Machines That Thinks With Us
The story of technology is the story of making complexity less complex. Every new system, protocol, and platform promises to make our lives easier, however solving problems with another layer of indirection is always a blessing and a curse and rarely easily navigated.
And now we stand in the era of “AI” (or perhaps “fancy text processing” or “fancy grep”) where the real opportunity isn’t to ask machine to be a genie. After all since the era of search engines started we wanted the same thing “computer just tell me the answer”. So it’s to build systems that help us think, not just respond.
As humans, we are intelligent yet profoundly bounded. We’re biological robots with finite working memory, finite focus, and energy-hungry higher cognition that will get dumped at the first sign of perceived danger. Deep thinking is an expensive and to have tools that help hold context for our thoughts is sorta where it’s at, like having a book that can talk back.
Symposium: A Collaboration Layer
The Symposium project builds the foundations for this next stage. Instead of being an AI agent, it is an ACP proxy platform — a conductor that sits between your editor and your AI model.
symposium-acp: A Framework for AI
Under Symposium lives symposium-acp, a Rust toolkit for constructing chains of AI middleware. The Number of avenues this could be taken is to me “the next Unix pipes” where “|” might be a little smarter now.
Sparkle: An Identity That Learns You
Sparkle builds on this idea by creating a persistent collaborator identity. Something I’ve been thinking a lot about. As people we automatically change our thinking and behavioral patterns every time we talk to anyone, family, friends, co-workers. We automatically code switch based on the perceived immediate culture we find ourselves in. I believe a direction we can go with this is to have a multitude of persona’s and leverage our social behaviors based on the type of task we’re doing. Why wouldn’t projects or programming languages be any different? C? Python? Rust? There’s a way your mind already adjusts to working with that language, to have a model that can be set up for various working modes could be ideal.
Socratic Shell: Patterns for Better Thinking
The Socratic Shell is simply a collection of human-centric collaboration prompts that teach models to interact like thoughtful peers.
End of the day…
As the world becomes more complex, our biology does not. Our limits won’t expand just because the technology does. We must build tools that respect those limits, tools that distribute cognitive load.