Geoffrey Litt's Malleable Software Vision

Table of content

Geoffrey Litt is a design engineer and researcher who builds tools that let ordinary people customize their software. He did his PhD at MIT in human-computer interaction, advised by Daniel Jackson, with a thesis on building personal software with reactive databases. Before academia, he worked at Panorama Education building software for teachers.

Litt researches malleable software — computing environments where users adapt tools to their needs without engineering teams. He works at Notion after years at Ink & Switch, the independent research lab exploring the future of computing. See Malleable Software for the concept and its implications.

The Core Idea

From the Malleable Software essay (June 2025), co-authored with Josh Horowitz, Peter van Hardenberg, and Todd Matthews:

“The original promise of personal computing was a new kind of clay — a malleable material that users could reshape at will. Instead, we got appliances: built far away, sealed, unchangeable.”

The manifesto argues:

The “application” model is fundamentally flawed
AI coding helps but isn’t sufficient alone
Users deserve the ability to customize their tools

AI-Generated Tools

Litt’s December 2024 blog post describes an unusual AI use case:

“What’s unusual here is: the AI didn’t write a single line of my code. Instead, I used AI to build a custom debugger UI … which made it more fun for me to do the coding myself.”

While working on a Prolog interpreter, Litt hit tricky bugs in unification. Instead of having AI fix the bugs, he had it build a visualization tool to help him understand what was happening. The debugging became enjoyable rather than frustrating.

Background

Period	Work
2019-2023	PhD at MIT CSAIL — Programming interfaces
2020-2025	Ink & Switch — Research on malleable software
2025-present	Notion — Design engineer

His PhD focused on end-user programming: how can ordinary people modify their software without learning traditional programming? His thesis, “Building Personal Software with Reactive Databases,” showed how spreadsheet-like techniques can let users customize web apps without writing code.

The Home Kitchen Metaphor

From the Dialectic podcast (June 2025):

“When I say malleable software, I do not mean only disposable software. The main thing I think about … is actually much closer to designing my interior space in my house. When I come home I don’t want everything to be rearranged. I want it to be the way it was. And if I want to move the furniture or put things on the wall, I want to have the right to do that.”

Malleable ≠ disposable. It’s about crafting an environment over time.

Ink & Switch Projects

The lab has produced influential prototypes. Litt worked on several of these directly:

Potluck — Gradually enrich text documents into interactive personal software
Embark — Dynamic documents for travel planning
Wildcard — Customize web apps using a spreadsheet view
Automerge — CRDT library for collaborative editing

The lab also pioneered local-first software, a set of principles for software that keeps data on your device while still enabling collaboration. Litt’s work connects these technical foundations to the user experience question: how do we actually give people control?

At Notion

Notion CEO Ivan Zhao has described the app as LEGO bricks — building blocks users combine freely. Litt’s research aligns with this vision. His role as design engineer bridges research insights with product reality.

AI’s Role (And Limits)

From the Malleable Software essay:

AI coding holds potential but isn’t sufficient alone. Problems:

AI-generated code still requires understanding to maintain
Disposable one-off tools don’t compound into personalized environments
The infrastructure for persistent customization doesn’t exist yet

The path forward requires both AI capabilities and new kinds of software architecture.

Litt writes about his own AI coding practice in “Code like a surgeon.” He uses AI for secondary tasks (writing documentation, spiking out changes, fixing clear bugs) but keeps primary design work hands-on. The goal: spend 100% of time on what matters.

Imbue Podcast (November 2025)

In Malleable Software and Human Agency, Litt discussed:

Technical, economic, and infrastructural barriers
Inventing new UI components for the AI age
Improving agent-human collaboration
How AI affects the creative process

Core question: how can everyday people shape their software like clay so humans have more power and agency?

Key Insights

Principle	Implication
Software should be clay	Users reshape tools to fit their lives
AI helps but isn’t enough	Infrastructure changes needed
Chef knife over avocado slicer	General tools beat specialized gadgets
Environment, not disposable	Customizations should persist and compound

Links

Next: Armin Ronacher’s Agentic Coding Practice