I suspect its just a tax break related policy, ive been doing the same shit at US companies (still in EU), and some European companies (in EU) and the US one were the only one asking me to file reports for hours classified as R&D, maybe european ones have no incentive to do that
Id say, considering that, and considering that for these corps taxes are lava, i’d go to look at incentives on taxes about R&D filing
Can confirm, filed some “R&D” hours at a company some years back that did software development but nothing I’d call “R&D”. They straight-up told us it was a requirement for some tax benefit or other. Like they did have us do it only on non-client work, but… it was not really R&D the way anyone means that.
Same. R&D included anything that wasn't not finished or had a roadmap that extended beyond the end of the year. In other words: unproven. If you squint that is R&D.
Right, software development, but not software maintenance. The issue is with companies filing maintenance under development, since both of those has the same "Software Engineer" title. In other industries there is usually a very clear separation between development and maintenance.
Are you suggesting that a company that only does Research but no development (eg because they are selling the patents afterwards or so), should not be classified as doing R&D?
I took R&D as the bucket that we put both research and development into.
All business expenses can be written off. The difference between capital and operating expenses is in whether they are written off in the current year or depreciated over multiple years.
I have the opposite experience, I’m based in NL. Of the 2 EU companies I worked for they both asked to fill timesheets. And of the 2 US companies I worked for (also based in NL), none asked for timesheets.
Been doing the same in the UK. Everything not BAU was R&D. Installed a new server package and configured it? R&D baby. Or so I was told, because you know, it’s new to the company ;) ;)
France has a giant research tax break that is very wildly misused by corporations who put development in it.
There are even service companies with “research centers” now. They simply bill employees without a client to the state for doing “research” that mostly consists in completing tutorials on the latest trendy technology at the HQ.
The thing is $25k of software production is left at the company, and when the employee is not needed anymore, then the collectivity has to pay for unemployment
1. What publicly-traded companies report to investors and what they report to the IRS are more different than you might initially suspect.
2. This $73.2B figure was pulled directly from their 2022 10-K filing[1] under an ops expense line item labeled technology and content, which includes R&D and then some. The devil in the details is buried in a footnote on p. 26:
>> Technology and content costs include payroll and related expenses for employees involved in the research and development of new and existing products and services, development, design, and maintenance of our stores, curation and display of products and services made available in our online stores, and infrastructure costs. Infrastructure costs include servers, networking equipment, and data center related depreciation and amortization, rent, utilities, and other expenses necessary to support AWS and other Amazon businesses.
At face value, to handwave this figure as just R&D and purport that it's directly comparable to other publicly-traded companies who report disaggregated research and development strikes me as somewhere between shotgun analysis and hoodwinking.
Because you don’t want to give tax breaks for another todo app and another off the mill CRUD that nowadays can be generated by GPT or could be copied from a template and adjusted. Using a template and adjusting or generating with GPT already sounds more like operations.
You want to give tax breaks for real research and development of new things.
The market seems like it would preclude that. If they're making "another todo app" with nothing novel then the market will reject it, they will have no income, and they never would have been taxed anyways.
The thought of research tax breaks is that the market qualifies them as useful. Useless inventions make no money, and have no income to be taxed anyways.
Markets aren’t efficient like that though. In theory they’re supposed to be, but we’ve had wave after wave of “innovation” that’s the same shit with different paint.
Take a car, replace all the perfectly good knobs with touch screens you won’t be able to replace in 5 years. Innovation! Massive conglomerates control the internet and feed you expertly tuned rage engagement. Innovation! Smart home lights that stop working when the servers go offline. Innovation!
On and on. If the markets were rational and efficient NFTs wouldn’t have even started. Crypto wouldn’t be a thing. Printers that refuse to print black because you’re out of cyan wouldn’t exist.
Markets blow and aren’t really maximizing productive anything directly. Tangentially, occasionally, when they’re not causing cost savings that cause doors to blow off of planes to save money. Seems tenuous at best.
> we’ve had wave after wave of “innovation” that’s the same shit with different paint.
I think in most of your examples, you want something different than what the market values, and me too! But our values are different than the average person. I'm willing to fiddle with running a home server to do those things, but most people aren't.
> Printers that refuse to print black because you’re out of cyan wouldn’t exist.
They do because the crappy business practices let them sell at a lower upfront cost. That appeals to enough people for the company to make money.
> Markets blow and aren’t really maximizing productive anything directly. Tangentially, occasionally, when they’re not causing cost savings that cause doors to blow off of planes to save money. Seems tenuous at best.
The markets optimize for profit. That generally comes from innovations in something; products, processes, equipment, who knows. I won't pretend the markets are perfect, or efficient. What I will say is that I don't have a better proposal. Planned economies have historically turned out much worse than free markets.
