nsbk

Safari routinely using 20+ Gb of memory with a handful of tabs open. Safari tabs refusing to close. Unresponsive System Settings window. Random application freezes and crashes, Apple Music not playing music. This is on a 32Gb M1 Max. My M1 Air on Sequoia doesn't experience any of these issues, even if it has half the unified memory.

The hardware looks amazing! Too bad they will ship with Tahoe installed. I’m not upgrading until I see in which direction the next Mac OS release goes

I’m right where you are. Very happy Apple customer since my first PowerBook G4. Currently have an M1 Max, an iPhone 17 Pro, the iPad Pro, HomePod, Apple TV, and Watch Ultra.

All the _just works_ feeling and reliability seem to be gone. Tahoe is so unstable that I now restart the Mac every day, when in the past it happened on software updates only. Apple Music is another huge mess, I can’t comprehend how can it be so unreliable.

Looking for exits as well and kind of looking forward to migrating to Graphene OS, self-hosted Immich, and Navidrome

Another alternative (although Mac OS-only) is [0] OrbStack. Some devs in my team are running it as a more performant alternative to Docker Desktop for Mac and they are very happy so far.

[0]: https://orbstack.dev

Since it is all about money, I just did vote with my wallet and cancelled the Max subscription

One of the many use cases, but basically yes. Other use cases: Home automation, remote backups, media servers, photo libraries, AI assistants... you name it!

At this point Tailscale is working so well and I'm so happy with it that I'm afraid it's time to start migrating to Headscale [0] for my home network. The rag pull may just be too painful otherwise!

[0]: https://headscale.net/

The US averages 23 pedestrian deaths per million people per year. The EU averages 8. The US fatalities have increased by 50% since 2013, while in the EU have decreased by 25% in the same time frame.

https://www.cdc.gov/mmwr/volumes/74/wr/mm7408a2.htm

There is no lizard brain. The "triune brain" theory has been debunked by modern neuroscience for years.

I got the models all the way around. Nemotron-speech ASR is the one that is English-only. Magpie TTS is multilingual and can do both English and Spanish

Make sure you’re using the nemotron-speech asr model. I added support for Spanish via Canary models but these have like 10x the latency: 160ms on nemotron-speech vs 1.5s canary.

For the LLM I’m currently using Mistral-Small-3.2-24B-Instruct instead of Nemotron 3 and it works well for my use case

I'm putting together a streaming ASR + LLM + streaming TTS setup based on Nvidia speech models: nemotron ASR and magpie TTS, pipecat to glue everything together, plus an LLM of your choice. I added Spanish support using canary models, as magpie models are English-only and it still works really well.

The work is based on a repo by pipecat that I forked and modified to be more comfortable to run (docker compose for the server and client), added Spanish support via canary models, and added Nvidia Ampere support so it can run on my 3090.

The use case is a conversation partner for my gf who is learning Spanish, and it works incredibly well. For LLM I settled with Mistral-Small-3.2-24B-Instruct-2506-Q4_K_S.gguf

https://github.com/nsbk/nemotron-january-2026

I actually forked the repo, modified the Dockerfile and build/run scripts targeting Ampere and the whole setup is running seamlessly on my 3090, Magpie is running fine and using under 3Gb of memory, ~2Gb for nemotron STT, and ~18Gb for Nemotron Nano 30b. Latencies are great and the turn detection works really well!

I'm going to use this setup as the base for a language learning App for my gf :)

Any ideas on how to add Ampere support? I have a use case in mind that I would love to try on my 3090 rig

They definitely do change things and not always for the better. I used to start running/walking workouts from the Apple Watch telling Siri something along the lines of "let's get running!" or "let's go for a walk" to which it would react starting the right kind of workout. Since WatchOS 26 it tells me it doesn't know where I am and something about privacy settings, and that's it. Asking in robot mode it does work though: "start running workout"

My thoughts exactly! Although I may end up getting some Mini M1/M2 variant with Asahi Linux instead

I’ve always wanted a personal eclipse orb and now I can buy one. Amazing. https://anycrap.shop/product/personal-eclipse-orb

I switched to Jellyfin last year and never looked back. The only thing I find lacking is the Apple TV App, I tried Swiftfin but it stutters the whole time when playing high quality UHD content. I tried Infuse and it works much better

It is the main database for a huge Rails app. They adopted Mongo right when its popularity started to decline. I always thought it was a very poor choice since the day I joined.

It is a especially bad choice considering that a lot of the data stored in it is IoT-like and the system creates a single document per event :facepalm:

Some of the insights match my personal experience and preferences. At $dayjob we're migrating from Mongo to TimescaleDB (now TigerData ¯\_(ツ)_/¯) which is basically a PostgreSQL extension for time series data and couldn't be happier. We are getting better performance and massive storage savings.

On the analytics side of things we are starting to use DuckDB for some development efforts, but we are keen on potentially replacing some or all of our Snowflake usage with DuckDB.

Interesting timing, they have recently reached out to my $dayjob. We will be probably be running a workshop on our (massive) dataset with them. I'd like to evaluate the performance of a couple of analytical models we've manually built against whatever this model can do based on some prompts. Exciting times!

I thought it was tequila

I'm short on time and couldn't read the whole article, but apparently metabolic health is a huge factor that doesn't seem to be pointed at here. Obese parents, both mum and dad, multiply the likelihood of child autism by something like 4x. It could be that the obesity epidemic is the cause of the rising autism rates. Same goes for other metabolic health impacting factors such as diabetes, hypertension, and the likes.

Both sound great, I’ll probably read both and decide which series to cover first. Thanks!

That sounds like a lot of fun indeed! Thanks for the advice

What entry point would you recommend to a total newcomer to the series?

Nice! I drive a 2005 Audi with all physical controls. I intend to keep it for as long as it keeps running, so if the trend continues and more manufacturers join it I may be able to skip the whole huge touchscreen madness altogether

Or 42 in Cuttlefishish

I’d love to buy that config as my personal laptop, but the problem is that my 512/16 M1 Air still works so well for my use case that I can’t find enough reasons to justify the expense. M6 Air maybe!

It is indeed! I used to connect to a Synology NAS as remote backup location through Tailscale and the transfers would peak at around 5 MB/s.

Now I mount the NAS volumes to a host at the same location and sftp to it. It’s still dog slow at 30MB/s but that’s the NAS limitation.

Direct access to the NAS can also be achieved via Subnet Routers

https://tailscale.com/kb/1019/subnets