Tools for Thought as AI Copilots
Jeff Tang, CEO of Athens Research, on Pinecone and the AI stack
The real product in tools for thought is better thinking, not better storage. Athens was built around helping a user see relationships across messy ideas, people, and projects, so the note is only raw material for a higher level job, which is pattern finding and problem solving. That is also why Tang saw AI and vector search as potentially leapfrogging manual note workflows, because they can summarize, connect, and retrieve meaning from unstructured information faster than a person tagging and organizing everything by hand.
-
Athens used a graph model to surface bidirectional links, properties, and relationships, especially for generalist work that spans product, design, engineering, and go to market. The value was seeing how pieces connect, not capturing one more page of notes.
-
The same idea shows up outside note taking. Runway argues spreadsheets and finance tools matter when they help a team offload complicated mental models and understand the business together. In both cases, the winning product is the one people reach for to think through decisions.
-
This is why AI changes the category. Once language models can summarize text, find common threads, and pull useful context from a large corpus, the interface can shift from manual writing and filing toward an always available reasoning layer. That is the broader pattern behind AI assistants and workflow products like Zapier as well.
Going forward, tools for thought are likely to converge with AI copilots that work quietly inside everyday workflows. The category will be won less by who offers another editor, and more by who best turns scattered company context into live assistance that helps people understand, decide, and act in the moment.