WEEKLY DIGEST
Week in AI: Aug 17 — Aug 21, 2026
The top stories from our coverage this week.
How does Z.ai boost GLM-5.3 performance without retraining the 743B base model?
marktechpost.com/2026/08/14/z-ai-ships-glm-5-3-with…
AI-generated books make up 20% of Amazon's self-published catalog but account for 12% of sales, dragging down revenue per human-written title.
the-decoder.com/ai-generated-books-are-flooding-ama…
HumanTracker benchmark reduces motion tracking errors by 31% against human perception.
huggingface.co/papers/2608.13555
@AnthropicAI's Dario Amodei says AI centralizes power by nature, with open models just shifting control to chip owners.
the-decoder.com/anthropic-ceo-says-ai-centralizes-b…
Debate training reduces reward hacking in AI systems using LLM judges.
alignmentforum.org/posts/BB8o7b8A4Aykeksvw/debate-t…
Language model weights carry unique signatures to verify their lineage, not just their training data.
huggingface.co/papers/2608.14929
OpenAI dissolved its catastrophic AI risk team, reassigning the work to other groups.
the-decoder.com/openai-dissolved-the-team-built-to-…
Agility Robotics paths to public markets vary, with SPACs, reverse mergers, and traditional IPOs each offering distinct advantages.
therobotreport.com/robots-on-wall-street-non-tradit…
Memory-based AI agents improve over time but are fragile to task order and underspecification.
arxiv.org/abs/2608.18066v1
Domain data scarcity forces LLMs to repeat content, undermining token-per-parameter ratios.
huggingface.co/papers/2608.14071