Today's AI Pulse
OpenAI unveils ChatGPT Images 2.0 with multilingual text and ‘Thinking’ mode
OpenAI rolled out ChatGPT Images 2.0, a next‑generation image model that adds multilingual typography, real‑time web research, and agentic reasoning. The new ‘Thinking’ mode lets paid users generate up to eight coherent images per prompt, including floor plans, maps and UI mock‑ups. The feature expands the model’s ability to work with uploaded content.
Also developing:
This Startup Thinks Slime Mold Can Help Us Design Better Cities
Cambridge startup Mireta has built an algorithm inspired by slime mold’s natural network-building to optimize urban infrastructure—claiming it can improve transit times, reduce congestion and factor in constraints like flood zones, traffic patterns and budgets. The team translated observable slime mold behaviors into rules that reproduce efficient, resilient pathways without using live organisms, and has applied the approach to about five projects since launching this year. Experts say the method shows promise for bottom-up, adaptive planning but caution it may struggle to overcome political, social and stakeholder barriers to real-world implementation.
From Slop to Sotheby’s? AI Art Enters a New Phase
Generative AI is moving from novelty to institutional recognition as artists using tools like Midjourney and Runway gain large followings, museum placements and auction sales—most notably an AI-generated bitcoin NFT by Henry Daubrez that sold for $24,000 at Sotheby’s and...
Voice‑Chat GPT Cuts My Phone‑Screen Addiction
Okay, I’m clearly addicted to my phone. But one thing is helping. Usually I’m constantly checking and replying to texts and emails and socials, scrolling AI articles, reading papers. But when I’m rushing to get ready, it makes me late for...
Anthropic Connects Claude to Microsoft Teams, Outlook, and OneDrive
Anthropic has integrated its Claude AI assistant with Microsoft 365—including SharePoint, OneDrive, Outlook and Teams—allowing Claude to surface and analyze documents, emails, chats and meeting summaries directly in conversations. The Microsoft 365 connector, built on Anthropic’s open-source Model Context Protocol,...
Researchers Find Adding This One Simple Sentence to Prompts Makes AI Models Way More Creative
Researchers from Northeastern, Stanford and West Virginia universities have introduced Verbalized Sampling (VS), a prompt-level technique that boosts creativity in LLMs and image generators by adding a single instruction—"Generate 5 responses with their corresponding probabilities, sampled from the full distribution."...
OpenAI's Sora 2 Sends Shockwaves Through Hollywood
OpenAI vs Hollywood: Why Sora 2 by OpenAI is freaking out the movie industry? https://t.co/DsFiZ7oXbI #movie #openai #sora2 #soraArt #AIart #aivideo #hollywood @PawlowskiMario @chidambara09 @Ym78200 @CurieuxExplorer @efipm @bigfundu @sayedflah @Ronald_vanLoon @cyngn @belindabeibi @odisseiaalfa @DigitalColmer @MyCompanionsAI @KirkDBorne @patricegorissen @jeffkagan @EdwardKens50830 @enilev @insom_ai333...
AI Amplifies Social Engineering and Malware Threats
Social engineering, malware and future of cybersecurity in AI https://t.co/YkVbQefP9z #SocialEngineering #CyberSecurity #AI #artificial_intelligence @SpirosMargaris @PawlowskiMario @mvollmer1 @gvalan @ipfconline1 @LaurentAlaus @Shi4Tech @Fisher85M @kalydeoo @Ym78200 @Nicochan33 @chboursin @3itcom @Fabriziobustama @sallyeaves @helene_wpli @ahier @rwang0 @EvanKirstel @RLDI_Lamy @Analytics_699 @Khulood_Almani @tewoz @chidambara09 @IsabellePiel29 @ClementIsa @SvetBnov...
New York Bans AI-Enabled Rent Price Fixing
New York Gov. Kathy Hochul signed a law banning landlords from using AI-enabled price-setting software, making New York the first U.S. state to outlaw algorithmic rent pricing after several city-level bans. The statute not only forbids using such tools to...
