2026-01-07
As prediction markets scale, reliable and credibly neutral resolution mechanisms will become critical. The evolution of social media content moderation is a very tight analogue for this and points to four likely developments: (1) The development of internal expertise in the
Financial Times
Polymarket is disputing that the US mission to capture Nicolás Maduro constituted an “invasion”, refusing to pay out bets on a contract with $10.5M in wagers
Prediction market disputes US raid amounted to an invasion in fight over more than $10.5mn in wagers
2025-12-02
This is a super interesting and deep document from Anthropic detailing Claude's values and charge. You can see some conceptual stretching going on here where “safe” is being recast to justify reducing refusals because it would be “unsafe” to be “unhelpful” to users. This seems [image]
Simon Willison's Weblog
A Claude user gets Claude 4.5 Opus to generate a 14K-token document that Claude calls its “Soul overview”; an Anthropic employee confirms the doc's validity
This appeared to be a document that, rather than being added to the system prompt, was instead used to train the personality of the model during the training run.
2025-11-14
Asking an AI to grade itself for political bias is not the right way to assess political bias.
Axios
Anthropic open sources a method to score AI model political evenhandedness; Gemini 2.5 Pro got 97%, Grok 4 96%, Claude Opus 4.1 95%, GPT-5 89%, and Llama 4 66%
Ina Fried / Axios :