Our manifesto

Civilization advances through infrastructure that rewards truth-seeking and enables collective exchange of ideas. We’re concerned by current AI writing tools, which have quickly become everyday intellectual infrastructure for many humans, but overwhelmingly replace human experiences and care, hurting human agency.

In this project, we’re building an AI partner – which we term Effortful – that supports human thought through (1) novel knowledge representation interfaces and (2) connecting users’ private writing spaces in controlled, productive ways.

We claim that there are two gaps in existing tools that Effortful addresses:

(1) Writing is truth-seeking’s best friend–we believe that to clearly think, one must write. We envision technology as a great potential partner in this process, but current model interfaces are cumbersome and sycophantic: writers waste energy manicuring context for LLMs, and chatbots default to pleasing rather than challenging. Instead of supporting intellectual rigor, we’ve found they offer more help than is epistemically healthy.

Effortful helps people think through writing by developing an epistemically healthy, visual-centric interface. As one writer we interviewed said: “Writing and reading is nonlinear… you’re in the world of ideas and there’s nothing linear about an idea. I’d love any tools that break that linearity and help you look holistically.” Our core interest is in designing bi-directional text-graph interaction, transforming linear writing into dependency graphs mapping key lines of argument, evidence, and ideas. Adding text creates new nodes in real time, while manipulating nodes – for instance, adding a premise or exploring different conclusions – auto-generates new text that extends a user’s reasoning along multiple paths. Unlike conversational LM interfaces, which encourage copy-paste behaviors and fragmented thinking, Effortful’s seamless interaction lets users explore different abstraction layers while remaining rooted in their original ideas/thinking.

(2) Truth-seeking thrives through clusters; take the Enlightenment-era Royal Society of London or present-day founder networks in Silicon Valley. We think that this paradigm of collective, human-grounded truth-seeking will and must continue even in the new rapid age of AI. However, existing interfaces are generally isolating, single-user text boxes.

With Effortful, we want to explore cross-pollinating knowledge networks that connect users’ private writing spaces in controlled, productive ways. Imagine if Robert Boyle, Christopher Wren, and other natural philosophers had easy shared access to each other’s research notebooks. Through smart nudges and group brainstorming sessions, Effortful’s AI will identify conceptual overlaps between users’ knowledge graphs and enable sharing of relevant ideas, bringing this into reality.

In its first iteration, Effortful will be an open-sourced web platform that is both a user-facing writing environment and a continuous research testbed. From its usage, we hope to rigorously investigate collaborative thinking in the age of AI.

We plan to measure project success computationally—tracking user network interconnectedness and document embedding drift as indicators of creative growth—and qualitatively through interviews and surveys with pilot users, building upon our initial grant research.

We’re really excited about where we’ll take this, and we’re optimistic about the potential for AI to support truth-seeking and human creativity. We’d love for you to follow along in our journey!