By playing, you agree to uListen's Terms and Privacy Policy and YouTube's Terms of Service.
Summaries, chapters, transcripts, and FAQs on this page are made by uListen, not YouTube.
CHAPTERS
- 0:00 – 2:35
The Promise and Letdown of the Second Brain
Steph Smith frames the long-held idea of a second brain and asks whether it has delivered. She notes that computers have been directive tools, while consumer AI can now handle plain-language prompts and unstructured data, and she introduces the Mem founders.
- •People often spend more time organizing a second brain than using it
- •Computers to date have required discrete instruction from the user
- •Consumer AI can now process simple language prompts and unstructured data
- •Guests are Mem founders Kevin Moody and Dennis Xu
- 2:35 – 4:49
The Hidden Cost of Searching for Information
Steph Smith cites a statistic that knowledge workers spend 2.5 hours a day searching for information. Kevin Moody thinks that is an understatement and describes the hidden time sink of redoing work that already exists. Dennis Xu says search is only the beginning of the value and asks about the mundane tasks that follow it.
- •Knowledge workers spend about 2.5 hours a day (roughly 30%) searching, per a 2016 stat
- •Kevin Moody says the figure is likely still relevant or higher because unstructured content is easier to produce and share
- •Teams often reinvent the wheel without knowing the work already exists
- •Dennis Xu says the workflow of a knowledge worker will change over the next few years
- 4:49 – 8:47
Why Technology Hasn't Delivered Seamless Collaboration
Steph Smith traces knowledge technology from writing to the internet and asks why collaboration is still inefficient. Dennis Xu argues past tools were broadcasting devices that lacked hyper-personalized consumption. Kevin Moody explains how organizing information burdens the creator and rarely helps others.
- •Past tools mainly broadcast information rather than helping people digest it
- •The same search by different people should yield different results
- •Information's most valuable use often comes weeks, months, or years later
- •Traditional knowledge bases impose the creator's structure, which makes little sense to others
- •Kevin Moody sees an opportunity for information not to need organizing, with AI matching it to context
- 8:47 – 12:45
Eliminating the Folder with LLMs
Steph Smith asks what new approaches LLMs open up. Dennis Xu explains that the folder is a skeuomorphic holdover from filing cabinets and that LLMs add the missing reasoning. Kevin Moody argues information should be organized into an arbitrary number of retrieval contexts, retrievable through natural language.
- •The folder mimics the physical filing cabinet from early computing
- •Computers lacked contextual understanding of human language and could not reason
- •LLMs demonstrate reasoning that can be paired with computers' vast storage
- •Kevin Moody contrasts automatically generated folders with organizing for arbitrary ways of recalling information
- •Users can ask natural language questions with no categorization step
- 12:45 – 15:03
Proactive AI Assistants Beyond Clippy
Steph Smith raises proactivity, where AI surfaces useful information unprompted. Dennis Xu jokes that Clippy was ahead of its time and says proactivity only works if the system deeply understands the person. He describes flipping the model so the machine acts and then confirms with the user.
- •Proactivity could surface work others in the organization already did
- •Proactiveness must avoid being annoying through deep understanding of a person
- •Interactions with computers are still limited by explicit instructions, even in English
- •The machine could act first and confirm afterward
- 15:03 – 18:34
Cost, Monetization, and Value Beyond Retrieval
Steph Smith asks about the cost of running such a system and how it can be monetized. Kevin Moody describes an individual free and premium plan and says the cost of compute is falling by an order of magnitude every several months. Dennis Xu argues the value goes beyond retrieval and could be compared to hiring help rather than to a software subscription.
- •Mem aims to democratize access to a personal AI
- •Individual plan with free and paid premium tiers
- •Compute costs are coming down by an order of magnitude every several months
- •Traditional note apps' value ended at search, while AI can help think through problems
- •Pricing can be compared to hiring an assistant rather than a typical subscription
- 18:34 – 21:03
Personalization and Adaptive AI
Steph Smith shares her wish for an AI that calls her out on her bullshit and holds her to her goals. Kevin Moody jokes that Mem may add that as a setting. Dennis Xu says true personalization means the AI adapts and learns who you are, not merely searching your data, which Mem calls adaptive AI.
- •Steph Smith wants an AI that holds her accountable to her stated goals
- •Kevin Moody says a 'call me out' option might become a knob
- •Access to personal data alone doesn't feel personalized
- •Adaptive AI learns about the user and applies it in future interactions like a human would