Silicon Valley Execs Spend Fortunes on High‑IQ Embryos
Inside Silicon Valley’s Growing Obsession with Having Smarter Babies Tech execs are paying tens of thousands to find brilliant dates or select high-IQ embryos. ‘They want to raise high-performing children.’ Read 👉 https://t.co/eC8HTcXPr6 @WSJ #AI #IoT https://t.co/yXYu192VXg
JSON Prompting Turns Vague Ideas Into Structured, Reusable Layouts
What makes JSON prompting great? Structure. It turns a vague idea into a clear, inspectable layout. You can isolate changes, copy sections, reuse logic easily. Gemini Flash 2.5 (Nano Banana) handles JSON prompting flawlessly. Prompt in 🧵↓ https://t.co/SsisTHWW1y
Ars Live Recap: Is the AI Bubble About to Pop? Ed Zitron Weighs In.
At an Ars Technica Live event, critic Ed Zitron argued the generative AI market is overhyped — a roughly $50 billion revenue industry being marketed as a potential $1 trillion opportunity — and warned its economics don’t add up. He...
Start Using MLX with Guidance From the MLX King
This is today. Learn how to get started with MLX from the MLX King himself. DON'T MISS IT! https://t.co/F4SbYuS8GN
Open Agent Builder: Free Visual AI Workflow Playground
Ace. @Firecrawl_dev just dropped Open Agent Builder, an n8n-style visual playground for building AI agents. Design and deploy agent workflows from a simple canvas. Add Firecrawl, LLMs, logic nodes, and watch it all run as an API 🔥 100 % free and open source. Repo...
Why RL Won — Kyle Corbitt, OpenPipe (Acq. CoreWeave)
Kyle Corbitt, co‑founder and CEO of OpenPipe (recently acquired by CoreWeave), explains the industry’s shift from supervised fine‑tuning to reinforcement‑learning‑based agent training. He argues that 90% of AI projects stall not because of capability limits but due to reliability gaps,...
AutoTrain Paper Reaches Five Citations After Months
Opened google scholar after several months and to my surprise AutoTrain paper has 5 citations now 🤩 https://t.co/zjuZnIMIwb
Free Limited-Time Access to Moondream3 via Luminal Demo
Demo getting praise in AI community. Fast!
Pt.2 Beyond Phishing: Cyber Threats in the Age of AI with Four Flynn
The post serves as a detailed guide to Professor Hannah Fry’s 2025 podcast series, spotlighting Episode 10 Part 2 which examines the human side of cybersecurity, the motivations of attackers, and AI‑driven defenses such as passkeys and risk‑based authentication. It also recaps the...
AI Threatens Human Memory Retention
Is artificial intelligence fading human memory? https://t.co/Numsk5NJ0i #artificial_intelligence #AI @lexfridman @KirkDBorne @Ronald_vanLoon @erikbryn @antgrasso @sallyeaves @Nicochan33 @HaroldSinnott @mvollmer1 @marcusborba @CatherineAdenle @vinod1975 @YuHelenYu @ingliguori @fogoros @ylecun @TerenceLeungSF @SourabhSKatoch @wil_bielert @HsrYueli @faustospain @Whats_AI @kaifulee @demishassabis @marek_rosa @AndrewYNg @BernardMarr @CurieuxExplorer @andresvilarino @KentPage @kimgarst @raehanbobby...
AI Boom May Be a Bubble, Here’s Why
Why AI boom might be a bubble? https://t.co/N3vrJhnhX7 #AI #artificial_intelligence #TechNews #investments @AlbertoEMachado @Eli_Krumova @postoff25 @Khulood_Almani @anand_narang @NutritiousMind @baski_LA @TanyaSinha_ @devaang @AlAmadi1 @jeancayeux @enilev @efipm @mvollmer1 @Nicochan33 @RagusoSergio @FrRonconi @Shi4Tech @sallyeaves @LaurentAlaus @Fabriziobustama @smaksked @MargaretSiegien @PawlowskiMario @gvalan @Ym78200 @mikeflache @EduardoValenteI @ipfconline1...
