Interestingly, stagnation started around 2014 (in the number of questions asked no longer rising,) and a visible decline started in 2020 [1]: two years before ChatGPT launched!
It’s an interesting question if the decline would have happened regardless of LLMs, just slower?
> The unhelpful feedback was a consistent push to dumb down the book (which I don't think is particularly complex but I do like to leave things for the reader to try) to appease a broader audience and to mellow out my personal voice
Interestingly, this was my exact experience when working with a publisher (Manning, in my case), and it was the main reason I decided to part ways when writing my book (The Software Engineer’s Guidebook). While I did appreciate publisher’s desire to please a broader crowd by pushing a style they thought would broaden the appeal: but doing so makes technical books less attractive, in my view. And even less motivation to write!
In my case, self publishing worked out well enough with ~40,000 copies sold in two years [1], proving the publisher’s feedback wrong, and that you don’t need to dumb down technical books, like this specific publisher would have preferred to do so.
Even if it wouldn’t have worked out: what’s the point writing a book where there’s little of the author (you!) left in it. Congrats to OP for deciding to stick to your gut and write the book you want to write!
If you think the issue of devs using fake identities is a problem limited to the Fortune 500: I talked with a 6-person startup who very nearly hired a person who could have been from anywhere else than they claimed, including North Korea. All they know is the candidate used an AI filter to make them look like a Polish person [1] - and this startup recorded when they caught this faker.
This is a full-remote startup and they have now added a mandatory in-person interview to their recruitment loop.
Amusingly, in their case, using local job boards did not help: they got candidates pretending to from Poland or Serbia, yet not speaking the language.
A little sad to see how each episode like this casts more doubt and uncertainty into full-remote interviewing.
I’ve been talking to close to a dozen current and former Amazon employees since the end of 2021, and reached out to people publicly sharing attrition observations. Amazon attrition is really bad, even in AWS (Amazon employees all agree that AWS is a much better place to work in tech than Amazon retail).
At the same time, Amazon has had to pay very high for new hires. Hiring managers used to need to get L10 approval to go out of bands on offers outside bands, but starting late 2021, they could go out of bands up to ~20% higher with no approvals.
To top all of this, Amazon has a 6% internal URA (non-regretted attrition) target: something that has been in place for close to a decade. The Amazon Music group revolted late 2021 as reported by the Big Technology publication in-depth [1] as they didn't have 6% of people to fire, at the very time when they could not hire. Yet they were forced to hit this URA target, just like all other orgs in Amazon. Apparently this “rebellion email” leaked across the company and is sparking Amazon-wide outrage, and adds fuel to the fire.
Amazon has been extremely frugal for the skillset they hire for, very demanding and have created a stressful workplace. They still force 6% of engineers to be fired, even when their own leadership opposes this (Amazon Music).
I wonder why they can’t hire. Also, why is Amazon still part of the FANG abbreviation when they have this culture across many of their organisations?
It’s an interesting question if the decline would have happened regardless of LLMs, just slower?
[1] An annotated visualization of the same data I did: https://blog.pragmaticengineer.com/are-llms-making-stackover...