Google Sucks at Marketing

Shaan Puri demonstrates how AI tools are becoming proactive assistants that can handle significant portions of knowledge work, from communication to research to team management. He shows that the shift isn't just about asking AI to do tasks—it's about AI automatically figuring out what needs to be done based on context.

Key Tools and Use Cases:

• NotebookLM (Google):

  • Google has powerful tools but "sucks at marketing them"—you need to "poke around" to find them
  • Can transform any podcast into a slide deck presentation automatically
  • Example: Dropped in a YouTube link to an hour-and-a-half AI podcast, asked it to "turn this into a slide deck"
  • Created an entire presentation called "Summoning Ghosts: The Decade of AI Agents" with quotes, graphics, and structure
  • Value: Get the gist of long-form content without watching/listening to the full thing
  • Not perfect—some charts look "kinda crazy" and might have things that are "a little bit off or a little bit dumb"
  • But legible and useful enough to decide if you want to dive deeper

• WhisperFlow:

  • "I barely type anymore"—works across any app (Slack, Facebook, Instagram, email, iMessage)
  • Just hit "start flow" and speak naturally
  • Learns your personal style: removes periods if you think they "look too formal," separates lines, adds bullet points
  • Understands commands like "scratch that" to remove mistakes
  • Key insight: "Typing is a pretty slow thing to do, it actually requires a lot of energy and you don't really realize it until you're able to just stop doing it"

• ChatGPT for Communication:

  • Uses it to manage workplace conflicts productively
  • Example: "Sometimes if I get in an argument with one of my employees...what they did angers me but I need to be productive about it"
  • Uploads the email and asks: "My instinct is to write this angry thing but obviously I know that's not right, can you help me write this email in a way that gets me what I want"
  • "It's already doing a lot of the talking anyway"

• Nebula (by Furqan):

  • Early prototype that connects to Slack, Gmail, Google Calendar
  • Proactive, not reactive: "You don't need to ask me what to do, I should be able to figure out what to do based on what's going on"
  • Automatically creates prep docs for upcoming meetings
  • Reads code changes on GitHub and summarizes them without being asked
  • "If something changed, I'm gonna tell you about it"
  • Best for: Teams of 10-30+ people with daily standups, code pushes, multiple communication channels
  • Not as useful for small 3-person teams where everyone already talks directly

The Bigger Shift:

• From Prompted to Proactive:

  • AI is moving from "you have to tell it what to do" to "it figures out what needs doing"
  • "Every day I'll just proactively search for stuff and write stuff to you about the things you care about"
  • Set up workflows like "keep checking my dashboards and giving me summaries and tell me when anything pops off"

• The Challenge:

  • "Staying on top of all this stuff is really hard"
  • Tools are evolving rapidly and many powerful ones are under-marketed or in early stages