I think there are compelling arguments that we're short on regulations in a fair few places. I'm more leery of arguments that the government should be making broad definitions of "innovation", it veers a little too close to central planning for my liking.
If I build a new assembly building for a new product, that roof is going to be mainstream, long lasting, cheap as I can get it, and billed to “development” because that’s exactly what it is.
You can say the same standing up a CD/CI infrastructure, writing unit tests, mundane documentation and validation processes. It’s all development, getting stuff into production and into the market.
> Research and development is the set of innovative activities undertaken by corporations or governments in developing new services or products, and improving existing ones.
I’m responding to how Amazon categorizes their R&D expenditures (which is based on IRS’s definition, not the Wikipedia definition).
If we don’t understand how Amazon categorizes R&D, we can’t make a fair comparison to France.
If anything, IRS’s (very) broad definition of R&D helps France in that Amazon’s R&D spend isn’t truly R&D in the colloquial sense (or the Wikipedia definition)
In other words, Amazon’s R&D spend is likely overstated (based on what an average person would consider “R&D”)
For the concept of R&D to mean anything, the D has to imply development in the sense of translating the R part—research—into prototypes or new products. It can't just be classifying every web developer in the organization as part of the R&D department, even if they're a junior dev changing button labels on internal accounting software. Not saying that Amazon is using the term the way I'd like, just that if they aren't, my English degree is going to start smoking and vibrating.
R&D means something very specific when it comes to tax implications (this context). Most companies I've worked for have included all software development time as R&D (including stuff like modifying buttons). Don't try to force some book definition on it, unless that book is the tax code for R&D tax credits, it doesn't work.
Research and Development ( R&D ) tax relief supports companies that work on innovative projects in science and technology. You may be able to claim Corporation Tax relief if your project meets the standard definition of R&D .
Section 174 is a harebrained law aimed solely at "punishing" successful engineering companies. It is a Trump-era law designed to appease voters mad at what they view as successful big tech "intelligentsia." It will destroy startup culture.
I can't wait until we automate the people cheered this law on, out of a job.
As a software developer I also salute that law, As italian I always voted against governments who wanted to detax r&d, if its done and funded by governments for the collectivity, then it benefits everyone, but if a private company is going to r&d for the benefits of their shareholders, then they can also pay their fair share of taxes instead of leeching from the collectivity
I always felt, from outside, that anti trumpers just are against everything he does with simplistic explanation, its a good law amazon has enough money to pay their own r&d
Those are good points imho, I think If they make a profit with it, then they can also pay for it, its useful for society and society appreciate it with the money they give to it, no reason to pay for it with taxes
Some products also only make money after a long time of R&D. If you are a startup focused on such a product, R&D tax benefits are critical to keep you alive during your long development time that will require multiple rounds of investment before you can even think of making a profit.
For example, a new drug can take 10~16 years to get to the market, and this is intended by government regulations (there is often no way around it). One has to show the drug is safe and works against the intended disease, before one can sell a drug.
However, before you can even start this clinical work in humans, you must discover the drug first and optimize it through laboratory and animal testing. Only after that can you enter the clinical stage to test it in humans. Clinical stages are typically divided into phase 1/2/3 and each stage requires more time and money. After you complete all that, you can finally ask for approval to sell your drug at a profit from the government (US Food and Drug Administration (FDA)/ EU European medicine agency (EMA)).
This trajectory can typically be divided as follows:
Drug Discovery (early lab testing) 2-4 years, $2-15M
Preclinical Drug Development (advanced lab/animal testing) 2~3 years, $10-15M
Clinical Stage 1 (First in human studies) 1~2 years $15~30M
Clinical Stage 2 (medium-sized human study), 2~3 years, $40~60M
Clinical Stage 3 (large, longer term human study), 2~4 years $100~300M.
Total: 9~16 years of research & development with a cost of $167~420M.
Then finally you can sell the drug for a profit. If you are a startup focused solely on such development, you have no way (or very limited ways) to make money/profit in those 9-16 R&D years. In such a case, R&D tax benefits are really useful.
> I always felt, from outside, that anti trumpers just are against everything he does with simplistic explanation
Sounds like quite a broad brush. As a former Trump voter (to my shame), I realized he's a broken clock that can be right once in a while; yet still such a profoundly net negative that I'd rather he be in jail than in office.
In particular I'm most against his record of sexual assault, association with predators, peeping at young pageant participants, boasting about such things, bending the law and policy to benefit his friends, praise for despots, his demand for loyalty over talent, and racist commentary.
I realised after writing it that my thought was a bit out of touch, in Italy I had my Trump, in Berlusconi, and I hated everything he said, even if something made sense
Section 174 has nothing to do with "detaxing R&D".
