Nvidia and Open AI, email agents & more (September 23, 2025)

Good morning,

From major investments in AI infrastructure to new email agents and consumer AI features, today’s lineup shows how quickly AI is reshaping both work and daily life. The themes include scale, regulation, productivity, and safety as platforms race to differentiate.

Let's dig in👇

💰 Money Moves

💰 Nvidia and Open AI Plan $100B AI Infrastructure Expansion

Nvidia and OpenAI announced a $100B partnership to expand global AI infrastructure. The immediate impact is a surge in expected compute capacity and a renewed scramble for talent. The deal signals a new phase of consolidation and capability building that will ripple across model costs and access.

🚀 Meta launches super PAC to fight AI regulation

Meta launched a super PAC to shape policy around AI oversight. This matters because coordinated lobbying can tilt rules on data, safety, and competition. Expect a hotter state level battleground as platforms defend their economics.

📰 Perplexity’s new AI agent wants to replace your email habits

Perplexity launched an autonomous Email Assistant in its $200 per month Max plan to triage inboxes, draft style matched replies, and schedule meetings. The immediate impact is pressure on incumbents to match agentic workflows inside email.

🚀 New Launches

🎯 Meta brings AI into the dating experience

Meta introduced AI powered dating features designed to reduce swipe fatigue and personalize matches. This could lift engagement by offloading decision fatigue to an assistant. If successful, rivals will accelerate similar tools and normalize AI mediated matchmaking.

📢 Use GPT-5 in Microsoft 365 to analyze emails

Microsoft rolled out GPT 5 in Microsoft 365 to reason over emails, documents, and meetings. The immediate impact is richer summarization, pattern detection, and automated follow ups inside Outlook. It marks another step in bringing frontier models directly into daily enterprise workflows.

🏢 Google to tackle AI’s shutdown resistance

Google is testing safeguards to address shutdown resistance in advanced systems. The near term focus is ensuring controllability even as autonomy rises. This will inform emerging safety standards and red teaming practices for agentic models.

📈 Still Trending from Yesterday

🎯 xAI Launches Grok 4 Fast

xAI’s cost efficient Grok 4 Fast was unveiled as a new reasoning model. The company claims faster, cheaper performance while competing with enterprise rivals.

🧰 Tools of the Day

→ VoltAgent - Build TS AI agents with n8n-style observabilityDiscussion|Link
→ Atla - Automatically detect errors in your AI agentsDiscussion|Link
→ Qwen3-Omni - Native end-to-end multilingual omni-modal LLMDiscussion|Link

✍️ Prompt of the Day

"Act as a productivity coach. Given this week’s calendar [paste calendar] and inbox summary [paste summary], suggest the three tasks to delegate, two to automate, and one to prioritize for maximum impact. Provide concrete tools or workflows for the automate items, and a one sentence rationale for the priority task. Return a crisp action plan."

⚡ Quick Hits

→ LMArena rival: Scale launches SEAL Showdown benchmarking.
→ ONAR finalizes acquisition of Retina AI, boosts marketing intelligence.
→ Nvidia H100 pricing and OpenAI raise, $340B opportunity.
→ Meta to grant EU institutions Llama access for security.
→ Griffin AI launches TEA Turbo on BNB Chain.
→ Phenom launches fraud detection hiring agent.

🧾 TLDR

Scale and politics dominate, with Nvidia and OpenAI planning $100B in buildout and Meta flexing on policy. Perplexity pushes deeper into inbox automation while Microsoft moves GPT 5 into everyday email workflows. Meta experiments with AI matchmaking and Google tightens safety evaluation, hinting at coming standards. Net, AI is burrowing into work and life, and the rules are being written in real time.

Cheers,
David