Prompt Notes #3 - weekly AI digest

AI goes fast, helping you (and myself) catching up.

Welcome to the 3rd edition of Prompt Notes, your new favourite weekly AI digest.

This edition will cover the week from 16 February to 22 February

What caught my attention?

Sam Altman’s recent behaviour: his BG2 interview, where he basically told early investors they could sell their shares if they weren’t happy, his salty response to the Claude ads, and now this new meme‑worthy - but very revealing - line about AI’s energy consumption.

TLDR

• OpenAI raises revenue forecasts but adds $111 billion to cash burn projections due to spiralling AI training and running costs. (The Decoder)

• Anthropic launches Claude Code Security to detect vulnerabilities missed by conventional scanners, triggering a sell-off in cybersecurity stocks. (The Decoder)

• NVIDIA is reportedly set to invest $30 billion in OpenAI. (The Decoder)

• OpenAI CEO Sam Altman states that AGI is close, superintelligence is not far off, and internal models are accelerating research. (The Decoder)

• ByteDance restricts its AI video tool, Seedance, following Disney’s legal threats over IP infringement. (The Decoder)

---

Rest of the news

• A new benchmark tests AI models on finding specific photos in personal collections, revealing sobering results. (The Decoder)

• Anthropic’s study shows that AI agents primarily thrive in software development, with limited adoption elsewhere and low user-granted autonomy. (The Decoder)

• India hosts a four-day AI Impact Summit attended by executives from OpenAI, Anthropic, Nvidia, Microsoft, Google, Cloudflare, and heads of state. (TechCrunch AI)

• Anthropic rolls out new desktop features for Claude Code to automate more of the development workflow. (The Decoder)

• Microsoft’s new gaming CEO vows not to flood the ecosystem with endless AI-generated content. (TechCrunch AI)

• A Google VP warns that LLM wrappers and AI aggregators may not survive due to shrinking margins and limited differentiation. (TechCrunch AI)

• Google releases Gemini 3.1 Pro, an LLM capable of handling more complex forms of work with record benchmark scores. (TechCrunch AI)

• UAE’s G42 partners with Cerebras to deploy eight exaflops of compute through a new system in India. (TechCrunch AI)

• Article explores the concept of launching AI data centers into orbit to mitigate Earth’s environmental impact from massive generative AI facilities. (Wired AI)

• Amazon Web Services attributes a 13-hour outage in one system to its AI coding assistant Kiro. (The Verge AI)

• InScope raises $14.5M to automate financial statement preparation using AI. (TechCrunch AI)

• Trump administration repeals Biden-era Mercury and Air Toxics Standards on power plants amid rising electricity demand from AI data centers. (The Verge AI)

• Nvidia expands early-stage investments and ties with India’s AI startup ecosystem through investors, nonprofits, and venture firms. (TechCrunch AI)

• Accenture ties employee promotions to usage of its AI tools, monitoring individual logins. (The Decoder)

• CEOs of Read AI and Lucidya state AI tools replace tasks rather than workers. (TechCrunch AI)

• Discussion highlights the intense competition for AI researcher talent concentrated in Bay Area companies. (The Verge AI)

• OpenAI partners with Reliance to integrate AI search into JioHotstar with two-way streaming links in ChatGPT. (TechCrunch AI)

• Warner Bros. accuses ByteDance of deliberately training Seedance 2.0 on its characters, alleging copyright infringement. (The Decoder)

• Apple advances development of smart glasses, a pendant, and camera-equipped AirPods, targeting smart glasses production for late 2026. (The Decoder)

• Airbnb reports that traffic from AI chatbots converts to bookings at a higher rate than Google search traffic. (MarTech)

• Shopify highlights its AI advantages in recent earnings report. (Stratechery)

• High-maturity brands increase spending on generative engine optimization for visibility in AI search. (MarTech)

• Fei-Fei Li’s World Labs raises $1 billion to develop AI world models for understanding and operating in 3D environments. (The Decoder)

• Google integrates DeepMind’s Lyria 3 beta into the Gemini app for generating 30-second AI music tracks from text, images, or videos. (The Verge AI)

• Google publishes its 2026 Responsible AI Progress Report. (Google AI Blog)

• Anthropic releases Claude 4.6 Sonnet with 1 million token context, improved web search, and adaptive thinking for complex coding. (MarkTechPost)

• 17 US-based AI companies have raised $100M or more in 2026, including three rounds over $1B. (TechCrunch AI)

• Ricursive Intelligence raises $335M at a $4B valuation in four months, driven by its founders’ reputation. (TechCrunch AI)

• Alibaba releases free open-weight Qwen3.5, a hybrid linear attention and MoE model matching top Western models with 17B active parameters per query. (The Decoder)

• ByteDance plans to improve safeguards on Seedance 2.0 after accusations from Disney, Paramount, and Hollywood groups of copyright infringement. (The Verge AI)

• Google AI introduces WebMCP protocol for direct, structured website interactions enabling AI agents to bypass screenshot-based browsing. (MarkTechPost)

• OpenClaw is released as a self-hosted personal AI assistant integrable with WhatsApp, Telegram, Slack, and Discord for tasks and voice interaction. (MarkTechPost)