The US tax code already requires that self-classified R&D expenditures be treated as capital-expenditures, and subject to a 5-15 year depreciation period.
Section 174 requires that any and all software development costs be treated as R&D expenditures, and thus capitalized/amortized.
The total tax paid at the end of the depreciation period will be the same, but the upfront tax in the first year of a given expenditure is now 5-15x higher (depending on where the work was done).
> I can't wait until we automate the people cheered this law on, out of a job.
If your goal was "foment another wave of far-right nationalism in America", this is a good idea. Judging by the rest of your post, however, I don't think that's what you had in mind.
In Australia we have a tax break that means 80% of all R&D costs are repaid by the government. Everything we can ever possibly classify as R&D is classified as R&D. If I take a dump on company time it's R&D because I'm a software dev and therefore thinking about developing stuff while I'm in there.
That's not actually allowed, it's meant to be research with an "unknown outcome" or some similar wording. But lots of companies ignore yhat and claim regardless.
Interpreting that is my least favourite part of the R&D tax incentive - as far as I'm concerned practically everything has an unknown outcome. I can't figure out what we're doing in software and sometimes that even stretches to include after a product has been built. Most ideas have been had before in some form or another - it is just nobody at this company knows about it.
But genuine R&D would look very similar. There isn't a good test to distinguish "I don't know how to do this thing" from "nobody knows how to do this thing". I don't have the time, resources or inclination to know what everyone else knows. And then having to second guess what the tax office thinks everyone knows, because I know they'd accept some of the things I'm doing as research. It just isn't clear where they think the line is.
Wait so if I'm a software dev in Australia whose hours are classified as R&D, and my employer pays a 20% tax rate, can I file my withholdings reduced by 80% too?
It's a process where you claim the rebate from the government. You need to have an ABN, and I think be incorporated (? not sure). And then, yes, you can claim away to your heart's content.
When I worked there I had to fill in the forms for my team (Internal tools development). R&D was for all new feature development, time spent working on tickets/support/etc. were all non-R&D. Typically a 6:1 ratio R&D:not.
I don't know about your thorium power plant but Tesla is giving you battery tech, and John Deere (~2B$ in R&D) is essentially giving you a fleet of robot farmers.
These are not flying cars, but batteries that are 5 times cheaper in 10 years is impressive. I don't know much about farming but it looks like now, tractors do most of the job, and the guy in the cab is just checking to see if everything goes well. Farmers are still needed, be we need less and less of them for bigger and bigger fields.
Tesla and Panasonic and CATL are doing R&D on batteries. And Tesla and Panasonic are known long-term partners. So, how do they share the R&D? Do they research different topics (chemistry, manufacturing, etc...), does Tesla pays Panasonic to do the R&D? Directly or indirectly?
Tesla must have played a major role regarding battery technology. Other EV manufacturers too, but I suspect Tesla more than the others since they were the biggest player for quite a while (and still are by some metrics), have loads of money, and offer battery products other than cars.
Cost of Sales: 272,344
Fulfillment: 75,111
Technology and Content: 73,213
Sales and Marketing: 32,551
General and Administrative: 8,823
Other Operating Expense (Income), Net: 62
So Technology and Content is the $73bn number.
I think Fulfillment covers warehouse and delivery costs.
I am really confused as to why this is considered as a tax break strategy. Operating costs directly lower earnings, where as R&D might have to be amortized over years - at least in certain european countries. Isnt this reinvesting in your business? Also spending (or donating) money lowers your taxes by lowering your earnings.
I mean, sure, maybe they are cooking their numbers.
My own hypothesis was that the first table and the second table don't measure the same thing. Which would mean the headline is comparing apples to oranges.
Seems to me like products with entirely different use cases. Alexa's not so good at generative text content, but ChatGPT isn't very good at triggering kitchen timers and home automation shortcuts via voice command.
There is nothing technically stopping ChatGPT from hitting up endpoints with JSON other than OpenAI's forced limitations.
It has excellent voice to text, excellent text to voice, excellent comprehension (it can understand vague commands like "set all the lights in the house to a Christmas color theme") and it knows how to generate JSON in a particular format that you can specify. That plus an output scraper that knows where to send the JSON to is pretty much all you need to make this work. It's so close to a commercial product that I can literally taste it. And yet here we are, having to hack together our own solutions... with local LLM's that are about as good as GPT 3.5.
For me, the pinnacle of smart home voice automation would be a fully local voice interpreter that runs on some low-power device that I already have (like, say, an Apple TV), that supports some fixed, explicit set of commands that are easy to reference and/or modify, that has rock-solid reliability and that runs deterministically.
