Anthropic released Claude Opus 4.8 on May 28, claiming it outperforms Google’s Gemini 3.1 Pro and OpenAI’s GPT-5.5 in agentic coding, financial analysis and autonomous tasks. A Google staff engineer allegedly used confidential Year in Search data to place $2.75 million in Polymarket bets, netting $1.2 million in illegal gains.
On May 28 Anthropic introduced its flagship Claude Opus 4.8, an upgrade over Opus 4.7 that it says surpasses OpenAI’s GPT-5.5 and Google’s Gemini 3.1 Pro across synthetic tests such as agentic coding, financial analysis and autonomous computer tasks. The model also adds honesty features to flag uncertainties and reduce unsupported claims while dynamic workflows allow orchestration of hundreds of subagents.
Federal prosecutors allege a Google staff software engineer used internal Year in Search data to place roughly $2.75 million in Polymarket bets from October 15 to December 4, winning about $1.2 million. The engineer is charged with commodities fraud, wire fraud and money laundering and has been placed on leave as the company evaluates further action.
Google is providing a financial backstop for a 15-year, $7 billion triple-net lease underpinning Hut 8’s $3.25 billion financing of the River Bend AI data center in Louisiana. The arrangement features a 3% annual rent escalator and renewal options that could raise total contracted value to $17.7 billion, ensuring stable cash flows without future refinancing risk.
Finance