How Anthropic’s ‘Skills’ Make Claude Faster, Cheaper, and More Consistent for Business Workflows
Anthropic launched 'Skills,' a capability for its Claude AI that lets organizations package instructions, code, and reference materials into reusable, composable folders the assistant loads on demand, using "progressive disclosure" to avoid context-window limits. Skills work across Claude products and...
Anthropic Not Targeted, Actually Partnered with White House
A new Bloomberg story describes Anthropic as a “White House target” after my tweet the other day. Nothing could be further from the truth. Just a couple of months ago, the White House approved Anthropic’s Claude app to be offered...
AI Learns Game Rules From a Single Play
Super smart robot breakthrough: I had to ask Grok why this paper is important. Here is its answer: >>> Imagine you're playing a video game, like Minecraft, and you've only played it once. Normally, to get really good, you'd have to play...
Spotify Says It’s Working with Labels on ‘Responsible’ AI Music Tools
Spotify announced a formal agreement with major labels — Sony, Universal, Warner, Merlin and Believe — to jointly develop “responsible” generative AI music products and to stand up a dedicated research lab and product team. The pact emphasizes upfront deals,...
Future Jobs: Human Neurons and Stem Cell Computing
What jobs will AI never replace? @BatsouElef asks. Every job I can think of an AI can eventually do. Even being your mom or playing sports. So I'm asking myself "what kinds of jobs will we come up with after AI and...
Turn ChatGPT Into Your Strategic Decision Partner
🤯 What if ChatGPT could make you smarter — not just faster? Most people use it for busywork. I’ll show you how to use it for what actually matters: career moves, negotiations, strategic planning, and high-stakes decisions. Introducing my new...
Is Agentic Commerce a Thing?
Forrester analysts outline “agentic commerce” as nascent AI systems—composed of foundation models and task-specific agents—that can take actions to assist or conduct parts of the shopping journey, from discovery to, in rare cases, checkout. The firm distinguishes owned experiences (brand-controlled...
AI Now Drives Avatars, Making Real-Time VTubing Suit‑free
You can just transfer your entire performance to any character now. Fed it my guitar playing to drive this avatar, and it holds up well. Wan 2.2 Fun Control and Animate are powerful primitives. At the rate things are going, this'll be...
Rigorous Evals and Error Analysis Accelerate AI Agent Development
Readers responded with both surprise and agreement last week when I wrote that the single biggest predictor of how rapidly a team makes progress building an AI agent lay in their ability to drive a disciplined process for evals (measuring...
ACE Prevents Context Collapse with ‘Evolving Playbooks’ for Self-Improving AI Agents
Stanford and SambaNova introduced Agentic Context Engineering (ACE), a framework that prevents “context collapse” by treating an LLM’s context as an evolving, itemized playbook updated incrementally by Generator, Reflector and Curator modules. In evaluations ACE outperformed strong baselines—improving agent-task performance...
AI‑Powered Seedream 4.0 Turns iPhone Shots Into Viral Ads
Marketers: How do you introduce a new product to the world? One that no one knows they want? And you really wanna make it viral? That's my storytelling challenge for Illusion of Life, which has just shipped to its first customers a...
San Francisco's AI Gold Rush Revives Its Frontier Spirit
LIVE FROM DREAMFORCE: Marc Benioff (@Benioff), Chair, CEO and Co-Founder of @Salesforce, on the spirit of San Francisco and why AI is its next gold rush. “There’s a metaphysical spirit to San Francisco, a spirit of adventure, of innovation, of freedom....
Inside the Web Infrastructure Revolt over Google’s AI Overviews
Cloudflare has automatically updated robots.txt files on roughly 3.8 million domains and rolled out a new Content Signals Policy—covering about 20% of the web—to let site operators opt out of AI uses (ai-input and ai-train) while distinguishing traditional search from...
Unlock Existing ChatGPT Auth, Not New Sign‑In Layer
Huge if true. Wrote a bunch about the Sign in with ChatGPT opportunity back in April. “Today, there’s no secure way to link a ChatGPT account to another tool. No OAuth. No scoped tokens. No way to persist context responsibly. What’s needed...
Real-Time World Models Promise Holodeck‑Level Robotics
OK @drfeifei was already near the top of my list of favorite people alive, but this pushed her up to the top. Infinite real time realities in a Holodeck. This work has sped up my expectations about robotics. @adcock_brett told @Benioff...
