What's this about?
Ollama — the most popular tool for running open models locally with one command — published a blog post celebrating the open-model wave (Kimi K3, Qwen, Gemma…) and confirming an $88M funding round, alongside its $20/month cloud subscription for models too big to run at home.
What exactly is Ollama?
Ollama is an app that turns "download and run an AI model on your own computer" from an afternoon of setup into a single command (`ollama run gemma`). It wraps the open-source inference engine llama.cpp with easy packaging, a model library, and an API that other apps plug into. That convenience made it the default on-ramp for local AI — and made it controversial, because the hard part it wraps is someone else's project.
What's the community saying?
HN 108 points, 46 comments — unusually hostile for a funding post
- "Stop using Ollama" — the top comment is just a link to an essay by that name (12 replies); the llama.cpp-attribution grievance resurfaces: rebranding upstream features as Ollama's own.
- Quality, not just ethics — critics say Ollama is slower than plain llama.cpp and its default quantized models are noticeably worse than community ones (e.g. unsloth's).
- VC money = enshittification clock — "note to self: don't use any Ollama service unless it's calling an industry-standard API."
- The defense — it genuinely is the easiest way to plug open models into agents and assistants, and the $20 subscription is generous with a no-logging pledge; "how did they raise that much if they're so hated?"