Why I ignore new LLMs for at least 2 weeks
With everything changing so relentlessly fast, a lot of us have been feeling overwhelmed. New announcements every day urging us to drop everything and start using some new and shiny thing, lest we fall behind.
My technique for dealing with this:
I ignore every new announcement for at least 2 weeks.
Because when something new emerges and suddenly gets a ton of attention, it can be for one of two reasons:
-
It's genuinely important and worth your attention.
-
It is overhyped and not as important as it seems.
Or it can be somewhere in between: valuable, but still overhyped.
If it's closer to #1, then in 2 weeks, you'll still be able to learn about it and use it then. You won't lose anything that matters.
If it's closer to #2, you saved yourself much distraction, anxiety, and cortisol. Putting you ahead of everyone who got caught up in the hype.
For example, when OpenAI once announced a new LLM, which I'll call SomeNewLLM (because its name does not matter). It was all over Hacker News on its day of release. The week after, when a colleague mentioned it, my response was:
"What's SomeNewLLM?"
I had no idea, because I make a point of ignoring hype. It is only when hype dies down that we clearly see whether there is actual value.
The other benefit of waiting:
By the time you do pay attention to it, enough people have run it through its paces that you can find good articles and content explain how to get the most out of it.
Again, this new LLM is a good example. A few weeks later, folks had created some quality content about its pros and cons for different kinds of tasks. To the point your favorite chat AI can already give good guidance about when to use it, and when not to.
Before that point, all you can get are low-information "hot takes", which are more entertainment than education.
That's the job of influencers, who rely on stimulating your emotions in order to monetize your attention. I hope you realize that's not always in your best interest.
And then there's the fact that by the time the hype dies down, all the early bugs have probably been fixed.
(In the extremely-rare case it's important AND urgent, you'll hear about it from people around you. You will know.)
This isn't just about LLMs, of course. It applies to pretty much all breaking news about technology, and how we work with it as professionals.
P.S. If anything, 2 weeks might be 2 soon. There may be an argument for waiting months as a general rule, with exceptions on a case by case basis. In the long run, that strategy might have an even better net gain.
P.P.S. If you want to learn about AI coding principles and practices without wasting your life on shiny hype, go here:
https://boostedcode.com