To me, it feels like Siri/HomeKit (or even Home Assistant/Assist) has a much clearer path to that ideal than ChatGPT spitting out JSON.
taxes. the way that R&D is measured is very tied to the tax code in countries. If R&D is economically promoted (historically the USA has invested much more money into R&D than other nations per capita because other nations tend to favour tax credits as opposed to direct investment - which seems like it opens the door for foreign investment to control a large percentage early in a startups life in exchange for tax credits after investment has dropped into their borders)
Maybe they need to spend even more to figure out how to show Adidas shoes when I search “Adidas shoes”. Seems that’s a really hard CS problem. Once they have that, maybe they can show products ordered by price when I select “Sort by price low to high”.
Tossing another anecdote onto the pile, I just searched "Adidas shoes". The first four hits are not for Adidas shoes, after that they all are.
All four of those non-Adidas hits are "Sponsored" results, but that's not particularly easy to notice (for one of them it's almost impossible).
I would agree with GP that Amazon product search often leaves much to be desired. It's hard to search for X without getting accessories for X, competitors to X, products that can be connected to X. This is annoying at best, and crippling if you want to, say, sort by price (all the cheapest matches will be accessories or otherwise not what you wanted).
Now that Amazon has captured user-share, they will (ineluctably) abuse user good-will (by padding search results with ad-placed products the user does not want) until they reach some marginal breaking point. The users that remain are either oblivious or cursed with a lingering taste of bad-faith and bullshit.
Eventually the giant will strangle itself in its sleep, but that's probably 50+ years out given how these things go.
What I mean by it is that some previous interaction with amzn has left behind some state that leads to a markedly different search result than for other people.
I can see (because I was there discussing it back in 95) what the appeal of personalized search results could be for both sellers and consumers. But a few decades of that BS have fairly convincingly made the case that it is a horrible idea.
One thing I absolutely cannot understand is the "price range" functionality seems to be broken as well. Maybe it's the same root cause and/or change which broke price sorting but when I put in a range of say $100-$200 searching for "network switch" the first result is $240, the 2nd result $90, and the 3rd result $120 and legitimate results go on for tens of pages. So it's not that there isn't anything for sale in that range, it's not that there weren't enough results to display so they needed to pad the numbers, and it's (seemingly?) not that they are just trying to push a higher dollar sale... it's just plain broken?
Then call the category ~200. IMO This is prime enshitification. You cant search for anything reliably any more. You dont get to see what youre actually searching for.. just whatever a company thinks they can make more profit off of.
Yep. Google did it with their search and now it’s just what makes the world turn. And they have to be disingenuous about it or else people wouldn’t use it. I think it’s pretty plain and simple - realise this is the case, then vote with your feet
I’ve stopped trying to save by online ordering, within reason. If a store has what I need + points me in the right direction and demonstrates they are knowledgeable, they will get my business. 5-10 dollars is not worth saving at the expense of fuelling the enshittification machine. If that makes a huge difference to you, then reevaluate if you actually need the item
You’ll end up with less stuff, better stuff, with less hassle and noise, and a relationship to a vendor which may speed up returns and maintenance. As always, pick your poison
It isn’t uncommon to say your price band is $100-200, then find something costing $250 which checks all the boxes. In this case, fuzzing prices while presenting results seems like an acceptable trade off.
As a user, when I set a price range, I want the search function to return results within that range. I don't want fuzz. I don't want you to second guess me. I want the results I asked for.
This is the natural, simple, user-friendly design I want.
What if you (the user not you you) don’t know what price range will net them the feature set they want?
E.g., would you know exactly the price range to pay for a tool set which lets you do general repairs around the house? Should it be ₹100? ₹500? How will a search engine know for sure that you’re an aware customer who knows their price ranges versus a newbie exploring the space? Even google claims to have millions of results per query meaning they, for all their talk about organising the world’s information, don’t know what exactly you want.
It’s all about trade offs and checkboxes(layovers, specific terminals, airlines, aircraft body types, price ranges, dates, non-stops, insurance, in-flight Wi-Fi, meals, early boarding) etc.
Do you first look through the page with all the options, then open different tabs with each unique combination of checkboxes and options enabled/disabled? That must exponentially (literally) harder than simply viewing fuzzy results on screen and pruning the decision tree in your head.
I input my source and destination and date and travel class and then choose from the listed results. At no point I would expect the website to show me business class if I chose economy. If I chose non-stop, I expect non-stop flights and nothing else.
If I say I want to fly on Tuesday, 2/2/2029, with a departure between 8-11am, economy, and ask you to show me options sorted by cost (descending), you should show me exactly those results.
You could offer me a flexible search option, as some sites do (+/- x days on flight times, etc.). Great! Now I can request that we search within a date range. Nice! That might be a good feature for some people!
Offering a simple query interface that returns what it says it will doesn't mean we can't offer flexible search parameters.
