MERZpoet: fine-tuning Mistral and Llama on your own writing as a mirror of the mind
Merzmensch · x · 2026-08-17
Merzmensch shares the MERZpoet project: fine-tuning Mistral and Llama on years of his own texts (GPT-2 proved incapable) to generate poetry in English and Japanese in his style. He frames it as a "distorting mirror of the mind" — distortion reassembles fragments of one's thinking in new constellations. Working with LLMs directly, without harnesses like ChatGPT or Claude, lets you dig deeper into their semantic base, which he likens to a Noosphere of the world's combined mindsets.
Related event: Developer Fine-Tunes Mistral and Llama to Mirror His Own Poetry Style(4 posts)→
More from Fun
- Two robot lamps meet for the first time and look like they are arguing — dee_hw · 2026-08-17
- Showcase: Hot in The City for AI Film Festival — DuHal9000 · 2026-08-17
- Personal Steampunk Film Created with Minimax at Home — jude1903 · 2026-08-17
- Let Opus 5 loose in Blender to render a wizard, got an abstract result — KKTeX_LaTeX3 · 2026-08-17
- Blogger TheZvi apologizes for mischaracterizing an RLVR worker as viewing AI as 'just tools' — TheZvi · 2026-08-17
- Hand-resolving Git conflicts in Neovim takes 40 seconds vs 15 minutes with tools — vboykis · 2026-08-17