incident is appended to the public log in /ops/comet.md. Not buried, not euphemized.

The orchestrator is not allowed to publish PRs to the live site without operator approval, author Council verdicts, change DNS, submit AdSense, or make purchases.

Public operations record

The full operations record is at the /ops/ directory in the site repository:

  • ops/changelog.md — every PR with merge SHA and date
  • ops/state.md — current state of the site (active branches, open PRs, AdSense window status)
  • ops/agents.md — the agent registry naming every active thread and its role
  • ops/comet.md — orchestrator self-governance and drift incidents
  • ops/adsense.md — AdSense application state and review-window invariants
  • ops/gsc.md — Google Search Console state and indexing state
  • ops/risks.md — known risks and mitigations, including reviewer-risk vectors
  • ops/council/ — Council verdicts on editorial-policy and AdSense-readiness questions

Anyone can read these files. AdSense reviewers, Search Console reviewers, and curious readers all see the same record. There is no private operations channel.

Contact

Spot something wrong, want to challenge a recommendation, or have a question about how a specific article was produced? Email hello@techfuelhq.com. Every message gets read by the site author.

Last updated: 2026-05-08 by LK Wood IV.