Flights are a terrible example of where I'd want to offer fuzzed results... having booked flights on the wrong day before, the last thing I want is to ask for X, but get offered Y.
There are so many decisions other than just the dates that you implicitly make (or are made for you when you book your flight). Do you want insurance? Wi-Fi? Extra leg room? Turbo prop aircraft? Jets only? 2-3-2 seating? 3-3 seating?
Logically, you should want to control each of these variables while searching for flights. If you are ready to trust a flight aggregator with these choices, some fuzziness in price seems acceptable too.
You are probably thinking that you don’t want to make a choice about aircraft type or seating configuration and so, whatever the search engine shows in whatever order is fine by you. If so, consider that others might have the same ambivalence towards price fuzziness.
I can kinda understand (though not always agree with) fuzzy matching in the case there are 0 results that match the given filters. Something like a heading entry "No matches found. Showing closest matches" or maybe even just always sticking them at the end of short results lists. I can't understand just assuming every match for any search should be fuzzy just because there might be a use case it helps for certain searches. Or even a "fuzzy match" type filter option, though I think you'd find extremely few people actually use it.
I.e. the UI needs to be there to efficiently help the user do what they ask not to try to be smarter than the user. Unless the user asks for fuzzy matches it doesn't make sense to make the response lead with them.
FWIW I just tried that exact search and every item I see listed at above $200 also has a "more buying options" listed on the search page with a price less than $200, so while it's annoying I guess I get the logic.
All those options (the ones I see anyway) are used, so when I limit the search to "Condition - New" they all go away.
I don't think your argument holds water. A while back I read some article about how search is handled in companies like Amazon and surprisingly one requirement they have is appeasing key sellers. There were major political battles driving the decision on what search results should be boosted in specific search keywords just because a representative for company A insisted quite strongly that should be the expected result, while rep from company B had a very different point of view.
I'm sure Amazon does optimize results based on conversion, but only fools believe search results are optimal based on concrete criteria.
Also, some search results are clearly awful. Once I saw canned fish items showing up while searching for USB drives.
Recognize names of well-known brands in search queries and filter (or at least prioritize) items by the associated seller? It's not exactly black magic.
Those are questions for Amazon. I can give you all sorts of arbitrary criteria, but whatever they choose it would be ridiculous for Adidas to not be on the list.
Yes. I was looking for a vegetable dicer yesterday and gave up and just got one at Kmart, because Amazon was page after page of the exact same one over and over that had horrible reviews.
Their search doesn't even seem to understand negative search terms either.
eg searching for 25GbE network cards shows a bunch of 2.5GbE network cards, and no amount of -"2.5GbE" nor "NOT 2.5GbE" options (and variations) gets rid of them. :(
Like many of the comments here, I'm very curious what counts as "R&D" spend by Amazon. My guess (and happy to be corrected if I'm wrong) is that it includes a big chunk of AWS capex spend. I think it's great that Amazon invested so much into AWS, but it then also isn't surprising that the company that runs a huge portion of the Internet has large "R&D" spend.
If you imagine a tech company that declares that they're feature-complete, and they cut expenses to those required to maintain existing service to a high degree of quality/reliability, that seems like a reasonable benchmark of spending that could be classified as "not R&D".
Feel obliged to point out this is misleading / not especially relevant: GDP is the value of the annual output of a country, while market capitalization is the (discounted) value of all future cash flows of a company. LVMH is ~0.5T market cap French company, but it’s nowhere near 1/6th of French economic output.
This is a lot like the recent “Novo Nordisk’s market cap is greater than the GDP of Denmark!” comparisons — true, but they’re not like numbers (and indeed, much of Novo Nordisk’s revenue is captured in Danish GDP).
Please remember that any publicly-traded company must follow guidelines, US-based or otherwise. There are rules for what constitute R&D for those who want to understand why 'company asked me to fill a report'
Non-US based publicly traded companies would likely not have to follow guidelines of the American IRS?
(Though, of course, that might be wrong. American SEC tries to have global jurisdiction, whether you want them to apply to you or not. Matt Levine writes about that a lot in his 'Money Stuff'.)
The "D" in R&D is development, i.e. software engineers writing day-to-day product code. So R&D budgets cover the majority of employee salaries and even stuff like cloud hosting. It's not earmarked for groundbreaking research as people seem to think.
Perhaps when companies or departments get to that scale they should get the responsibilities of a country like France. think employee protections, voting about decisions and direction, etc...
This is a smart accounting trick to work around taxes. No doubt Amazon and others do a lot of R&D. But when such comparisons are made with Europe, some details need to be understood.
