Seasoning and profitability rules are why S&P does not have as steep of a drawdown or as long as a recovery as Nasdaq over the last 30 years of market performance.
The S&P recovered from Dotcom bottom in ~7 years while the Nasdaq-100 took 15 years. Likewise Nasdaq took 3.5 years and the AI hype to recover back to its COVID highs in 2024 while S&P had the same recovery in about 2 years.
This is the downside to Nasdaq having higher returns in tech bull markets.
So the indices have a very different volatility profile by design, we should be happy to have the choice rather than have them all converge to the same product.
They've gotten far more aggressive in the last year.
I've been subscribed for 15+ years, at one point I added the food subscription because my wife wanted it.
First of all they log you out way more often than they ever did, so constantly prompted to log back in. This is of course because after login they try to up-sell you to more packages.
I am more likely to exit entirely than add more packages at this point.
The investment case for the AI labs is to point at recent hockey stick revenue growth and say "this, til 2030".
So to hit that you would assume you'd need - a wider base of use cases outside SWE, a wider user base in each use case, and higher usage per user.
Meanwhile, we have had a drumbeat of headlines of tech & non-tech companies putting in spending monitoring, caps, model downgrades, etc since the end of Q1... mostly on the primary good agreed use case - SWE assistance.
And this ignores the AI labs costs to provide the models entirely, this is purely whether revenue can keep going to the moon or not.
Its interesting that we've already gotten to the point sama is acknowledging the issue.
In some ways.. since Microsoft is known for maintaining backwards compatibility whereas Apple is not, I think 3rd party devs are just not incentivized to care about Windows ARM compatibility.
Further, it doesn't seem like Microsoft made x86 emulation as seamless or performant as Apple did during the various MacOS CPU architecture changes.
Every use case I've looked at has been a minefield of app incompatibility and poor performance under x86 emulation.
"Enterprise" people. There's this whole other world of legacy enterprise software where people do things that run companies, write large-scale software with an ancient version of .NET, and run this software on Windows servers.
... yet they still struggle to try and port their code to a newer .net version that does run on ARM. And do a bunch of other work tasks that can utilize this hardware.
Here is a particularly irritating example. Cabinet Vision, some CAD type of software.
Requires ancient .NET. That actually is available for Arm though.
Required Jet DB driver 2010, which doesn’t exist on Arm, although it’s only needed for the installer.
Requires SQL Server embedded 2012 and 2016, which don’t exist on Arm at all. Yep, both versions.
Also required PowerShell version 2, which was deprecated in 2017, although they magically figured out how to fix that once Windows 10 was EOL’d and Win. 11 doesn’t support v2.
The vendor has zero plans to ever support this on Arm.
They will eventually get their lunch eaten by a new competitor who decides to just release a macOS version.
I try to avoid Microsoft products these days (modern dotnet is pretty nice though), but from my past experience, I believe every bit of this and have run into all of these with other software, including the requiring of multiple versions of MSSQL Embedded, which is just unholy.
I remember when they first pivoted from multiperson multitouch tables to tablets. It sounded like a really cool device - even got me to walk into a Microsoft store.
Then I realized that it used the same shitty Windows with the same shitty registry that I had mostly avoided for my whole life to that point. I certainly wasn't jumping in on that tablet.
> the software I might want to use doesn’t support arm
Doesn't Windows come with something like Apple's Rosetta to do on the fly translation? I expect it wouldn't work with games, but most other kinds of software should work.
Rosetta worked quite well for Apple so I would expect Microsoft could do something similar.
I kind of find it hard to believe they would use the name prism...
When I read it, the first thing I thought of is the NSA program named PRISM.
Anyway, I was curious so I googled for differences between the Apple and Microsoft programs and Apple included x86 translation circuitry on their CPU. I wonder why Microsoft didn't do the same?
I mean, I used to be - with the disclaimer that I worked at Microsoft for a while (left in 2019), there was a hot minute when Surface devices were good and on an upward trajectory to become great. Microsoft was doing interesting things with new form factors and interface devices-- the Surface Book, Studio and Dial weren't all hits, but they were some of the only noteworthy experiments in PCs-- and they actually cared about build quality in a way pretty much no other PC manufacturer did.
Then Panay left, Windows 11 has been a debacle, and Nadella seems to give zero fucks about anything which isn't Copilot or Azure, so the Surface momentum that they spent so much time building has just coasted to a complete stop. It's sad.
Moreover, the Surface Pro 1 and 2 used Wacom EMR styluses --- still regret not getting one, but then Samsung did the Galaxy Book 12 (which was about perfect), so I was _finally_ able to replace my Fujitsu Stylistic ST-4110 (the Toshiba Encore 2 Write was a necessary stop-gap).
These days I use a Samsung Galaxy Book 3 Pro 360 (w/ a spare which I panick bought when I wasn't sure if they would do a Book 4 --- now they're up to a 5), Kindle Scribe Colorsoft, Samsung Galaxy Note 10+, and have a Wacom One on my MacBook (both of which need upgrading....)
I do windows (Desktop) dev for work, and am interested in running local AI. I'd be really happy to have a machine with 64GB shared memory and a decent x64 emulation.
Sure but theres a level of uncertainty being expressed for a paid service that you don't see elsewhere.
Imagine if every time you booked an uber it was like "your drive may crash the car". Or whenever you ate at a restaurant, the waiter said "there is a chance the chef will poison you". Or your bank statement said something like "these numbers may be wrong".
Each side gets the smooth brains they crave because for ~50% of the population it's become a team sport / religion situation. A majority of people have not thought through 1/10th of the policy positions they automatically support/reject based on the team hat color.
There's a quote I will mangle and I forget the source of that's something like "If you agree with all the positions of your chosen political party, either you have thought through every option and came to the same conclusions on dozens of topics, or you haven't thought through anything".
> because for ~50% of the population it's become a team sport / religion situation
It's kinda even worse. We can only have two parties because of FPTP, and turnout is about 60% of voting-eligible population on average. We know from recent popular vote that that 60% is split roughly 50-50.
So 30% of the voting population is Team Blue Hat, and 30% is Team Red Hat.
If you can get 15.1% of the population to vote for an unserious clown, they'll win their respective party's nomination. And in most states, one of the parties is a pretty sure lock in the general.
15% is a pretty low bar. Compare: Up to 30% of people think "chemtrails" are at least a "somewhat" real thing, and 5% of people believe vaccines contain microchips. So the bar for getting a guaranteed win is somewhere between those two wacky and easily disproven beliefs.
The S&P recovered from Dotcom bottom in ~7 years while the Nasdaq-100 took 15 years. Likewise Nasdaq took 3.5 years and the AI hype to recover back to its COVID highs in 2024 while S&P had the same recovery in about 2 years.
This is the downside to Nasdaq having higher returns in tech bull markets.
So the indices have a very different volatility profile by design, we should be happy to have the choice rather than have them all converge to the same product.
reply