I’m old enough to remember Usenet, email distribution lists, and the early days of web forums. The value of those systems was in the years of conversational context lying in wait. Traversing the deep, wide knowledge took time. Distilling it took attention and effort.
Then came the search engine wars. Altavista was good for some things. Yahoo! for others. Google showed up with pretty much the same results, lies about intent to monetize, and a lot of hype. Now servers reject my emails because they don’t say “@gmail.”
Google shortened time spent with the content, flattened and trimmed context to twenty pages, max. I spent less time searching for content, which felt good. I spent less time consuming context, which was less exhausting.
It came at a price. Expectations got higher and higher. Search algorithms changed to anticipate wants. Content was contextualized for me. Sometimes I liked that.
Skip over a few things, and we get to AI augmented web browsing. Search results begin with an AI summary. Browser sidebars and URL bars prompt LLMs with a couple of keystrokes.
From what I hear and see from users, most of the prompts generated by and for these assistants are the wrong way around. They ask for summaries. LLMs get tasked with churning out shortened versions of content written at fourth-grade reading levels and riddled with hallucinations their designers have no interest in fixing.
I use them the opposite way. I am constantly asking Gemini and ChatGPT to give me more. More context. More sources. More everything.
Act like a researcher at a policy think tank that leans moderate for american politics. Based on the text below, taken from [url], provide historical policy context for the salient points. Cite specific legislation and/or established court precedent when possible. Write at a graduate school level. Do not assume the reader understands jargon.
The above prompt has been hugely helpful not just in understanding my moderate left friends, but also for expanding topics. The response to this prompt is followed up by several more questions. Please explain X legislation in light of Y. Provide more historical context for A, B, and C court cases. What do you mean by [prominent political movements I should probably already know about]?
When I feel mildly informed, I return to Google, looking for more official or trusted documentation. I frequently end up at Wikipedia, Amazon, or the library, looking for long form texts.
A long the way I assume that the AI summary was factually incorrect, or socially imprecise, but that I do not know how.
After a few hours, I learn a lot. I learn more than I did a couple of years ago before AI became available. And I like that.