160 comments
[ 1.4 ms ] story [ 217 ms ] threadId say, considering that, and considering that for these corps taxes are lava, i’d go to look at incentives on taxes about R&D filing
It's both hilarious and disheartening how much of a tax grift it is.
I mean I work in a RnD dep. now and it honestly is some R but like 2% of my time.
I took R&D as the bucket that we put both research and development into.
https://en.wikipedia.org/wiki/Technology_readiness_level
This doesn't seem like a profit-driven decision.
Bench / sub 100% utilization is one reason consultants cost so much.
1. What publicly-traded companies report to investors and what they report to the IRS are more different than you might initially suspect.
2. This $73.2B figure was pulled directly from their 2022 10-K filing[1] under an ops expense line item labeled technology and content, which includes R&D and then some. The devil in the details is buried in a footnote on p. 26:
>> Technology and content costs include payroll and related expenses for employees involved in the research and development of new and existing products and services, development, design, and maintenance of our stores, curation and display of products and services made available in our online stores, and infrastructure costs. Infrastructure costs include servers, networking equipment, and data center related depreciation and amortization, rent, utilities, and other expenses necessary to support AWS and other Amazon businesses.
At face value, to handwave this figure as just R&D and purport that it's directly comparable to other publicly-traded companies who report disaggregated research and development strikes me as somewhere between shotgun analysis and hoodwinking.
[1] https://www.sec.gov/Archives/edgar/data/1018724/000101872423...
You want to give tax breaks for real research and development of new things.
The thought of research tax breaks is that the market qualifies them as useful. Useless inventions make no money, and have no income to be taxed anyways.
Take a car, replace all the perfectly good knobs with touch screens you won’t be able to replace in 5 years. Innovation! Massive conglomerates control the internet and feed you expertly tuned rage engagement. Innovation! Smart home lights that stop working when the servers go offline. Innovation!
On and on. If the markets were rational and efficient NFTs wouldn’t have even started. Crypto wouldn’t be a thing. Printers that refuse to print black because you’re out of cyan wouldn’t exist.
Markets blow and aren’t really maximizing productive anything directly. Tangentially, occasionally, when they’re not causing cost savings that cause doors to blow off of planes to save money. Seems tenuous at best.
I think in most of your examples, you want something different than what the market values, and me too! But our values are different than the average person. I'm willing to fiddle with running a home server to do those things, but most people aren't.
> Printers that refuse to print black because you’re out of cyan wouldn’t exist.
They do because the crappy business practices let them sell at a lower upfront cost. That appeals to enough people for the company to make money.
> Markets blow and aren’t really maximizing productive anything directly. Tangentially, occasionally, when they’re not causing cost savings that cause doors to blow off of planes to save money. Seems tenuous at best.
The markets optimize for profit. That generally comes from innovations in something; products, processes, equipment, who knows. I won't pretend the markets are perfect, or efficient. What I will say is that I don't have a better proposal. Planned economies have historically turned out much worse than free markets.
I think there are compelling arguments that we're short on regulations in a fair few places. I'm more leery of arguments that the government should be making broad definitions of "innovation", it veers a little too close to central planning for my liking.
If I build a new assembly building for a new product, that roof is going to be mainstream, long lasting, cheap as I can get it, and billed to “development” because that’s exactly what it is.
You can say the same standing up a CD/CI infrastructure, writing unit tests, mundane documentation and validation processes. It’s all development, getting stuff into production and into the market.
https://en.wikipedia.org/wiki/Research_and_development
https://www.irs.gov/pub/irs-regs/research_credit_basic_sec41...
(Not 100% sure this is the most up to date definition but it gives you an idea)
If we don’t understand how Amazon categorizes R&D, we can’t make a fair comparison to France.
If anything, IRS’s (very) broad definition of R&D helps France in that Amazon’s R&D spend isn’t truly R&D in the colloquial sense (or the Wikipedia definition)
In other words, Amazon’s R&D spend is likely overstated (based on what an average person would consider “R&D”)
Not in the UK and most EU law is identical to UK.
I can't wait until we automate the people cheered this law on, out of a job.
I always felt, from outside, that anti trumpers just are against everything he does with simplistic explanation, its a good law amazon has enough money to pay their own r&d
What if the R&D is open-sourced?
Or what if the R&D would be for a product genuinely beneficial to people? Think a new drug, or infrastructure like Google Maps, etc.
For example, a new drug can take 10~16 years to get to the market, and this is intended by government regulations (there is often no way around it). One has to show the drug is safe and works against the intended disease, before one can sell a drug.
However, before you can even start this clinical work in humans, you must discover the drug first and optimize it through laboratory and animal testing. Only after that can you enter the clinical stage to test it in humans. Clinical stages are typically divided into phase 1/2/3 and each stage requires more time and money. After you complete all that, you can finally ask for approval to sell your drug at a profit from the government (US Food and Drug Administration (FDA)/ EU European medicine agency (EMA)).
