FRED enables combining (and tweaking/scaling/etc) charts pretty easily on their site. Check out the 'edit graph' button. You can use the ID's of the individual charts (like IHLIDXUSTPSOFTDEVE) when searching for a table to add.

Here are those 4 tables combined: https://fred.stlouisfed.org/graph/?g=1EDHP

Nice tip. Thanks.

There's also section-174, which big corporations don't care about but kills smaller businesses left and right.

Yeah 174 is crazy: https://youtu.be/1ecu0YsCGxg

Never listen to theo.

Maybe... :-) Where is he wrong?

I mean, section 174 was a big deal to the large corp I'm in when it was getting started. We had to clarify R&D time separately.

I hear a lot of static about this on hacker news and nowhere else.

I work for a large company and we were already capitalizing the time we spent developing software. From an accounting perspective it makes sense that software development effort produces a semi-durable asset that a company can use, sell, etc.

For smaller companies, the math is the same.

For startups very specifically, who expect to go to the moon or go bust in 5 years, I can believe that having a lower initial burn rate (because they could represent all developer salaries as opex and deduct in the immediate calendar year) was beneficial.

Also, the tax law changes actually disincentivize offshoring software development by requiring an absurdly long amortization schedule for foreign-developed software. Yet offshoring is currently as intense as it has ever been since the early 2000s. (If we expect the tax law changes to have a big effect on developer hiring, we would expect a decrease in offshoring given the disincentives).

I wonder how much of this drop can be simply explained as having reached a certain industry maturity. I assume at some point most of the fundamentals are in place, so we need less people?

Programming isn't an industry, per se. It's a skill, like accounting, sales, communications or law. There are definitely many businesses for which it's a core strength, but any large company has some need for it.

Maturity is definitely an issue. If you listen to what the industry is selling these days it's pretty much just AI. Before that it was DeFi and VR. There hasn't been a new smartphone since.. the smartphone.

I have more thoughts here: https://news.ycombinator.com/item?id=41495082

I think programmer is the last job I would replace with AI.

That sounds like something a programmer would say.

That's funny, but I also don't doubt that the industry, should it pursue this path, is going to be on a hiring spree to get engineers back to fix the mess they've wrought a few years from now.

It's the idea programmers should be last which is kind of arrogant/funny.

First we replace judges, doctors, police, generals, state officials, politicians, etc etc, and only then the programmer, the true foundation of our society!

(It shouldn't need stating that replacing any of these professionals with AI systems as they exist today is a terrible idea. Yes, even the politicians)

I probably wouldn't replace any of those with a system that excels in generating plausibly sounding bullshit.

LLMs are good at processing natural language. I think only a few human tasks are suitable to be fully replaced with AI. Things like translation might be a good candidate, with some proofreading.

Journalism is an obvious one, since most journalism today is just rewriting information gathered from other media or social media, the same thing LLMs do. Insightful opinion writers or actual investigative journalists would be an exception.

I could see LLMs easily doing the job of newspaper sports reporting, for instance, taking the stats and a few main facts from last night's game and turning it into a narrative article. It just wouldn't be good at long-form pieces that go into depth about motivations and strategy.

"Pursuing this path" kicked-off 2 years ago when the tech industry collectively conducted the first round of unprecedented-in-scale (until the following year) 'right-sizing'.