SAP's CTO just admitted something most enterprise vendors won't: AI isn't a technology upgrade, it's a business model overhaul. While startups chase the latest model releases, the companies that actually run the world's supply chains are quietly figuring out what happens when your "operating system" gets intelligence.
01
SAP's CTO: AI is a business model shift, not just a tech upgrade
SAP CTO Philipp Herzig sat down with No Priors to explain how the enterprise giant is approaching AI across its 400,000 customers. Herzig described SAP as "the operating system of a company" that handles everything from finance and HR to supply chain and manufacturing. His key insight: AI represents a fundamental business model transition, not just new features bolted onto existing software. He also argued that large language models alone aren't sufficient for predictive analytics in enterprise environments.
Why it matters: SAP runs the back office for most Fortune 500 companies. When they say AI changes business models, not just workflows, every CFO using their software is about to discover what that means in practice.
Google's Ads Advisor gets three new AI safety features
Google announced new "agentic safety and policy features" for Ads Advisor, their AI assistant for Google Ads management. The update focuses on protecting and streamlining ad accounts, though Google provided limited details on what the three specific features actually do.
Why it matters: Google is testing AI agents on billions of dollars in ad spend. If these safety features work, expect every other ad platform to copy them. If they don't, expect some very expensive AI mistakes.
Product manager Josh Woodward shared that Google's NotebookLM can now automatically label and categorize sources you upload. The feature appears designed to help researchers and students organize large collections of documents without manual tagging.
Why it matters: This is the unglamorous AI feature that actually saves time. Every grad student and consultant who drowns in PDFs just got their filing cabinet automated.
Product manager builds Star Fox game in 15 minutes with GPT-5.5
Peter Yang continued his tradition of testing new AI models by asking them to build classic video games. This time, GPT-5.5 combined with Codex successfully created a working Star Fox game after just 15 minutes of prompting, complete with gameplay mechanics and graphics.
Why it matters: Yang's game-building tests have become an unofficial benchmark for AI coding ability. When a model can build a functional 3D space shooter in a quarter-hour, we've crossed into "AI game developer" territory.
DeepSeek V4 arrives with 1M token context and Huawei chip support
China's DeepSeek released V4 Pro and V4 Flash, their first major update since V3 in December. The models feature 1 million token context windows, new compression techniques that reduce memory usage by 90%, and critically, support for Huawei Ascend chips instead of NVIDIA hardware. Latent Space Newsletter called it roughly equivalent to "Gemini 3.1, GPT 5.4, Opus 4.6 level" performance.
Why it matters: DeepSeek just demonstrated that Chinese AI can match Western labs while running on non-American chips. The geopolitical implications of AI independence just became very real.