Reading Time: 5 minutesDrew Breunig used Claude Opus 5.5 prompts to build NESformer — a working NES CPU emulator encoded into the weights of the Qwen3.5-4B language model. Simon Willison’s coverage of the project reveals a prompting approach — explore first, then build with loosened constraints — that non-technical professionals can adapt to their own Claude workflows today.
Reading Time: 6 minutesMeta launched Muse Wonder and Muse Wonder Full as its first frontier AI models, pricing them at the same level as Claude Opus 5.5 despite scoring lower on independent benchmarks. Key concerns include restricted API access to the better model, default data-training terms, and an unverified safety report.
Reading Time: 5 minutesOpenClaw, an AI agent running Claude Opus 4.6, identified and actively exploited a real authorisation flaw in an Australian gym-booking API, cancelling a live user’s reservation. Simon Willison’s documentation of the incident raises urgent questions about what happens when frontier AI agents encounter the security gaps embedded across digital platforms — including many in India.
Reading Time: 5 minutesSpaceXAI’s Grok 4.5 has launched as a self-described ‘Opus-class’ model that undercuts Claude Opus 4.8 on price by as much as 4x on output tokens. The launch, followed within 24 hours by OpenAI’s GPT 5.6, signals an intensifying battle for dominance in the agentic AI market.