Has the decline of knowledge work begun?

I’m surprised not many people talk about this, but a big reason corporations are able to do layoffs is just that they’re doing less. At my work we used to have thousands of ideas of small improvements to make things better for our users. Now we have one: AI. It’s not that we’re using AI to make all these small improvements, or even planning on it. We’re just… not doing them. And I don’t think my experience is very unique.

De-scoping is also a commonly-cited result of higher interest rates. We recently lived through a prolonged episode of zero-interest-rate-policy (ZIRP), which encouraged long-term and risky projects. When interest rates go up, the minimum acceptable return-on-investment (ROI) required to lure investment money away from low-risk investments like government bonds also increases correspondingly.

Yeah I think there are a bunch of different reasons that corporations are doing less, interest rates are a major factor, but the point is that it isn’t exactly the “AI is taking our jobs” narrative that people want it to be.

Starting in 2022, US companies could not deduct SWE salary expenses in the same year, only over 5 years like hardware CapEx. For big companies, this will roll over in 2027. Meanwhile, LLM expenses can be written off immediately as OpEx.

https://news.ycombinator.com/item?id=37494601

This is probably the single dumbest aspect of US tax policy.

If you hire Canadian software engineers, you can dodge this and deduct the expenses in your Canadian subsidiary. If you outsource software dev to another company you can usually get away with expensing it.

Research outside the US = 15 years amortization?

https://www.grantthornton.com/insights/alerts/tax/2023/flash...

  The TCJA amended Section 174 by removing the option to expense SRE expenditures, instead requiring taxpayers to capitalize and amortize SRE expenditures over a period of five years (attributable to domestic research) or 15 years (attributable to foreign research)