A personal news feed, made public.

Site launched April 2026 · Rebrand April 21 · Editor-in-Chief · Chicago

I was always going to build something like HumanOperator's, but the first version of the concept was wrong. I framed it as "an editorial publication for practitioners." True, but indistinct. The honest version showed up on April 21, 2026 when I rewrote the brief: HumanOperator's is a personal public news feed. Mine. Built around what I actually consume (podcasts, YouTube, Substack, articles) and the thoughts those things trigger. A newspaper made for me, made for people like me. The operating philosophy still anchors the publication, but the model is "Substack × Flipboard × personal journal," not "another editorial publication trying to compete with Stratechery."

The pivot: from publication to personal feed

The original positioning was the safe one. Practitioners write publications; this would be a publication. But sitting with it, I realized I was always going to be the wrong person to maintain that for very long. I'm not a full-time journalist. I'm a builder who reads constantly. Asking "what should the publication cover this week?" produces nothing. Asking "what triggered something in you when you were consuming content this week?" produces material.

So the pivot was three layers, not one:

The mental model from one of those captures: Capture is the comments section turned inside out. Instead of leaving my reaction in someone else's thread where it disappears, I capture it on my own surface where it becomes the seed for my public output. The reader becomes the author. The response becomes the work. This is the operational substance of the rebrand. It required a tool that didn't exist three days earlier to make real.

The operating philosophy still holds

"We are living through the most consequential workplace transition in a generation. And almost everyone is asking the wrong question. The question is not 'will AI replace me?' The question is: where does the human belong?" From the Operating Philosophy, humanoperators.ai/philosophy

That thesis didn't change. The operational shape did. Where does the human belong? In the systems that consume content, react to it, capture the reaction, and turn it into public work. The publication isn't separate from that system; it IS that system, made public. I'm modeling the answer by living it.

I have been that human for 20+ years. I built my first context system at 16 with paper inboxes at a hospital in Oak Park, Illinois, on my birthday, the first day I was legally allowed to work. Quality systems at a telecom startup. SharePoint at a pet insurance company. Supabase and Claude Code today. The tools changed; the problem never did. HumanOperator's is the public artifact of how I'm thinking through the present chapter, and the consumption-and-reaction surface that makes thinking through it visible to anyone watching.

HumanOperator's is not the fear camp. Not the hype camp. It's the practitioner camp: the people who have been inside operational systems long enough to know exactly where the human belongs. And now it's also a model: this is what a personal news feed looks like when it's branded, public, and built on the same desk principles you could apply to your own.

The brand throughline

The publication's identity mark is a typographic signature: ⟨Human⟩ ⟨Operator⟩, set in DM Mono with angle brackets, never decoratively. It appears in the ticker, the masthead, the footer. The angle brackets aren't styling; they're an IP marker. ⟨Human⟩ is a defined role in a system architecture. ⟨Operator⟩ is the practitioner who understands this, not from a framework but from having built the systems that proved it.

The pattern is the positioning. Every piece on the site reinforces a 20-year throughline: systems-builder-at-intersections. Operations veteran who learned context architecture before the term existed. The publication exists to make that pattern legible: first to me, second to readers who recognize themselves in it.

Where it sits in the ecosystem

This isn't the only site. It's one of three pillar sites that share a design system but speak to different audiences:

The three palettes are calibrated by surface temperature. Coolest at analyticgator.ai (institutional). Middle warmth at humanoperators.ai (paper, the publication). Warmest at alfonsoherrada.com (personal: you are meeting someone here). Each site signals its register before a single word is read.

Content from the same source can land on different sites depending on which audience needs it. A piece on context architecture might run as a client case study on analyticgator.ai, an editorial piece on humanoperators.ai, and a portfolio entry on alfonsoherrada.com: three different framings of the same intellectual property, each appropriate to the reader.

The editorial structure

HumanOperators.ai is structured like a newspaper, not a blog. Six beats, each with a color, each with a voice register:

Each beat color appears as a 10px horizontal rule above story cards on the homepage. The ticker above the masthead pulls the most recent piece per beat. The masthead itself is centered (the one editorial-convention exception to "never center everything"), with the publication mark and dateline inline.

