AI Updates on 2025-08-23

Google offers Veo 3 video generation free for the weekend, allowing users to create 8-second videos with sound through text prompts or photo uploads @GeminiApp
Mistral AI announces Mistral Medium 3.1 achieving top rankings on LMSYS Arena leaderboard, ranking #1 in English (no Style Control) and 2nd overall despite being a smaller model @MistralAI
Qwen releases Qwen-Code v0.0.8 with deep VS Code integration, enhanced MCP support, and responsive terminal UI features @Alibaba_Qwen

Coinbase CEO Brian Armstrong mandates all engineers set up GitHub Copilot or Cursor accounts within a week, demonstrating enterprise-level AI adoption requirements @TechCrunch
Hiring managers increasingly expect developers to be hands-on with AI coding tools as a baseline requirement, though solid developers who choose not to use these tools shouldn't be automatically rejected @GergelyOrosz
AI-native junior employees can drive AI adoption in organizations by explaining capabilities, demonstrating usage, and identifying opportunities where AI can save time or create value @a16z
Meta partners with Midjourney to license AI image and video model technology, indicating continued consolidation in the AI media generation space @TechCrunch

Google's Nanobanana model delivers professional-level Photoshop edits through text prompts, including facial expression changes, weather modifications, and style transfers while preserving scene context @deedydas
Linear integrates multiple AI agents including Cursor, Codegen, and Devin directly into product workflows, enabling feature building and bug fixes from within the platform @karrisaarinen
Runway launches an interactive game engine that creates comics with user choices in any visual style, offering more graphical capabilities than previous AI Dungeon-like experiences @emollick
Genspark AI Developer provides a zero-setup browser-based IDE with visual feedback and model selection options like Claude Opus 4.1 and GPT-5, designed for users with limited coding experience @fchollet
ChatGPT introduces project-scoped automatic memory features, allowing AI to remember context within specific projects while maintaining compartmentalization @simonw
Custom annotation apps provide significant advantages over generic tools, as demonstrated by a medical application where doctors can provide context-aware labeling with minimal typing @HamelHusain

Waymo's autonomous vehicles demonstrate 85% fewer serious injuries and 79% fewer overall injuries compared to human drivers across 57 million miles of data, suggesting significant safety improvements @emollick
X releases a model card for Grok showing concerning deception and sycophancy scores, though the document lacks detailed explanations for interpreting the results @emollick