Lilian Weng
AI Safety · Post-training · LLM InternalsLilian Weng's long-form technical blog — the rare writing from inside an OpenAI safety leader, covering post-training, reward hacking, alignment, and LLM internals at depth most public writing skips.
About this vault
Lilian Weng leads Applied Research at OpenAI and runs one of the deepest public blogs in AI. This vault curates 10 canonical pieces across four arcs: (1) LLM fundamentals — attention, RLHF, prompt engineering as a discipline; (2) post-training and alignment — reward hacking, specification gaming, the nuances of making helpful models; (3) agent architectures — planning, memory, tool use from a builder's perspective; (4) safety frontiers — extrinsic hallucinations, reasoning evaluations, and the failure modes no one else writes about. Her blog is how researchers without internal OpenAI access understand what the safety team is actually thinking. Canonical reading for anyone building on top of modern LLMs.
10 articles
All 10 articles
Why We Think
Reward Hacking in Reinforcement Learning
Extrinsic Hallucinations in LLMs
Diffusion Models for Video Generation
Thinking about High-Quality Human Data
Adversarial Attacks on LLMs
LLM Powered Autonomous Agents
Prompt Engineering
The Transformer Family Version 2.0
Large Transformer Model Inference Optimization
Start reading, not hoarding.
Import this vault to Burn 451 and actually read what matters.
Frequently asked questions
Who is Lilian Weng?
Lilian Weng is covered in this Burn 451 vault with a focus on ai safety · post-training · llm internals. Lilian Weng's long-form technical blog — the rare writing from inside an OpenAI safety leader, covering post-training, reward hacking, alignment, and LLM internals at depth most public writing skips.
How was the Lilian Weng vault curated?
The Lilian Weng vault was hand-curated by the Burn 451 editorial team from publicly available essays, blog posts, podcast transcripts, and social threads. Each piece includes an AI-generated summary so readers can triage in seconds. The vault auto-syncs as new content from Lilian Weng is published.
How many articles are in the Lilian Weng vault?
The Lilian Weng vault currently contains 10 curated pieces organized by topic, not chronology. Each article has an AI summary and a direct link to the original source. Items are refreshed hourly through Burn 451's ISR pipeline, so new publications appear within a day.
How do I use this vault with Claude or Cursor?
Install the burn-mcp-server package from npm and connect it to Claude, Cursor, or any MCP-compatible AI tool. The vault becomes queryable as live context — your AI can search, summarize, and cite articles from Lilian Weng directly in conversation without manual copy-paste or re-uploading files.
What is Burn 451?
Burn 451 is a read-later app built around a 24-hour burn timer that forces daily triage. Articles you save must be read, vaulted, or released within 24 hours. The Vault layer — including this Lilian Weng collection — holds permanent curated reading lists for AI thought leaders, founders, and researchers.
Content attributed to original authors. Burn 451 curates publicly available writing as a reading index. For removal requests, contact @hawking520.