It takes a while to figure out technology

20.11, Friday 24 Nov 2023

The first web pages and the first web browser, WorldWideWeb, were published December 1990. Mosaic launched in 1993 and became the first commercial web browser in 1994.

I made my first e-commerce purchase in 1997 maybe 98. It was a crazy heavy resin gargoyle, mentioned here, and the way I bought it was I browsed the website and then sent an email saying what I wanted to buy and giving my credit card number.

7 years in and it wasn’t obvious yet that you could type your credit card number in an input field.

Amazon filed their 1-Click patent in September 1997. It was granted in 1999. Here it is: Method and system for placing a purchase order via a communications network

The idea is that you have previously put your credit card number in an input field. And then using a cookie, the website remembers who you are. Then you can hit Buy Now without having to re-enter your details.

Apple paid $1m to license 1-Click in 2000 (Wikipedia).

Look, it was kinda obvious then, the people who sold me my gargoyle aside. The patent being granted was a little controversial. But as an idea it was not obvious obvious.

Even being conservative about the timespan, from 1994 when the web became the most popular service on the internet, to 1997 when the patent was granted, that’s three years and still the idea of “putting your credit card number in a box and the server remembers it” was novel enough to allow for a patent.

I was still buying software in a box off a shelf into the 2000s. Software was still a business with inventory; it was measured in terms of stock, not in terms of customer acquisition cost and retention. How long did it take for the web to replace boxed software with SaaS? 15 years? And we’re still figuring out the best ways to make a pricing page.

All I mean is that it takes a while to figure things out.

With the web, all the pieces were there from the early 90s.

We didn’t get till 1999. That’s when the idea of UGC - “user-generated content” - started going mainstream. Blogs themselves didn’t go mainstream till, what, 2001? 2004?

OpenAI released GPT-3 in June 2020. That was good enough for chat. The interface wasn’t cracked until November 2022 when OpenAI released ChatGPT. 2 years!

The technique behind chat agents is called Retrieval-Augmented Generation, RAG. It was invented in May 2020 (arXiv). It’s a fundamental building block, dead simple: you concatenate the prompt with a relevant document retrieved from a database using vector search (which is surprisingly good). But it wasn’t well known until mid 2023.

Inventing takes time!

I keep coming back to this tweet from Nat Friedman, ex CEO of GitHub and now deep into AI.

The multiple cantilevered AI overhangs:

  1. Compute overhang. We have much more compute than we are using. Scale can go much further.

  2. Idea overhang. There are many obvious research ideas and combinations of ideas that haven’t been tried in earnest yet.

  3. Capability overhang. Even if we stopped all research now, it would take ten years to digest the new capabilities into products that everyone uses.

And you know what, that tracks for me.

So I don’t feel I’m ever in a hurry with new technology. I’m not saying don’t do the work. Do the work like crazy.

Because we are imagination bottlenecked.

Share techniques and ideas widely.

Demo freely.

Get the obvious ideas out of the way and together we’ll come up with the good ones.

This, by the way, is why London is such a great scene right now.

More posts tagged:
Follow-up posts:

If you enjoyed this post, please consider sharing it by email or on social media. Here’s the link. Thanks, —Matt.