I have a problem most scientists don’t think about until it’s too late: I need to communicate what I do across LinkedIn, Twitter, TikTok, Instagram, YouTube, and a personal website — and it all needs to sound like me. Not a polished version of me. Not a marketing-approved version of me. The actual me, with the goofy tilt and the patient anger and the Feynman quotes.
For a while I did what everyone does. I wrote posts when I felt inspired. I forgot platforms for weeks. I’d nail a LinkedIn piece and then realize I hadn’t touched TikTok in a month. The content was fine when it existed. The problem was the gap between what I wanted to say and the operational capacity to say it consistently.
So I built a system. And because I’m an infrastructure person, I over-engineered it. Beautifully.
The Architecture
aDNA — Agentic DNA — is a knowledge architecture I’ve been developing for federated science. It organizes everything into three primitives: modules (atomic capability units), datasets (the data layer), and lattices (directed graphs that compose modules into workflows). The same architecture that lets you wire a DNA sequencing pipeline to a protein structure predictor turns out to be surprisingly good at wiring a blog post to a Twitter thread to a TikTok script.
The content vault is structured as a triad: What (brand knowledge — voice, identity, narrative, visual), How (operations — campaigns, missions, sessions, pipelines), and Who (people — collaborators, community, governance). Every piece of content I produce can trace its lineage through this triad.
Voice as Infrastructure
The first thing I built wasn’t a publishing pipeline. It was a voice architecture. Twelve registers — from “The Joy of Not Knowing” to “The Neighborhood Is the Lab” to “Science Belongs to All of Us” — each with defined purposes, triggers, and constraints. Eighteen anti-patterns that define hard failures: things Stanley would never say. A decision tree for blending registers based on platform, audience, and content type.
This sounds excessive until you realize what happens without it. Without voice infrastructure, every piece of content is a cold start. You sit down, stare at the blank page, and try to remember who you are as a communicator. With it, the question shifts from “what should I say?” to “which register blend serves this audience on this platform?” That’s a tractable question. That’s engineering.
The Campaign Model
I structured the entire build as a phased campaign — 42 missions across 6 phases, from foundation (migrating existing brand context) through ingestion (extracting structured knowledge from transcripts and articles) through quality review (an iterative improvement system that ran 50 review cycles across 10 quality dimensions) through forge wiring (connecting content generation pipelines) to validation.
Each mission decomposes into objectives. Objectives fit in a single working session. Sessions produce artifacts and end with a structured status report. It’s the same OODA cascade — observe, orient, decide, act — that runs a military campaign or a clinical trial. Applied to content.
The result: 20 post seeds generated from raw identity context, each with three narrative angles and platform adaptation tables. A publishing pipeline that routes content through voice compliance checks, format validation, and multi-reviewer quality gates before anything goes live. Five forge wrappers that produce different content types — website pages, video scripts, speeches, presentations, graphic novels — all drawing from the same underlying brand knowledge.
What I Actually Learned
The surprising thing wasn’t that the system works. It’s what the system revealed about the brand itself.
When you force-structure your identity into twelve voice registers and test every piece of content against eighteen anti-patterns, you discover things about yourself you didn’t know. I learned that my default register in public talks isn’t wonder — it’s builder’s invitation. I learned that my use of patient context follows a strict protocol that I’d been applying intuitively for years without naming it. I learned that the mentorship pattern from my DeepChem GSoC work is the same pattern that drives Science Sundays and the World Genome Academy — I just hadn’t drawn the line between them.
The vault didn’t create the brand. It made the brand legible to itself.
Why I’m Sharing This
Science is a collaborative community driven effort. So is content. The aDNA architecture is open-source. The vault structure is forkable. The voice architecture pattern — registers, anti-patterns, decision trees, quality gates — works for any brand that needs to maintain consistency across platforms without losing authenticity.
You don’t need my specific registers. You need your own. But you need them written down, tested, and wired into a system that catches you when you drift. That’s what infrastructure does. It makes the hard thing repeatable and the repeatable thing improvable.
If you’re a scientist, a patient advocate, a community builder, or anyone who needs to communicate complex ideas across multiple platforms without losing your voice — this is a solved problem. Not easily solved. But solved. And the solution is a lattice, not a template.
What can’t you do with that?