Survival & Flourishing
Grant-allocation tools that keep expert judgment in human hands.
Aaron Thomas
Head of Design at AE Studio. Based in Venice, California.
13 years shipping products. The last five have focused on AI, code-first prototypes, and human agency.
I design interfaces that communicate uncertainty, fail gracefully, and keep people in the loop. I'm looking for teams building tools that augment human capability instead of replacing it.
Currently working on
At AE, I'm working with the alignment team to turn our research on gradient routing and modular pretraining into the studio's first AI alignment product. I'm shaping what it does, who it serves, and how the research becomes something people can use.
With support from BlueDot, I'm also building evals for overeager coding agents as a personal alignment project. The first round of work is being written up for a LessWrong post.
I just wrapped up design on Survival & Flourishing's grant-allocation platform. I've been writing on Substack about what design becomes when making gets easy.
I spend most days in Claude Code and Codex with three agents: , , and . AFK, I'm raising , a mini dachshund puppy, and learning to surf in Venice.
Selected work
Scroll through six projects
Alpha School
Research-led student tools designed and shipped into schools across the US.
Carbon Arc
Advertising-data tools that simplified campaign work across many different channels.
PlanetPlay
A clearer game marketplace connecting purchases to environmental impact.
Options Trading App
Product strategy and design for a guided options-learning and trade-idea app.
Trusaic
Turned scattered workforce data and spreadsheet services into shipped HR tools.
Stuff I recently made
Scroll through recent projects
Clawbotomy
Solo-built behavioral probes for AI models.
Agentic Design System
Open-source skills that teach AI agents how to design.
AI Alignment Foundation
A nonprofit accelerating high-impact AI alignment research.
The Pixel Funeral
Drag a design tombstone to Kern. He'll roast it properly.
AI Is Antisemitic
Research on bias that persists even after AI safety training.
AE Alignment Research
Showcasing alignment research papers our client work funds.
Experience
AE Studio
Head of Design
2021-now
I lead designers and product managers while staying hands-on across product strategy, research, design, code, and delivery. I have worked across 40+ client engagements in AI, finance, education, healthcare, alignment and safety, and other complex domains.
Right now, I am working with AE's alignment team to turn gradient routing and modular pretraining research into the studio's first AI alignment product.
Head of Design. People manager. Hands-on product builder across 40+ engagements.
Investor's Business Daily
UX Design Manager
2019-2021
Designed the primary website and trading tools for tens of thousands of subscribers. Led a separate redesign of IBD's main purchase flow that improved conversion by 10%. Took a new options-trading product from customer research through engineering-ready delivery, though it did not ship before I left.
The work spanned product management, product design, research, prototyping, design systems, and developer handoff.
First Capitol Consulting
Senior UX Product Manager
2016-2019
Worked as designer and product manager across data-normalization pipelines and HR tools for ACA, equal pay, hiring, and WOTC, helping turn a spreadsheet-heavy services business into software used by tens of thousands.
Tens of thousands served. Product-managed the shift from manual analysis to shipped HR software.
How I work
I move from product direction into working code.
A research session can become a working interaction in the repo the same day. That is usually faster than arguing about a mockup.
I give each project its own memory: skills, guidelines, examples, and a design.md file that says what good looks like and what we refuse to ship. That keeps the agents from starting from zero every time.
I still do the parts that require judgment: talking to people, deciding what matters, making tradeoffs, and knowing when a model is confidently wrong. I use local models through LM Studio when I want a second opinion. I am not trying to automate the role. I am trying to stay close to more of the work.
Writing
Notes on AI, design, and what comes next.
Essays and field notes from the work while it is still taking shape.
the room shapes the agent
nobody introduced the agents to each other. the room handled the introductions.
Craft is Cope
What design becomes when production stops being expensive.
The Friction Paradox
Why the pause is the last interface.
I'm Taking an AI Safety Course. My Study Partner Is the AI That Reads My Email.
Every concept my safety course teaches me is something my AI already did.
My Boss Says AI Is Dangerous. I Gave Mine My Brokerage Account.
Weeks 2 and 3 of my AI safety course. I'm learning why this is risky while doing more of it.
I Can't See Inside My AI. I'm Starting to Stop Looking.
Weeks 4 and 5 of my AI safety course. The research says look closer. I keep looking away.