Mythos Leaks, Token Billing, Claude Routines, and OpenAI's Big Week (Image Gen 2.0 + 5.5)

The hosts attempt a chaotic "lightning round" covering major AI and tech stories: Anthropic's Mythos tool was reportedly accessed despite being considered too dangerous to release, yet it helped find serious Firefox bugs; Microsoft is criticized for new monetization moves including token-based GitHub billing, licensing "agents," and deeper Anthropic/Claude integration that may raise private-tenant data concerns. They discuss Claude "routines" that trigger on events for faster sales and help-desk responses, alongside ongoing risks like spyware incidents, OAuth-token attacks, and a Vercel hack. OpenAI's Codex is described as operating a Mac via the GUI while sending screenshots to servers, and ChatGPT Image Gen 2.0 is praised for consistent edits, web-informed infographics, and flexible sizing up to 4K. They also note ChatGPT 5.5 improvements, gas-powered data centers and climate impacts, UI differences between Microsoft and Google's Gemini integrations, hype cycles, and IPO/valuation speculation.

Hashtag Trending would like to thank Meter for their support in bringing you this podcast. Meter delivers a complete networking stack, wired, wireless and cellular in one integrated solution that's built for performance and scale. You can find them at Meter.com/htt

00:00 Cold Open and Sponsor 00:18 Botched Group Intro 01:09 Lightning Round Kickoff 01:24 Anthropic Mythos Leak 04:56 Microsoft Billing Moves 08:13 Claude Routines for Business 11:08 Agent Security and OAuth 14:21 OpenAI Codex Screen Control 15:38 Models Love Tool Use 17:41 ChatGPT Image Gen 2.0 25:51 OpenAI 5.5 Upgrade 30:34 Custom Image Sizes and 4K 31:54 Gas Powered Data Centers 35:08 Fries and Fat Panic 36:34 Data Centers Pay Up 37:59 Fast Power Options 40:01 Copilot vs Gemini UX 47:26 OpenAI 5.5 Lightning 48:17 Synapse Game Demo 52:15 Cyber Model Fizzle 56:23 IPO Hype Machine 59:41 Too Big to Fail 01:09:54 Reality Check Wrap