Sebastian Raschka's from-scratch video course on building reasoning models
- ID
- 39665
- Status
- summarized
- Published
- 26 Oct 2026, 11:11 PM
- Fetched
- 27 Oct 2026, 2:51 AM
- Provider
- Hacker News
- Category
- dev-community
- Original URL
- https://magazine.sebastianraschka.com/p/building-reasoning-models
- Source URL
- https://hnrss.org/best
Summary
- Score
- 8.0
- Created
- 27 Oct 2026, 2:51 AM
- Tags
- Audience
- developersai_ml_learnersai_agent_users
What happened
Sebastian Raschka released an eight-video, 11-hour course that builds reasoning models from a pretrained LLM in Python and PyTorch, covering evaluation, inference-time scaling, reinforcement learning with verifiable rewards, and distillation. It is a standalone sequel to his Coding LLMs from the Ground Up course. The Hacker News thread drew 309 points and 41 comments.
Why it matters
If you build or use agents, the course's engine-vs-harness framing is a concrete way to decide where to spend effort: improving the reasoning model with RLVR/distillation versus improving context management and tool access in the harness. The 11 hours of PyTorch code-along let you test inference-time scaling and distillation directly instead of treating reasoning as a black-box API mode. There is no Malaysia-specific policy, funding, or infrastructure hook in this item.
Discussion angle
Where does your team draw the line between the reasoning model and the agent harness, and after seeing this course, would you rather implement RLVR/distillation yourself or keep buying reasoning modes through an API?