Definition
Long-Term Conversational Memory
A technical capability that allows an AI agent to retain context across multiple sessions over days, weeks, or months - referencing prior conversations, tracking commitments made, and building a longitudinal understanding of the user. Distinct from in-session context (which resets after a conversation ends), long-term memory enables genuinely cumulative relationships. In accountability products like Commitify, this capability is the core retention mechanism: the value increases with each call because the AI remembers what you said last time and can follow up on specific commitments.
From the episode “AI Friends That Actually Call You (And They Remember Everything)”Hear it on the podcast1 episode
Learn the language of AI-native founders
Every week: the new terms, tools, and workflows founders are using, straight from the podcast.
Related terms in AI Fundamentals
LensEncoded expertise — the structured prism of how an individual thinks, weighs evidence, and judges. Distinct from the artifact it produces.LLM Benchmark ConvergenceThe narrowing of performance gaps between frontier models (GPT, Claude, Gemini) as all approach ceiling benchmarks on standard evaluations. Users switch models for marginal improvements, then switch back. Actual loyalty is driven by workflow integration, API familiarity, and developer preference rather than raw capability differences. The meaningful quality signal that persists through convergence: developers voluntarily paying 4–20x more for Anthropic than for cheaper Gemini-equivalent performance — a revealed preference that tells you more than benchmarks do.Local AIAI processing that occurs on a user's device rather than in the cloud.Local modelA model running on your own hardware or private infrastructure, so the data never leaves the building.Loss functionThe formula that scores how wrong a model's predictions are. Training is the process of driving that score down.Machine learning (ML)The branch of AI where computers learn patterns from data and improve at a task without being explicitly programmed with rules.Markov chainA model of a sequence of events where the odds of what happens next depend only on the current state, not on how you got there.ML modelA custom machine learning model trained on your own data, as distinct from a general-purpose LLM. More expensive and data-hungry to build, but harder to replicate because the training data is proprietary.