Areas of exploration
X-Thought Labs explores new approaches to LLM systems that can reason, adapt, preserve continuity and operate effectively across extended periods of time.
Exploring how LLM systems can preserve useful knowledge over time while remaining capable of revising earlier conclusions.
Our current work is a memory prototype that records where knowledge came from and when it was known, corrects without rewriting history, keeps obligations, and does not let a model turn text into a decision. See progress
Exploring how LLM systems can remain coherent and effective across extended periods, changing contexts and continuing tasks.
Studying how LLM systems can adjust the depth, structure and computational effort of reasoning to the demands of a situation.
Investigating whether LLM systems can become better at recognizing uncertainty, weaknesses and recurring patterns in their own reasoning.
Studying how LLM systems can balance reasoning quality, computational effort and the value of additional analysis.
