Summary
Matt Wolfe’s weekly AI news roundup centers on Sakana AI’s launch of Fugu and Fugu Ultra, a novel orchestrator model that routes prompts across multiple AI providers — including OpenAI and Anthropic — selecting the best underlying model for each task. Unlike traditional single-provider models, Sakana Fugu acts as an intelligent manager with built-in redundancy: if one provider goes down, it silently reroutes without losing progress.
Wolfe provides hands-on testing via the Codex CLI, building a Mega Bonk game clone and a Future Tools website clone using Fugu Ultra on its highest setting. The two projects consumed 22 million input tokens and 210,000 output tokens at a combined cost of $30 — a practical caution for users tempted to run Ultra at maximum from the start. Benchmark comparisons show Fugu Ultra matching or beating Fable 5 on LiveCodeBench and Google Proof Questions, with competitive results on Scode, though trailing on SWEBench Pro.
The roundup also covers Anthropic’s Claude Tag feature for Slack, which lets teams @-mention Claude directly in channels where it can take on multi-step projects, retain context over time, and proactively surface help when relevant. A sponsored segment introduces Nexos, an AI productivity dashboard aggregating ChatGPT, Claude, and Gemini access with a no-code agent builder. The episode is a solid weekly briefing for anyone tracking new model capabilities and enterprise AI integrations.
📺 Source: Matt Wolfe · Published June 26, 2026
🏷️ Format: News Analysis
