2025 LLM Year in Review | karpathy
karpathy.bearblog.dev · 1,775 words · saved by 16 readers
2025 Year in Review of LLM paradigm changes
2025 LLM Year in Review – karpathy 2025 LLM Year in Review 19 Dec, 2025 2025 has been a strong and eventful year of progress in LLMs. The following is a list of personally notable and mildly surprising "paradigm changes" - things that altered the landscape and stood out to me conceptually. 1. Reinforcement Learning from Verifiable Rewards (RLVR) At the start of 2025, the LLM production stack in all labs looked something like this: Pretraining (GPT-2/3 of ~2020) Supervised Finetuning (InstructGPT ~2022) and Reinforcement Learning from Human Feedback (RLHF ~2022) This was the stable and proven r
saved by
- Shagun Maheshwari
- Trang Đoàn
- Elizabeth Qiu
- Art M.
- Zhi Ying Sun
- Nhật Nguyễn Thế
- Dex Ter
- Yudhister Joel Kumar
- Liam Hinzman
- Kevin Xu
- Tony Kam
- Akira Yoshiyama
related reading
- 2025: The year in LLMssimonwillison.net
- AI in 2025: gestalt — LessWronglesswrong.com
- Guardian Angels: LLM Personalization for Productivity and Security · Gwern.netgwern.net
- Verifiabilitykarpathy.bearblog.dev
- What We’ve Learned From A Year of Building with LLMs – Applied LLMsapplied-llms.org
- Andrej Karpathy — AGI is still a decade awaydwarkesh.com
- As Rocks May Think | Eric Jangevjang.com
- Things we learned about LLMs in 2024simonwillison.net
- GenAI Handbookgenai-handbook.github.io
- The State Of LLMs 2025: Progress, Progress, and Predictionsmagazine.sebastianraschka.com
- State of AI 2025: 100T Token LLM Usage Study | OpenRouteropenrouter.ai
- The bitter lesson of LLM evalsparsed.com