Sunday Rundown #145: Deep Research & Deep Cuts
Your skimmable AI news catch-up of the week.
Happy Sunday, friends!
Welcome back to the weekly AI news roundup.
In case you missed it, here’s this week’s Thursday deep dive:
If you’re consistently missing out on my emails, remember to check your “Promotions” tab and mark whytryai@substack.com as a “Safe Sender.”
👩💻 AI releases
Amazon added AI merch design to the Shopping app, letting customers describe a design idea to Alexa to have it created and printed on a product.
Anthropic news:
Claude Fable 5 is a top-tier model that leads benchmarks in knowledge work, scientific research, software engineering, and vision. (But do you need it?)
Claude Managed Agents can now run on a schedule and securely access authenticated tools to perform tasks like data sync and reporting.
Deezer launched a free AI Music Detector that scans your playlists on Apple Music, Spotify, and 18 other services to detect and flag AI-generated tracks.
ElevenLabs released Avatars, persistent AI characters made from reference images that you can use for talking-head videos with lip syncing.
Google news:
DiffusionGemma is an open-source model that generates entire blocks of text simultaneously instead of token-by-token, making it 4x faster.
Gemini 3.5 Live Translate delivers near real-time voice translation across 70+ languages while preserving the speaker’s tone and pitch.
NotebookLM can now do multi-step research, write and execute code, and export to 11 new formats. (Only AI Ultra subscribers for now.)
Information Agents in Search can continuously monitor topics and send you updates with links when there’s new info. (Only AI Ultra subscribers for now.)
OpenAI introduced a Lockdown Mode that limits ChatGPT’s access to external connections and web search to guard against prompt injection attacks.
Perplexity added Deep Research to Computer, letting it spin up agents that turn complex queries into cited reports, decks, dashboards, and more.
🔬 AI research
Apple previewed a rebuilt Siri AI powered by Gemini, with a camera mode, cross-app sharing, onscreen awareness, and personalized context.
Meta will update the Edits app with an AI assistant and a desktop version to help you brainstorm ideas, analyze performance, and edit videos on a bigger screen.
📖 AI resources
“Agents’ Last Exam” [BENCHMARK]: a new tough benchmark that measures model performance across 1,490 real professional workflow tasks.
“Anthropic Public Record” [RESEARCH]: a survey of 52,000 Americans to gauge their opinions and feelings on AI deployment and societal impact.
“Claude Fable 5 - Full 319-page Breakdown” [VIDEO]: a comprehensive deep dive into Fable 5 by the always excellent AI Explained.
🔀 AI random
Anthropic had to disable access to Claude Fable 5 and Mythos 5 after the US government issued an export control directive aimed at preventing jailbreaks.
OpenAI submitted a confidential draft S-1 to the SEC as a formal first step toward a potential IPO.
🤦♂️ AI fail of the week
I wanted a man to cut himself shaving, not star in a bizarre horror comedy. (Also, very convincing “Ouch” there, bruh. Oscar-worthy.)
📹 Live & Learn update
I had to cancel my last Live & Learn session on short notice a few days ago (Friday).
But I’ll try again this coming Friday with a test of AI chatbots and standalone tools that let you make conversational edits to your images.
Join the session on Friday, June 18, at 2PM CET (8AM EST):
Also, please let me know what you’d like to get out of these live sessions by taking this quick chat-based survey.
Live & Learn concept in a nutshell:
I pick 3-5 tools in the same category
I test them live and rate them on 3-5 specific dimensions
I end up with a leaderboard based on my observations and ratings
See you there on Friday!