It’s not what you know, it’s who you know - more than ever.
The themes in this article are explained in a video cut from Jonathan’s Rootcause Summer School session and conversations leading up to it.
When everyone knows everything but not everyone knows everyone, it's not what you know but who know that matters even more.
Pointing out the shortcomings and failures of LLMs does not seem to impact the way people use them, after all:
Source: Reuters Digital News Report 2026
This means there is a large contingent of people using LLMs to understand the world, knowing that the information may be inaccurate. It suggests that the known shortcomings of LLMs are not enough to stop people using them, because the perceived benefits of using LLMs outweigh the negatives. The idea of knowledge, and feeling like you know something is more seductive than figuring out the truth.
The number of places in the information supply chain where AI is being used is increasing. If you think about how AI may be being used for research: writing reports, drafting executive summaries, summarising reports for research it becomes clear that information is being repackaged with an AI lens over and over again. Every time information is repackaged by AI, it risks becoming a little more homogenised, slowly nudging us all into similar thinking patterns. Rootcause founder, Jonathan Tanner, has noticed arguments he thought he deduced with Claude about topics such as cognitive offloading and an AI stock market crash crop up on blog posts and articles.
Is this the intellectual equivalent of an em dash? If we’re all working with AI, we’ll start to see the patterns of thinking the AI feeds to us crop up in i conversations about the world.
Polished sounding arguments are becoming increasingly abundant, knowledge is cheap and accessible to everyone, packaged up to be persuasive in any which way the user pleases. When we are no longer able to prove effort and understanding through a written piece, trust - or at least the ability to decide who to trust - can decay and we may revert to depending on networks, entrenching many inequalities we already see today.
What can we learn from this?
- Consider how to monitor what LLMs say about you and keep an eye on how to try and influence that (GEO)
- Try to balance awareness of the flaws of LLMs with understanding of where they are useful
- Consider how you can reach people that matter to your work in ways which are not meaningfully mediated by AI
- As mentioned in ‘Winning the Fight for the Future of Information’, build alliances everywhere you can in order to ensure that you have ‘network influence’ beyond your own reach