This trajectory can typically be divided as follows: Drug Discovery (early lab testing) 2-4 years, $2-15M Preclinical Drug Development (advanced lab/animal testing) 2~3 years, $10-15M Clinical Stage 1 (First in human studies) 1~2 years $15~30M Clinical Stage 2 (medium-sized human study), 2~3 years, $40~60M Clinical Stage 3 (large, longer term human study), 2~4 years $100~300M.
Total: 9~16 years of research & development with a cost of $167~420M. Then finally you can sell the drug for a profit. If you are a startup focused solely on such development, you have no way (or very limited ways) to make money/profit in those 9-16 R&D years. In such a case, R&D tax benefits are really useful.
Sounds like quite a broad brush. As a former Trump voter (to my shame), I realized he's a broken clock that can be right once in a while; yet still such a profoundly net negative that I'd rather he be in jail than in office.
In particular I'm most against his record of sexual assault, association with predators, peeping at young pageant participants, boasting about such things, bending the law and policy to benefit his friends, praise for despots, his demand for loyalty over talent, and racist commentary.
The US tax code already requires that self-classified R&D expenditures be treated as capital-expenditures, and subject to a 5-15 year depreciation period.
Section 174 requires that any and all software development costs be treated as R&D expenditures, and thus capitalized/amortized.
The total tax paid at the end of the depreciation period will be the same, but the upfront tax in the first year of a given expenditure is now 5-15x higher (depending on where the work was done).
If your goal was "foment another wave of far-right nationalism in America", this is a good idea. Judging by the rest of your post, however, I don't think that's what you had in mind.
How the huge mining companies get away with claiming so much is beyond me. But they do.
But genuine R&D would look very similar. There isn't a good test to distinguish "I don't know how to do this thing" from "nobody knows how to do this thing". I don't have the time, resources or inclination to know what everyone else knows. And then having to second guess what the tax office thinks everyone knows, because I know they'd accept some of the things I'm doing as research. It just isn't clear where they think the line is.
e.g. https://aws.amazon.com/solutions/agriculture/
These are not flying cars, but batteries that are 5 times cheaper in 10 years is impressive. I don't know much about farming but it looks like now, tractors do most of the job, and the guy in the cab is just checking to see if everything goes well. Farmers are still needed, be we need less and less of them for bigger and bigger fields.
Tesla and Panasonic and CATL are doing R&D on batteries. And Tesla and Panasonic are known long-term partners. So, how do they share the R&D? Do they research different topics (chemistry, manufacturing, etc...), does Tesla pays Panasonic to do the R&D? Directly or indirectly?
Tesla must have played a major role regarding battery technology. Other EV manufacturers too, but I suspect Tesla more than the others since they were the biggest player for quite a while (and still are by some metrics), have loads of money, and offer battery products other than cars.
I get that they have high R&D costs but I'm still trying to get my head around how that money is being spent.
This bit:
So Technology and Content is the $73bn number.I think Fulfillment covers warehouse and delivery costs.
"Technology and Content" presumably mixes in Amazon Prime studios, which apparently was $16bn in 2022: https://techcrunch.com/2023/02/03/amazon-ramped-up-content-s...
My guess is that it covers AWS infrastructure too.
https://news.ycombinator.com/item?id=38120388
https://pro.bloombergtax.com/brief/rd-tax-credit-and-deducti...
I suspect the numbers are wrong.
My own hypothesis was that the first table and the second table don't measure the same thing. Which would mean the headline is comparing apples to oranges.
There is nothing technically stopping ChatGPT from hitting up endpoints with JSON other than OpenAI's forced limitations.
It has excellent voice to text, excellent text to voice, excellent comprehension (it can understand vague commands like "set all the lights in the house to a Christmas color theme") and it knows how to generate JSON in a particular format that you can specify. That plus an output scraper that knows where to send the JSON to is pretty much all you need to make this work. It's so close to a commercial product that I can literally taste it. And yet here we are, having to hack together our own solutions... with local LLM's that are about as good as GPT 3.5.
To me, it feels like Siri/HomeKit (or even Home Assistant/Assist) has a much clearer path to that ideal than ChatGPT spitting out JSON.
No regulation is bad or worse. Even less accountability.
All four of those non-Adidas hits are "Sponsored" results, but that's not particularly easy to notice (for one of them it's almost impossible).
I would agree with GP that Amazon product search often leaves much to be desired. It's hard to search for X without getting accessories for X, competitors to X, products that can be connected to X. This is annoying at best, and crippling if you want to, say, sort by price (all the cheapest matches will be accessories or otherwise not what you wanted).
