nolist_policy

It's not Googles fault. It's our fault, llama.cpp doesn't support it, vllm can't really offload to host ram and while huggingface transformers supports it, it only can do f16 quant for MoEs like Diffusion Gemma. Unusable for low spec machines where diffusion really could speed up things.

We're sleeping on a amazing model right in front of our eyes. With diffusion models you can properly ban phrases. You can decensor and deslopiffy the model output without touching a single weight or doing a single backwards pass.

Checking the bus factor:

Btrfs regulars:

- 1 from Meta

- 1 from Oracle

- 4 from SuSe

- 2 from WDC

bcachefs:

- Kent Overstreet

Not true, Boris Burkov from Meta works on btrfs. And two people from WDC.

Only if you have quotas enabled IIRC.

> Sync, rotating snapshots, and deduplicated backups need to be kept entirely separate if you want any hope of maintaining your sanity.

Not if you use git-annex.

Yes, The two couldn't be more different.

GPT-3 is probably one of the last models trained without synthetic data. Meanwhile gpt-oss is rumored to be only trained on synthetic data.

I also miss the 2020 GPT-3 based AI Dungeon fine tune.

Then take DeepSeek V4 flash with AA score 52. Runs unquantized on 2x DGX spark with 1M context.

I don't get why they choose a tiling window manager.

I think you can also just use the random number generator (seeded with a secret key) as the watermark. Then the probability distribution is exactly the same.

Omarchy is not a product though.