0xc133

Well, I know what I'm upgrading tonight. (I didn't see any mention of this in the blog post, but they also implemented batch-upload for the vandelay tool, so exporting your archive into a new Stalwart server is much faster now, too!)

Anthropic released Claude Opus 4.6 on February 5, 2026. What year are you in, man?

I bought a Sonos soundbar (the Ray) to dip a toe in the waters, so to speak.

I’ll never buy another Sonos product ever again. Terrible experience all around. App is trash, soundbar drops off wifi for no reason, songs will stop playing over Airplay and it’ll just disappear as an Airplay speaker for hours and then show back up randomly and work fine again.

Why the hell would you ever let a TV on any network?

The cheapest Steam Deck available now is $789.

How were you accessing the model? I was using ollama.com’s cloud hosting the first few times I ran into it, and then it happened again the first time I tried Fireworks.ai before I made it even 10% of the way through the context window.

deepseek-v4-flash-0731 has been awfully prone to infinite looping output in reasoning for me, and once it hallucinated in the middle of going in circles that I had instructed it to start using Yoda-speak, which… I don’t have any idea where that came from.

*Forgejo

asciinema is so good! Love seeing the player on the site.

I was able to use rope/yarn scaling with llama.cpp to extend the context window to 256K and it seems to be pretty usable on the debugging-and-bugfixing session I have that’s 216K tokens deep. No infinite output loops, reasoning is still coherent, tool calls appear to be passing and failing at roughly the same rate as a fresh context window. Haven’t tried going to 512K or higher yet but 256K definitely seems usable to me on a single Radeon R9700.

I added these arguments to my llama-server call, using the official GGUF release from Meta’s account on HF.

`--rope-scaling yarn --rope-scale 2 --yarn-orig-ctx 131072 --override-kv muse-glimmer.context_length=int:262144,dflash.context_length=int:262144`

because once they run out of tokens they can’t do their jobs anymore

I discovered that barely-used enterprise Intel SSDs (DC S3610, S3700, S3710) were cheap on ebay a year or so ago and I may have bought a few dozen. These things have remaining endurance measured in petabytes. Sure, it’s overkill to have a 200GB enterprise SATA disk as the boot volume for a Raspberry Pi, but this is exactly my kind of overkill.

I also have become a maki convert and I really like it. I ran into an issue with the dynamic model provider that I should probably make a patch for; list_models doesn’t use the `<provider> models` output at all but instead tries to look up `<provider> resolve`’s base URL + /v1/models, which breaks on a provider like Z.ai which doesn’t have /v1/ anywhere in the path…