True Positive Weekly #176: chatbot trust, recursive self-improvement, vLLM speculative decoding
burkov · x · 2026-09-04
A new issue of burkov's weekly AI newsletter rounds up notable reads and projects:
- A psychologist explains why we trust chatbots and how to use them more wisely
- AI speeds up homework but correlates with poorer exam results — what it's doing to student learning
- A look at recursive self-improvement
- MIT study: generated images often can't be traced to any training data ("AI art has no author")
- Bayes alone doesn't make experiments faster
- Explainers on speculative decoding in vLLM on AMD GPUs and building a sandboxed coding agent from a raw shell
- A project on teaching an AI your taste — and verifying it learned judgment
More from Research
- SPACE cuts agent LLM calls by 78.9% while raising success rate on long-horizon tasks — dair_ai · 2026-09-04
- CoRL 2026 hosts first workshop on imperfect robotics data: failures, OOD, human surprises — RobobertoMM · 2026-09-04
- Baseten launches Base Labs, a research org for open-source AI with fully shared recipes — baseten · 2026-09-04
- Study: AI companions rival human friendship, users mourn forced separations — EricTopol · 2026-09-04
- MazeBench: New 3D Spatial Reasoning Benchmark Where Prior SOTA Agents Score Just 1% — patience_cave · 2026-09-04
- "World Models from Scratch": a hands-on open-source book launches its first release — Cohere_Labs · 2026-09-04