Teams is a terrible app, although Electron isn't the only reason for that: It needs a Gig of RAM to do things that older chat apps could do in 4 Meg.

The free ride of ever increasing RAM on consumer devices is over because of the AI hyperscalers buying all fab capacity, leading to a real RAM shortage. I expect many new laptops to come with 8GB as standard and mid-range phones to have 4GB.

Software engineers need to start thinking about efficiency again.

I haven't used 3.1 yet, but 3.0 Pro has been frustrating for two reasons:

- it is "lazy": I keep having to tell it to finish, or continue, it wants to stop the task early.

- it hallucinates: I have arguments with it about making up API functions to well known libraries which just do not exist.

Yes, it is very lacking in details. The Claude output would have been interesting, or a few logs or protocol dumps.

The lack of detail makes me suspect the truth of most of the story.

wow! Thanks for that.

Not in my experience with Gemini Pro and coding. It hallucinates APIs that aren't there. Claude does not do that.

Gemini has flashes of brilliance, but I regard it as unpolished some things work amazingly, some basics don't work.

It's very hard to tell the difference between bad models and stinginess with compute.

I subscribe to both Gemini ($20/mo) and ChatGPT Pro ($200/mo).

If I give the same question to "Gemini 3.0 Pro" and "ChatGPT 5.2 Thinking + Heavy thinking", the latter is 4x slower and it gives smarter answers.

I shouldn't have to enumerate all the different plausible explanations for this observation. Anything from Gemini deciding to nerf the reasoning effort to save compute, versus TPUs being faster, to Gemini being worse, to this being my idiosyncratic experience, all fit the same data, and are all plausible.

You nailed it. Gemini 3 Pro seems very "lazy" and seems to never reason for more than 30 seconds, which significantly impacts the quality of its outputs.

There fans are still good, but not the quietest anymore.

Noctua no longer manufactures them, they are now made by YS Tech.

I don't understand the threat model that banks worry about on rooted phones.

What is it? I can access their websites on a PC running as root or Administrator. What is the problem with rooted Android phones?

Mine is not 200MB on Android - the base apk is 67MB + 32MB for the ARM v8a specific libs. This is the code, the local caching and other data might make up the rest.

For Android, you can check [1] Download the apk, rename it as a zip and look inside to see the files.

A quick file analysis of the 67MB shows around 58MB of java code and some 32MB of ARM libs, 31MB of this is the libvideochat.

[1] https://www.apkmirror.com/apk/google-inc/gmail/gmail-2025-11...

Interesting timing, given that the PS4 has been hacked via various exploits, and the PS5 has just had its root encryption keys exposed.

Anna's archive has a great analysis of the Spotify data.

They identify a huge surge in tracks that few listen to after gen AI started.

The analysis is worth reading. The distribution is (Pareto)^3 ~99% of the tracks played are 1% of the catalogue.

> iAPX 432 Yes, this was a failure, the Itanium of the 1980's

I also regard ADA as a failure. I worked with it many years ago. ADA would take 30 minutes to compile a program. Turbo C++ compiled equivalent code in a few seconds.

Machines are thousands of times faster now, yet C++ compilation is still slow somehow (templates? optimization? disinterest in compiler/linker performance? who knows?) Saving grace is having tons of cores and memory for parallel builds. Linking is still slow, though.

Of course Pascal compilers (Turbo Pascal etc.) could be blazingly fast since Pascal was designed to be compiled in a single pass, but presumably linking was faster as well. I wonder how Delphi or current Pascal compilers compare? (Pascal also supports bounded strings and array bounds checks IIRC.)

> I wonder how Delphi or current Pascal compilers compare?

Just did a full build of our main Delphi application on my work laptop, sporting an Intel i7-1260P. It compiled and linked just shy of 1.9 million lines of code in 31 seconds. So, still quite fast.