How a piece gets made

This is the meta loop. HumanOperators.ai isn't published manually. Every piece flows through the Content Desk pipeline: five Claude Code skills that run the full Know → Position → Create → Voice-Check → Proof → Publish → Learn loop. Concretely, for a Founder's Desk piece:

1. /desk-story: Coach Me for content. Asks one question per turn, story-first, queries Open Brain for related stored thoughts. Outputs elicited material: the raw narrative, my actual phrasing, the moments the piece will turn on.
2. /desk-write: Generates the full draft from the elicited material. Platform-aware (Substack vs. LinkedIn vs. site card). Structure: scene → connection → pivot → insight → forward close. Runs an internal voice check before presenting.
3. /desk-voice: 15-point quality gate. Scores A1-D2. Flags AI slop, banned phrases, thesis-first openings, generic claims. Returns Ready (13-15), Revise (10-12), or Redraft (<10) with line-level feedback.
4. /desk-publish: HIL gate at every platform action. Confirms before each step. Canonical order: Substack → LinkedIn → X → Bluesky. Nothing ships without explicit approval.
5. /desk-archive: Post-publish. Archives to the Content Desk wiki, enriches Open Brain, logs to publishing timeline. Flags Career Desk bullet candidates if the piece contained achievement evidence.

The dual-output principle is what makes this not just a content workflow: every session produces both content and system training. A Founder's Desk piece about Career Desk is a piece for HumanOperators readers AND a bullet candidate for Career Desk resumes AND a story enriched in Open Brain for the next piece to draw on.

The site itself

The site is static HTML on Vercel. Four pages: index, archive, philosophy, shop. The build is intentionally light: no CMS, no React framework, no build step. Editorial discipline doesn't need infrastructure; it needs voice and design discipline.

• Linen palette. Middle warmth in the ecosystem. Background #f0ede6 reads as paper. Cooler than alfonsoherrada.com (personal), warmer than analyticgator.ai (business). Appropriate for a publication.
• Ticker above the masthead. Full-width --ink background, DM Mono 9px. Editorial beat badges scrolling. The publication's heartbeat, rendered as a real ticker, not a marketing hero.
• Beat color system. Five beat colors do most of the navigation. Once you've read three pieces, the color tells you the register before the headline does.
• Card grid with featured spread. Lead story occupies two columns. Sidebars run smaller cards. Beat color appears as a 10px horizontal rule above each card title: typographic, not decorative.
• Playfair Display + DM Mono + DM Sans. Same fonts as the rest of the ecosystem. Single typographic system across three sites with three different palettes.

The build, week by week

What's built

What I'd do differently

• Sit with the brief longer before building the site. The pivot to "personal news feed" should have been the original framing. It's the more honest version of what I was always going to maintain. Building the site under "publication for practitioners" first meant I locked in editorial register, beat structure, and 10 launch pieces under a positioning I rewrote three weeks later. Site design held up; the launch content needs another pass.
• Build Capture before the launch content batch. The 10 launch pieces were drafted from blank prompts. Half of them would have been better if they'd started from real captures of things I was actually consuming and reacting to in real time. The order should have been: pivot → Capture → 30 days of capturing → first launch piece drawn from captures.
• Don't draft the launch batch in one sitting. Ten pieces written in a single push reads in a different cadence than ten pieces written one per week with reader feedback baked in. The launch batch will need a second pass through Content Desk's voice gate before any of them ship. Pieces that felt strong at week one of writing now read as too prescriptive, too thesis-first.
• The shop page launched too early. Built /shop alongside the rest of the site, but there was nothing to sell yet. Empty shop pages signal "trying to monetize too early" — the opposite of the publication's positioning. Corrected May 4 by hiding it from public navigation across every brand surface; the page stays on disk as a sandbox for the eventual AI-native HIL-principles shop concept (year-2 R&D thread, not a launch item).
• Wire the archive page to real data earlier. /archive is a placeholder grid. As pieces ship, archive needs to be live the same day so the publication looks active. Should have set up the archive's data flow during the design week, not after.
• Don't underestimate the time between "site live" and "first piece live." The site shipped weeks ago. The Operating Philosophy shipped weeks ago. The first Founder's Desk dispatch is still in queue because the voice gate is strict. That's correct, but I should have planned for it instead of being surprised by it.

What unlocked the speed

• Three pillar sites with one design system. Master DESIGN.md + ALIGN.md + VOICE-GUIDELINES.md in desk-assets/. Site-specific DESIGN.md in each repo documents only the deltas. Shipping a new site means inheriting the master and writing a 50-line addendum.
• Static HTML. No framework, no build step, no CMS. Edit the HTML, push to git, Vercel deploys. The right tooling for editorial work is the absence of tooling.
• Content Desk in place before HumanOperators. The voice gate, CONTENT-STANDARD.md, the 5 voice signatures, the 6 content pillars, all written before the first HumanOperators piece. The publication didn't have to invent its own quality bar; it inherited one.
• The philosophy first. Wrote the foundational piece before the design. The design followed the writing ("a publication, not a blog → newspaper structure → ticker + masthead + beat colors"). When the design followed the writing, every design decision had an editorial rationale. The reverse usually fails.

Editorial designNewspaper layoutStatic HTMLVercelContent Desk pipeline⟨Operator⟩Brand IPVoice gate