[Maro]

I don't think it's a good idea to kepp up to date at a daily/weekly cadence, unless you somehow directly get paid for it. It's like checking stocks daily, it doesn't lead to good investment decisions.

It's better to do it more batchy, like once every 6-12 months or so.

[Luc]

How do you do that? Once you're out of the loop for half a year, it becomes harder to know what's important and what's not, I think.

[pilotneko]

Every release is novel. Once something has been around for a while and is still being referenced, you know it’s worth learning.

Waiting 3-6 months to take a deep dive is a good pattern to prevent investing your time in dead-end routes.

[SoftTalker]

Yes this is why I never buy the latest CPUs and try to never run the latest release of any software. Stay a (supported) release or two behind the bleeding edge, and you'll find stuff is more stable. Common bugs and other issues have been shaken out by the early adopters.

[swyx]

my conference is currently run on a 6 month batch https://www.youtube.com/@aidotengineer

and is curated by me/my team. hope that helps people keep up on the video/talk-length form factor (as in, instead of books, though we also have 2-3 hour workshops)

[Maro]

Some ideas:

1. Buy O'reilly (and other tech) books as they come out. This will have a lag, but essentially somebody did this research & summarization work, and wrote it up for you in chapters. Note that you don't have to read everything in a book. Also, $50 is a great investment if it saves you 10s of hours of time.

2. Talks on Youtube at conferences by industry leaders, like Yann LeCun, or maintainers of popular libraries, etc. Also, YT videos on the topic that are upvoted/linked.

3. If you're interested in hardcore research, look for review articles on arxiv.

4. Look at tutorials/examples in the documentation/repo of popular ML/AI libraries, like Pytorch.

5. Try to cover your blindspots. One way or another, you'll know how new AI is applied to SWE and related fields. But how is AI applied to perpendicular fields, like designing buildings, composing music, or balancing a budget? Trying to cover these areas will be tougher, because it will be more noisy, as most commenters will be non-experts compared to you. To get a feel for this, do something that feels unnatural, like watch TED talks that seem bullshity, read HBR articles intended for MBAs, and check out what Palantir is doing.

[ndjdjddjsjj]

It also becomes easier. If you missed the early 2000s SOAP hype for example you ... just saved a load of time. Maybe stepping back you can avoid langchain etc. and various other workaround tooling and see what wins.