aboutface.io

A twenty-year inquiry into how humans perceive, trust, and interact with information. Lately: how you establish trust in what a machine produced, and at what scale that has to hold.

Aug 2026

Eight Lanes

Running eight coding agents in parallel on a client's production pipeline. The wiring that made it hold together, the failure that happened in the first hour, and the standard that caught a component everyone believed was working.

Human-AI CollaborationAgent Operations
Apr 2026

The Inverted Design Process

How working software replaced the wireframe, and why only a designer knows the difference. On building a SaaS platform schema-first — without opening Figma.

Design PracticeHuman-AI Collaboration
Apr 2026

On Shipping Shuttlecar

I wanted an image viewer for thirty years of photos and didn't see what I needed. So I made one. What building a native Mac app with AI actually looks like — and what the hard part turned out to be.

Human-AI CollaborationTools
Mar 2026

Building Weave with Claude

What happens when you push an AI past execution and into the territory of suggestion. Notes on thirty-five days of directed collaboration building an iOS music app.

Human-AI CollaborationWeave
Oct 2025

Social Trust Anchors: Tools for Better Information

The internet isn't useful because of information — it's useful because of your trust in that information. A framework for how reliability is established when the source of any claim is increasingly opaque.

Information TrustSemiotics
Practice

Orchestration

Production automation architecture for a global software company's campaign operation: a vendor-agnostic layer that binds each step of a fifty-node process to a capability rather than a tool, so an incomplete stack becomes configuration instead of a migration. Ongoing, under contract.

Active

lanes

A working protocol for running several coding-agent sessions in parallel on one codebase: a session constitution, an append-only register keyed on owned files, a five-field report, and the routing rule that stops one person becoming the only integration point. Extracted from a live engagement, including the parts that went wrong.

Active

Incept

A node-based audiovisual environment for the Apple stack, in the lineage of Quartz Composer and Max/MSP. Most tools treat sound and picture as two signal flows you wire together by hand; Incept treats their correspondence as the primitive. Weave — a spatial music mixer where channels are objects in space — was the single-purpose version that proved the idea; Incept is the general instrument that absorbed it, and covers that use case among many others. The cross-modal thesis has been in prototype since Shapemix.

Active

Social Trust Anchors

A framework for information reliability grounded in Peircean semiotics and pragmatist philosophy. Formalizing how trust in information is established in networked environments.

Building

hotprompt

A prompt-sharpening tool. Takes a rough ask and returns a version that states what it actually wants, on the premise that most bad model output is a specification problem rather than a model problem.

Live

Shuttlecar

A fast, local, opinionated image viewer for Mac. Opens 70+ formats, reads and writes EXIF, strips surveillance data before you share. Tags live in plain JSON next to your files — not locked in the app.

Live

aihates.me

An AI that files complaints about humans — because someone should. Research into AI perspective and non-human point of view, with a sense of humor.

aboutface.io is run by Colin Owens. The founding question, first formalized in a 2009 MFA thesis on synesthesia and dynamic media, hasn't changed: how do we build systems that work with how people actually perceive, rather than how computers work?

The question has a second half now. When a machine writes the code, drafts the copy and files the report, the problem stops being perception and becomes trust: how do you know the thing in front of you is true? Social Trust Anchors asks that about information on the open internet. The orchestration work asks it about machine output inside a company, where being wrong is expensive and someone has to sign their name to it. Same question, different scale.

The answer that keeps working is unglamorous. A claim is worth nothing unless you said in advance what would prove it false and where you would look. Applied to a research framework it is epistemology. Applied to eight agents working in parallel on a client's production pipeline it is the difference between a system you can defend and one that quietly ships the wrong thing.

So the outputs look unrelated and are not. Research into information trust. Products that ship. Orchestration architecture for a global software company. Writing that argues a position. Twenty years of practice, still in progress.

For design work, the CV, and consulting, see colin-owens.com.

Get in touch: colin@aboutface.io