Instinct is the AI assistant every VC is chasing. Its founder, @noahrshinn, co-wrote Reflexion,…
Summary
Noah Shinn of Centra argues that organizational memory functions as a company brain to enable collective intelligence, moving beyond simple searchable databases to create a unified entity.
Memory is more than search.
Shinn distinguishes organizational memory from a glorified search box, describing it as a model that provides specific perspectives rather than just retrieving data.
Memory is much closer to a model
Organizations provide clearer utility functions.
Building memory for teams is more feasible than for individuals because companies have defined motivations, such as maximizing profit and selling products.
companies and teams you can define why they are doing what they are doing
AI architecture requires continuous evolution.
Shinn claims future systems must move beyond RAG or graph-based approaches toward models that constantly change and evolve through learning.
Memory collapses KPI analysis time.
In a 2,500-employee organization, Centra reduces the time required to identify the cause of KPI changes from one week to ten minutes.
Quantifying AI value remains difficult.
Enterprise pilots often fail because token usage is hard to estimate compared to fixed human salaries, making it difficult to prove profit impact.
I don't know how many tokens they are going to use
Read the original ↗Save to your Sponge
Summarized with Sponge. Quotes are the author’s own words.