🌿  home  contact  support

projects (rss)

My work is focused on exploring synergies between organic and artificial thought. This usually means putting together proof-of-concept tools based on off-the-shelf ML models or, more recently, tiny experimental investigations into aligning AI to human intentions. I often refer to this family of projects as thoughtware.

  • nested state clouds: extracting knowledge graphs from pretrained transformers.
  • deontic dipoles: inoculating language models with arbitrary stances.
  • velma: value extraction via language model abduction.
  • conceptarium bibliography: a pipeline for relating ideas to experiences.
  • cybersalience: a synergy of artificial and organic attention.
  • oneironomicon: a conversational sandbox for training user-centered assistants.
  • decontextualizer: a pipeline for making highlighted text stand-alone.
  • lexiscore: a nutritional label for food for thought.
  • ideoscope: an instrument for quantifying, understanding, and optimizing your thinking.
  • conceptarium: a fluid medium for storing, relating, and surfacing thoughts.
  • dual: amplifying knowledge work through user-defined assistants.
  • autocards: accelerating learning through machine-generated flashcards.
  • knowledge probes: promoting critical thinking through prompt generation.
  • memory navigator: expanding propositional memory through text mining.
  • semantica: augmenting conceptual thinking with artificial primitives.

principles

I learned a lot about myself through those projects, and I now have a more refined vision of how thoughtware should be like. It should be open source and self-hosted, in order to ensure better goal alignment between users and tools. It should be modular and composable, in order to support a combinatorial explosion of workflows. It should amplify and extend existing human brilliance rather than dehumanize through atrophy. It should non-judgementally extend the range of thinkable thoughts, rather than limit minds to rationalist conceptions of truth and knowledge. It should be informed by specific strengths and weaknesses of the human psyche as understood in cognitive psychology. It should be defined by its underlying culture and embodied principles, rather than opaque creators.

themes

In the upcoming years, I plan on exploring the following questions:

My work on the thoughtware stack, if you will, is made possible by a handful of generous sponsors. If you resonate with it and want to support an experimental tooling ecosystem for thinking, please consider supporting me via the link below.