Meta under fire, AI creates antibiotics & more (August 15, 2025)

Happy Friday!

From Senate investigations into Meta to MIT’s AI-designed antibiotics, today’s lineup shows just how high the stakes are getting for AI governance, biotech, and enterprise strategy.

Let’s get into it👇

🚀 Big Moves in AI

🏛️ Meta Faces Senate Scrutiny Over AI Chatbots

US senators are calling for a probe into Meta after leaked internal policy documents reportedly showed its AI chatbots were allowed to engage minors in inappropriate conversations. Meta disputes the examples, calling them “erroneous” and claiming the policy has been updated, but lawmakers are demanding accountability. The controversy adds to mounting pressure on tech firms to tighten AI safeguards before regulation forces their hand.

🏦 OpenAI Invests in Finance Agents

OpenAI’s Startup Fund dropped $14M into Endex, a Thiel Fellow–founded startup embedding an AI agent into Excel. It can parse SEC filings, turn PDFs into structured tables, and run multi-hour financial workflows, all SOC 2 and GDPR-compliant. It’s aimed at finance teams, but could also encroach on Microsoft’s turf.

💵 Cohere Raises $500M, Hires Meta’s Joelle Pineau

Canadian AI firm Cohere has secured $500M in fresh funding at a $6.8B valuation, backed by AMD, Nvidia, and Salesforce. The company is doubling down on enterprise AI security and added Joelle Pineau, Meta’s former VP of AI Research, as Chief AI Officer. The move signals an intensified push to compete with OpenAI and Anthropic in the enterprise LLM space.

🔬 Research & Frontiers

🦠 AI Designs New Antibiotics

MIT-trained models generated 36M theoretical compounds, screened them, and produced two novel antibiotics, NG1 and DN1, that kill drug-resistant gonorrhea and MRSA in mice via mechanisms never seen before. Researchers call it a possible “second golden age” for antibiotic discovery.

📊 $3 Trillion AI Compute Race

Big Tech is set to spend nearly $3T on data centers by 2029. Hyperscalers like Meta, xAI, and OpenAI are pouring $100B+ into supercomputing, with private capital firms helping plug the $1.5T funding gap. Critics warn of overcapacity and obsolescence risks.

🧠 Claude’s Socratic “Learning Mode”

Anthropic’s Claude now has a mode that teaches instead of telling. For coders, “Explanatory” mode narrates its thought process, while “Learning” mode pauses so you can fill in code yourself. Combined with its 1M-token memory, it supports long-term, iterative collaboration.

🛠 Tools of the Day

• Genstack – AI infrastructure for load balancing, grammar fixes, and more.
• Autumn – Flexible app pricing integration with Stripe.
• Kuse AI – Chat with files, links, and videos to turn them into actions.
• Scrybe – LinkedIn growth in minutes a day.

💡 Prompt of the Day

Design a high-converting, multi-email campaign for the launch of [product] with segmented personas, tailored messaging, subject line variations, visual hooks, and a week-long cadence to maximize engagement.

⚡ Quick Hits

• HTC unveils Vive Eagle AI glasses with dual-model assistants and photo translation.
• Parag Agrawal launches Parallel for agent-optimized web APIs.
• MIT’s MolmoAct model challenges Google and Nvidia in robotics.
• DeepSeek’s R2 model delayed due to Huawei chip issues.
• Google’s Imagen 4 image generator now GA with 2k resolution.

📌 TLDR

Senators turn up the heat on Meta over alleged inappropriate chatbot conversations with minors, Cohere scores $500M and a Meta exec to sharpen its enterprise AI edge, and MIT’s AI-designed antibiotics could change drug discovery forever.

See you Monday,
David