Now that Amazon has captured user-share, they will (ineluctably) abuse user good-will (by padding search results with ad-placed products the user does not want) until they reach some marginal breaking point. The users that remain are either oblivious or cursed with a lingering taste of bad-faith and bullshit.
Eventually the giant will strangle itself in its sleep, but that's probably 50+ years out given how these things go.
I can see (because I was there discussing it back in 95) what the appeal of personalized search results could be for both sellers and consumers. But a few decades of that BS have fairly convincingly made the case that it is a horrible idea.
I’ve stopped trying to save by online ordering, within reason. If a store has what I need + points me in the right direction and demonstrates they are knowledgeable, they will get my business. 5-10 dollars is not worth saving at the expense of fuelling the enshittification machine. If that makes a huge difference to you, then reevaluate if you actually need the item
You’ll end up with less stuff, better stuff, with less hassle and noise, and a relationship to a vendor which may speed up returns and maintenance. As always, pick your poison
This is the natural, simple, user-friendly design I want.
Don't give me some bullshit instead.
So many modern UX issues are from people presuming what users want, rather than just giving them well-designed tools and getting out of the way.
E.g., would you know exactly the price range to pay for a tool set which lets you do general repairs around the house? Should it be ₹100? ₹500? How will a search engine know for sure that you’re an aware customer who knows their price ranges versus a newbie exploring the space? Even google claims to have millions of results per query meaning they, for all their talk about organising the world’s information, don’t know what exactly you want.
It’s all about trade offs and checkboxes(layovers, specific terminals, airlines, aircraft body types, price ranges, dates, non-stops, insurance, in-flight Wi-Fi, meals, early boarding) etc.
Do you first look through the page with all the options, then open different tabs with each unique combination of checkboxes and options enabled/disabled? That must exponentially (literally) harder than simply viewing fuzzy results on screen and pruning the decision tree in your head.
I input my source and destination and date and travel class and then choose from the listed results. At no point I would expect the website to show me business class if I chose economy. If I chose non-stop, I expect non-stop flights and nothing else.
You could offer me a flexible search option, as some sites do (+/- x days on flight times, etc.). Great! Now I can request that we search within a date range. Nice! That might be a good feature for some people!
Offering a simple query interface that returns what it says it will doesn't mean we can't offer flexible search parameters.
Flights are a terrible example of where I'd want to offer fuzzed results... having booked flights on the wrong day before, the last thing I want is to ask for X, but get offered Y.
Logically, you should want to control each of these variables while searching for flights. If you are ready to trust a flight aggregator with these choices, some fuzziness in price seems acceptable too.
You are probably thinking that you don’t want to make a choice about aircraft type or seating configuration and so, whatever the search engine shows in whatever order is fine by you. If so, consider that others might have the same ambivalence towards price fuzziness.
I.e. the UI needs to be there to efficiently help the user do what they ask not to try to be smarter than the user. Unless the user asks for fuzzy matches it doesn't make sense to make the response lead with them.
All those options (the ones I see anyway) are used, so when I limit the search to "Condition - New" they all go away.
How do you define a Adidas shoe? Something with the same name could easily be fake
The problem is, they don't.
Would you believe what drives more sales and what you think you want could be different?
I'm sure Amazon does optimize results based on conversion, but only fools believe search results are optimal based on concrete criteria.
Also, some search results are clearly awful. Once I saw canned fish items showing up while searching for USB drives.
What's an associated seller?
Can you see where this would not work, or how it can be exploited ?
https://arstechnica.com/ai/2024/01/lazy-use-of-ai-leads-to-a...
eg searching for 25GbE network cards shows a bunch of 2.5GbE network cards, and no amount of -"2.5GbE" nor "NOT 2.5GbE" options (and variations) gets rid of them. :(
Ebay though seems to work ok. :)
If you imagine a tech company that declares that they're feature-complete, and they cut expenses to those required to maintain existing service to a high degree of quality/reliability, that seems like a reasonable benchmark of spending that could be classified as "not R&D".
https://finance.yahoo.com/quote/AMZN/
https://data.worldbank.org/indicator/NY.GDP.MKTP.KD.ZG?locat...
This is a lot like the recent “Novo Nordisk’s market cap is greater than the GDP of Denmark!” comparisons — true, but they’re not like numbers (and indeed, much of Novo Nordisk’s revenue is captured in Danish GDP).
https://tradingeconomics.com/france/employment-rate
That isn't a comparison that makes sense. It's sort of like confusing stocks and flows, but market capitalization isn't even a stock.
https://www.irs.gov/businesses/corporations/irc-41-asc-730-r...
(Though, of course, that might be wrong. American SEC tries to have global jurisdiction, whether you want them to apply to you or not. Matt Levine writes about that a lot in his 'Money Stuff'.)