Claude Becomes the First Model on AWS, GCP & Azure and Its Secret “Soul” Comes to Light

Your quick, no-BS briefing on the biggest developments in AI — product updates, infrastructure moves, safety news, and the breakthroughs shaping the future. Each episode cuts through the noise so you know exactly what matters and why.

In today’s episode we cover:

🔥 Perplexity Adds Long-Term Memory Across All Models
Perplexity now remembers your context across conversations even when you switch models, performing automatic context engineering behind the scenes. A big move toward persistent, multi-model assistants.
Source: https://www.perplexity.ai/hub/blog/introdu...

🧠 Hidden “Soul Document” Discovered Inside Claude 4.5 Opus
Researcher Richard Weiss uncovered an embedded identity/value document describing Claude’s ethics hierarchy, alignment principles, and operator-vs-user conflict rules. It reinforces Anthropic’s safety-first philosophy.
Source: https://gist.github.com/Richard-Weiss/efe1...

⚠️ Anthropic Disrupted an AI-Orchestrated Cyberattack — CEO Called to Testify
Dario Amodei will appear before Congress to discuss the attempted cyberattack and broader national-security risks, especially regarding China. Google Cloud CEO Thomas Kurian will also testify on hyperscaler security posture.
Source: https://www.axios.com/2025/11/26/anthropic...

☁️ Claude Becomes the First Frontier Model Available on AWS, GCP, and Azure
Anthropic’s partnership with Microsoft and NVIDIA makes Claude the only model with full multi-cloud availability. Companies can now adopt Claude without switching clouds — a major distribution advantage. Anthropic also committed to $30B of Azure compute.
Source: https://blogs.microsoft.com/blog/2025/11/1...

🇨🇳 Alibaba & Bytedance Use Southeast Asia Data Centers to Train Qwen & Deepseek
Facing U.S. GPU export restrictions, Chinese AI giants have shifted training runs abroad to access H200 and B300 hardware. The U.S. may loosen restrictions on H200 GPUs.
Source: https://finance.yahoo.com/news/chinas-tech...

📄 Andrew Ng Launches an AI Agentic Reviewer for Research Papers
Trained on historical peer-review data, the model achieves near–human-level review quality and could shorten the months-long academic review cycle.
Source: https://x.com/AndrewYNg/status/19930019227...

🔐 OpenAI Confirms Data Breach via Mixpanel Analytics Vendor
A Mixpanel security issue exposed API-related data, prompting OpenAI to remove the vendor and perform a full audit of its third-party ecosystem.
Source: https://www.theregister.com/2025/11/27/ope...

📉 Gemini 3 Pro Demand Forces Google to Change Access Limits
Google removed the 5 prompts/day guarantee for free users and restricted free-tier context windows to 32k tokens — far below Gemini’s million-token capability.
Source: https://support.google.com/gemini/answer/1...

📱 Leak Reveals Early References to Ads Inside the ChatGPT Android App
This contradicts Sam Altman’s historical stance against ads in AI assistants. How OpenAI monetizes recommendations without compromising trust remains an open question.
Source: https://x.com/btibor91/status/199471415263...

📊 Report: AI Adoption Among U.S. Businesses Is Starting to Flatten
Larger enterprises show signs of completed adoption curves, suggesting we're entering the optimization phase rather than rapid onboarding.
Source: https://www.apolloacademy.com/ai-adoption-...

💰 Databricks in Talks to Raise $5B at a $134B Valuation
With projected 2026 revenue around $4B, the valuation sits at ~32× — signaling continued investor enthusiasm for AI infrastructure companies.
Source: https://siliconangle.com/2025/11/30/databr...

📐 DeepSeek Releases Math-v2 — A Stronger Mathematical Reasoning Model
The new model significantly improves performance on math Olympiad–style reasoning benchmarks.
Source: https://huggingface.co/deepseek-ai/DeepSee...

📘 Harmonic Math’s “Aristotle” Solves Erdős Problem #124
Aristotle becomes the latest AI system to independently solve a long-standing number-theory challenge, highlighting rapid advances in AI-driven scientific discovery.
Source: https://x.com/vladtenev/status/19949228272...