New Chapter Releases Symbolic Verifier for LLM Evaluation
Chapter 3 on building a symbolic verifier for LLMs from scratch is now live: https://mng.bz/lZ5B . And with this, the first 176 pages of Build A Reasoning Model (From Scratch) are now available. This verifier is a useful method for...
Analytics: The Essential Tool for Content Marketers
🔴 If you're a content marketer, this is why you have Analytics! Some content gets real results. Some content …not so much. This is obvious to every marketer who's pushed the publish button. But it’s less obvious that content performs...
AI Products Mostly Re‑segment Existing Markets, Not Create New Ones
Most AI products are not creating new categories as much as they are resegmenting existing ones. The distinction is critical for a team to understand. These lessons have been learned over and over again by many of us.
Massive GPU Compute Powers Meta’s AI Talent Hunt
They ran 400,000 GPU hours. And that is how Meta recruits hyper-smart AI researchers. Oh, and the pay. :-)
Genie 3 Elevates World Labs with Generative Video Memory
Now this is fking cool. Genie 3 made the World Labs approach look dated. But now they drop a step towards the best of both worlds. Generative video w/ spatial memory. And a live demo to boot. Making the bitter lesson...
Y
World models = holodecks. But @ylecun gives a much more complete and precise answer. :-)
First Hour-Long RL Podcast for Robotics Enthusiasts
An hour of reinforcement learning. If you are listening, like I am, you are my kind of people! Or you might be working at a robotics startup. Which means you are my kind of people anyway. :-)
Accelerate Keras Models with Low‑precision Quantization
Run your models faster with quantized low precision in Keras! You can quantize any model (one of your own models or a KerasHub pretrained model) via `model.quantize(mode)`. Supports int4, int8, float8, GPTQ. Works with JAX, TF and torch. Links to guides in next...
Master Prompting: Five Techniques for Better Answers
Bad prompts get bad answers. Level up your prompting with our prompt engineering guide. Here are five techniques to get you started 👇 https://t.co/N8SVaBmms4
Discovering Poke: A Powerful Email‑Watching iPhone Assistant
This was a very interesting conversation with @marvinvonhagen founder of @interaction maker of “Poke.” This got me to use it even more on my iPhone. Great assistant that watches emails and more.
Best Ecosystem Wins: Partner Orchestration Drives AI Success
Why do 72% of AI projects never reach production? After 3 days at #IBMTechXchange, the answer was clear: The best ecosystem wins. ✅ AI scaled through partner orchestration 📷 Claude x IBM = strategic capability fill 📷https://t.co/ZTao4tnvpN + newsletter: https://t.co/HelZzurUa5...
AI Partnership Accelerates Fusion Toward Unlimited Clean Energy
Super excited to be collaborating with Commonwealth Fusion Systems @CFS_energy to use AI to accelerate fusion development - and move closer to a sustainable future with limitless clean energy
Funding General Intuition to Advance AI Spatial-Temporal Reasoning
Super excited about funding @PimDeWitte on this exciting journey towards AI understanding the physical world !
Real-Time 3D World Model Runs on One H100
Very excited to share @theworldlabs ‘s latest research work RTFM!! It’s a real-time, persistent, and 3D consistent generative World Model running on *a single* H100 GPU! Blog and live demo are available below! 🤩
Dagger’s Container Use Simplifies AI Agent Parallelism
Docker is overkill for running AI agents in parallel. @Dagger_io’s new Container Use is a much simpler, agent-first alternative: → No Dockerfile needed → Git-native isolation → Full logs & real‑time visibility + easy one-liner commands to do it all! 🙌 Let’s compare 🧵↓ https://t.co/etVd1uHheZ
OpenAI Slashes Microsoft Revenue Share, Gains $50B
Major shift at OpenAI: revenue share with Microsoft & partners to fall from 20% → 8% by end of decade, potentially netting @OpenAI an extra $50B. Big move on margins, and implying pressure on terms and structure. 🔗 https://t.co/kjyCLj1C6B #AI...