Top all-time · #17698
castform
Company blog · castform.com/blogaiprogramming
castform is the company blog at castform.com, writing about ai and programming. Since 2026, 4 different Hacker News readers have linked it 12 times, most often “I RL-finetuned an LLM to unslop my writing”. No feed was found, so read it on the site itself.
fine-tune and rl-train open models on your own data. frontier performance at a fraction of the cost.
Most-linked posts?
- I RL-finetuned an LLM to unslop my writing2 people · 2 mentions · first linked 2026-08-04
What Hacker News says?
castform founder here. while it's "rare for a specialized model to beat a strong general model" today, i think the tech/knowhow on how to do so is getting there. we see some early signs of this in industry e.g. https://shopify.engineering/sidekicks-continual-learning-loo... https://thinkingmachines.ai/news/learning-to-replicate-exper... https://castform.com/blog/elsa/ at the end of the day, models are only as good a…
Comment linking /blog/elsa/
on “Beating GPT-5.6 Sol on retrieval with 100x cheaper open models”Submitted Pokegents: Making multi-agent coding feel like a team as a story · 6 points
HN discussionSubmitted Prompt caching but for RL – 7.5x speedup on long-prompt/short-response workloads as a story · 4 points
HN discussionSubmitted Post-training to build better Roblox UIs as a story · 3 points
HN discussionSubmitted Automated red teaming with RL: attacker-defender co-training as a story · 3 points
HN discussionSubmitted Token-in-token-out RL training with any agent harness, via a proxy gateway as a story · 2 points
HN discussionSubmitted Castform now supports supervised finetuning as a story · 1 points
HN discussionSubmitted I fought the slop and I won (unslopping LLM output) as a story · 1 points
HN discussionSubmitted I RL-finetuned an LLM to unslop my writing as a story · 1 points
HN discussionSubmitted how we monitor our rl training runs as a story · 1 points
HN discussionSubmitted Designing dev onboarding for an agent-first world as a story · 1 points
HN discussionSubmitted I post-trained a model to reliably roll a die as a story · 1 points
HN discussion