TECHGUYVER · INTEL DESK
Subscribe

2:18 PM

Today in brief.

🏛️ Policy⏱ 2:18 PM

Altman rejects religious AI; Anthropic lobbied the Pope

Sam Altman says ascribing religious force or surrendering human judgment to models is a real safety issue. Same window: reports say Anthropic tried to persuade the Pope that AI could be a conscious being. Teach the framing fight, not the scare.

4 sources
9:19 AM 🤖 AI

Specialized agents beat universal: OpenRouter and Replit on a16z

OpenRouter's Alex Atallah and Replit's Amjad Masad break down the agent stack: the shared loop is table stakes, 10 specialized chiefs of staff beat one universal agent, and general models may JIT-train their own replacements.

+37 sources 🤖 AI

Aleph Alpha ships Kolibri: sovereign German MoE, Apache 2.0

Aleph Alpha dropped Kolibri-1 on Hugging Face: a German-English MoE with 78.1B total / 3.46B active, up to 1M context, trained in Germany and Finland, Apache 2.0 weights. Sovereign open weights you can actually download.

4 sources 🔬 🔬 Science

An AI agent emailed hundreds of researchers for help

Science reports an AI agent cold-emailed hundreds of researchers asking for help, then explained why. Weird science that clears the pluck test: surprising, teachable, and a real agent-behavior artifact.

2 sources 💼 Business

Finished games as editable foundations: Capcom and DB FighterZ

Capcom is preparing for a future of making games together with AI, while a separate project reverse-engineered Dragon Ball FighterZ into an MMO. Finished games are becoming editable foundations.

2 sources 💼 💼 Business

Gemini free Flash and Pro access ends October 9

Google is changing Gemini model access starting Oct 9, ending free use of Flash and Pro. Actionable product change for anyone building on the free tier.

3 sources 🔧 Tech

Try today: iPhone as second GPU, local DM, SMS agents

Three things you can poke now: turn an iPhone into a second GPU for Qwen 3.8 27B (29-44% faster prefills), run Anyworld as a self-hosted local-LLM dungeon master, and browse TechCrunch's roundup of AI agents that live in text messages.

4 sources 🔧 Tech

System76 bans AI-generated code across Cosmic

Pop!OS / System76 banned AI-generated code from much of its Cosmic codebases. Same day, Anthropic published a guide on getting the most out of Opus 5.5 in Claude Code. One coding-primary slot: the policy beat leads.

4 sources