Ask not what AI can do for Pullquote…

ChatGPT4o’s response to the prompt: “Create a pen and ink drawing that shows a giant wall of library card drawers with a tall rolling ladder in front”

Pressflex’s Pullquote extension for the Chrome browser, which makes it simple to store, tag, deep-link, and share quotes and facts, could help people tap into AI’s genius for processing ideas.

Background: Writing recently about AI’s osmosis with human thought and knowledge revived my interest in the idea that, in parallel to the massive “internal” work of upgrading AI’s algos, datasets and hardware, AI will also need lots of “external” help interacting with day-to-day subjective human needs, whether individually or collectively. In other words, as VCs and behemoths like Open-AI and Nvidia invest in strengthening AI’s brain-muscles-skeleton-digestion, AI also needs skin that’s both responsive to human touch AND invites human touch. Pullquote might help with this task.

Pullquote: Pullquote, as a day-to-day catalog of noteworthy snippets of an individual’s knowledge, facts, and interests, might serve as a human-centric skin for AI. Pullquote seems to be particularly well-suited to serve as an interface for AI: it’s lite — sitting in the Chrome browser and operated with a couple of clicks — and it adds metadata to quotes — users summarize and tag ideas/facts — and it creates a unified storage space for an individual’s interests. (The Pressflex team created Pullquote in 2011 to empower people to deep-link to a specific quote buried in the long text of a URL. We gradually added to this original function features for classifying, storing, and sharing quotes. In 2019, Google added “scroll-to-text” to Chrome, which covers Pullquote’s original use case. One limitation: Pullquote is desktop-only, because Chrome extensions don’t operate on mobile.) 

Talking with ChatGPT: To play around with how AI might enhance Pullquote and vice versa, I conversed with ChatGPT about Pullquote, then asked GPT to analyze some of my own Pullquote collections. Here’s AI’s take on my #Trump pullquotes, and ChatGPT’s take on my #carbs feed, and AI’s analysis of the feed of my recent Pullquotes

AI-suggestions for AI-in-PQ: No earth-shattering insights yet, but I did get a good list of possible AI-angles for Pullquote, some simple, some more complex. (Asking AI specifically for advice on driving network effects in Pullquote provided no extra insights.) Here’s the highlights I noted or extrapolated from GPT’s suggestions:

  1.  AI suggests tags when I grab a quote
  2. AI suggests people/tags to follow
  3. AI identifies a surprise connection of the day that sits at the intersection of ideas that interest you
  4. AI builds a newsletter based on tags you’ve used
  5. AI charts trends over time in your tagging
  6. AI suggests a summary for each quote you grab (currently our headline) 
  7. AI suggests a counterpoint for any quote you grab
  8. AI summarizes quotes you’ve given a specific tag to (our headline function)
  9. AI, like a tutor, suggests context for quotes or links to related information
  10. For each new quote, AI links to related quotes… or links to an essay integrating all related quotes for a given tag
  11. AI reveals links between similar quotes in my catalog… or other people’s?
  12. AI drafts an article (or states a thesis) based on quotes with a specific tag

Steps I took:

  1. To prime the task, I asked AI to summarize
  2. I asked GPT to suggest ways to integrate AI into PQ. 
  3. I asked AI to summarize my quotes at
  4. Since that summary was uninteresting, I asked AI to identify what areas were missing from my quotes. 
  5. I then asked AI to find a quote about one of the missing areas it has identified (Trump’s foreign policy) … it searched five sites and came up with something. 
  6. I asked AI to illustrate the ideas in the Trump quotes. (It refused!) 
  7. I asked AI for suggestions about using AI to drive network effects for PQ. 


  • I guess the most startling thing I realized is that ChatGPT can interact in multiple ways, right now, with Pullquote without any work on our part. 
  • The “what’s missing from these quotes” function was interesting. (In fact this is my favorite use of AI when I’m writing an essay — “what am I missing?” or “what are other examples of this phenomenon I’m describing?”
  • Newsletter engines like Substack have gained ascendence since we created Pullquote… good trend for PQ to ride?  

Network effects: I’d note that only one of the ideas above — AI recommending other Pullquoters — creates even weak network effects, exciting a user to beg friends to participate in Pullquote so the original user benefits. (Facebook and eBay have strong network effects; adding more users creates more opportunities for current users to communicate or trade.)  Network effects are crucial to growing web tools without spending money on marketing. Network effects are also important if I’m right that AI’s growth will be strongest when it facilitates socializing or other p2p communication/collaboration, just as social media fueled the Internet’s explosive growth. I think the biggest winner of the “skin for AI” game will be an AI service that enhances p2p collaboration, with AI as the medium for socialized creation or processing of information. 

Ideas to explore: 

  • PQ for work teams.