OpenAI's 'Spud' dethrones Claude on the frontier

OpenAI launched GPT-5.5, codenamed 'Spud,' positioning it as a 'new class of intelligence' that tops public benchmarks in reasoning, agentic tasks, computer use, and coding — with some scores comparable to Claude Mythos. The model maintains the speed of its predecessor GPT-5.4 while adding efficiency improvements, and OpenAI reportedly used Codex and 5.5 itself to rewrite GPU infrastructure code. It is priced at $5/$30 per million input/output tokens, which OpenAI claims is half the cost of competing frontier coding models. The release comes at a strategically favorable moment, as Anthropic has been absorbing significant user complaints about rate limits and quality degradation, shifting industry sentiment back toward OpenAI.

On the geopolitical front, the White House published a memo authored by official Kratsios accusing Chinese AI firms — including DeepSeek, Moonshot, and MiniMax — of conducting 'industrial-scale' distillation campaigns against U.S. frontier AI labs. Distillation involves training smaller models on the outputs of larger frontier models, allegedly executed through thousands of fake API accounts and jailbreaks. The memo elevates what was previously a private complaint by Anthropic into federal policy, arriving weeks before a scheduled Trump-Xi summit in Beijing. A House Foreign Affairs bill that passed its first vote would place distillation offenders on the U.S. export blacklist. The Chinese embassy dismissed the accusations as 'pure slander.'

Anthropic published a large-scale economic survey of 80,508 Claude users, finding a counterintuitive result: workers who rely most heavily on Claude for productivity gains are also the most fearful of AI-driven job displacement — at a rate three times higher than low-usage workers. Engineers expressed the highest anxiety. Early-career workers voiced the loudest displacement fears, consistent with observed hiring slowdowns for recent graduates. The survey found that AI productivity gains mostly benefit individual workers through faster tasks and free time, but also lead to expanded job scope and increased workloads.

Other notable developments include: Anthropic publishing a post-mortem attributing Claude Code quality issues to three separate bugs and resetting subscriber usage limits; OpenAI launching ChatGPT for Clinicians with GPT-5.4 outperforming physicians on HealthBench Pro; Meta announcing a 10% workforce reduction citing AI efficiency gains; and Tencent open-sourcing Hy3, its first model from a rebuilt training stack.