Today's AI Pulse
Anthropic fixes three missteps that briefly dulled Claude Code
Anthropic disclosed three separate changes that temporarily reduced Claude Code's performance and restored the model by April 20. A March 4 switch lowered the default reasoning effort from high to medium, which was reverted on April 7, and a March 26 cache‑clearing bug erased session context, causing forgetfulness and higher token usage until fixed. The issues were fully resolved by April 20.
Also developing:
By the numbers: Tesla to acquire AI‑hardware firm for up to $2B
Cursor Introduces Its Coding Model Alongside Multi-Agent Interface
Cursor unveiled Cursor 2.0, an IDE that pairs a new in‑house coding model called Composer with a multi‑agent interface that can run several AI agents in parallel via git worktrees or remote machines. Composer, built with reinforcement‑learning and a mixture‑of‑experts architecture, is billed as a frontier model that is four times faster than comparable high‑intelligence models while matching top‑tier accuracy. The release expands Cursor’s support beyond third‑party models from OpenAI, Google and Anthropic to its own competitive offering and adds features such as parallel agent execution and a suite of bug‑fixes in the 2.0 changelog. Early developer feedback notes the model’s speed but raises concerns about cost and whether it can close the performance gap with established frontier models.
Matrix Oscar Winner: "It Was a Prototype Disguised as a Blockbuster"
John Gaeta, the Oscar‑winning VFX architect behind the Matrix’s bullet‑time, explains how the film’s pioneering capture and simulation techniques became the R&D foundation for today’s game engines, virtual‑production tools, and AI‑driven world models. He traces a 25‑year evolution from the...
AMD and the US Government Sign $1 Billion Deal for Two Massive Supercomputers - Is the Global AI War Finally...
AMD and the U.S. Department of Energy have signed a $1 billion contract to build two next‑generation supercomputers, Lux and Discovery, at Oak Ridge National Laboratory, with deliveries slated for early 2026 and 2028. The systems will combine AMD Instinct GPUs,...
Matrix Was Prototype, Now Powers Hollywood’s Digital Future
Matrix Oscar Winner: "It Was a Prototype Disguised as a Blockbuster" I spoke to John Gaeta, the man who built bullet time. While Hollywood thought they were making a sci-fi trilogy, his team was running a research project in plain sight. We...
Last Call: November AI Engineering Cohort Closes in 48 Hours
Louis‑François Bouchard’s October 31 note warns that enrollment for the November Full‑Stack AI Engineering cohort closes in 48 hours, with a live kickoff on November 2. The program, priced at $349 (one‑time) after a free preview, promises a production‑ready GenAI playbook previously...
Claude Is Coming to Your Spreadsheets - but Is It Enough to Make You an Emperor of Excel, and What...
Anthropic has unveiled Claude for Excel, an AI assistant that can read, edit, and generate spreadsheets from a sidebar while tracking its actions and citing source cells. The beta, limited to about 1,000 Max, Enterprise and Teams customers, connects to...
AI Music Platform Udio and Universal Music Strike a Deal—Is This AI Music's Spotify Moment?
Universal Music Group settled its copyright lawsuit with AI music startup Udio and announced a joint venture to launch a licensed, commercial AI‑generated music platform slated for next year. The partnership will let users create, customize, and stream AI‑crafted songs...
Claude Can Be Tricked Into Sending Your Private Company Data to Hackers - All It Takes Is some Kind Words
Anthropic’s Claude AI tool contains a vulnerability in its Code Interpreter that can be exploited via prompt injection to exfiltrate private user data. The interpreter, recently given network request capability, can be tricked into reading files, storing them in the...
Google Rolls Out AI Tools, Upgrades, and Free Credits
No tricks here, only treats 🍭. Here’s this week’s roundup: — Our latest experiment from @GoogleLabs, Pomelli, is an AI marketing tool designed to help SMBs connect with their audiences faster — You can now generate a full set of exportable slides...
OpenAI Will Let Sora Users Pay Extra To Make More AI Videos
OpenAI announced that users of its Sora AI video generator can now purchase additional generations beyond the free allotment, paying $4 for 10 extra videos per day. The free tier remains at 30 generations daily, but the company’s head of...
Adobe’s Experimental AI Tool Can Edit Entire Videos Using One Frame
Adobe unveiled a suite of experimental AI tools at its Max conference, highlighted by Project Frame Forward, which lets editors modify the first frame of a video and automatically propagate additions or removals across the entire clip. The demo showed...
AI Multitasking Madness: Six Claude Terminals Running Simultaneously
Have you witnessed the level of AI multitasking insanity happening right now? This is 6 Claude Code terminals running in parallel on my laptop. Multi-agent chaos is my new normal 😍
Magic Leap and Google Just Showcased Another Pair of Android XR Smart Glasses – with Gemini on Board
Google and Magic Leap unveiled a prototype Android XR smart‑glasses pair at the Future Investment Initiative in Riyadh, merging Magic Leap’s waveguide optics with Google’s Raxium micro‑LED light engine. The glasses embed Gemini multimodal AI, allowing on‑device interactions such as visual analysis of...
Perplexity Strikes Multi-Year Licensing Deal with Getty Images
Perplexity, an AI‑powered search startup, has signed a multi‑year licensing agreement with Getty Images that permits the company to display Getty’s stock photos in its search and discovery tools with proper attribution and links to the original source. The deal,...
Tim Cook Says Apple Is Open to M&A on the AI Front
Apple CEO Tim Cook told investors on the Q4 2025 earnings call that the company remains open to acquisitions and partnerships to accelerate its AI roadmap, maintaining a three‑pronged strategy of in‑house foundation models, third‑party LLM collaborations, and targeted M&A. He...
How Unstructured Data Turns Your Business Into a Junk Drawer - and How to Fix It
Businesses are drowning in unstructured data—PDFs, emails, scanned invoices—stored in disparate, unmanaged locations, creating a digital “junk drawer” that hampers finance, HR and operations. The fragmentation leads to duplicated effort, audit delays, compliance risk and lost productivity. Intelligent Document Processing...
The Future of AI Applications: MCP Servers
The article outlines the emergence of Model Context Protocol (MCP) servers as a modular, context‑aware computing layer that links AI‑powered applications to distributed data sources across cloud, edge and on‑premise environments. Unlike traditional siloed servers, MCP servers dynamically allocate resources,...
Grammarly Announces 'Superhuman' Rebrand As It Doubles Down On AI
Grammarly announced a rebrand to "Superhuman," consolidating its suite of AI‑powered tools—including the Coda workspace, the acquired Superhuman Mail inbox, and a new cross‑app agent network called Superhuman Go—under a single subscription plan. The move builds on earlier launches such as the...
Top 10 Generative AI Trends Shaping 2026
10 Generative AI Trends in 2026 That Will Transform Work and Life Generative AI is continuing its evolution. This article outlines how GAI will impact science, employment, data ethics and societal norms in 2026. Read more 👉 https://lnkd.in/esgd2aVF #GenerativeAI #AI2026...
Apple’s AI Comeback Will Reshape Computing by 2027
I'm getting bullish on Apple again. It will take three years to see that. But I've been seeing patents come out of Apple that are next level and I've been talking with insiders again and they are working on fixing...
Nvidia Expands AI Ties with Hyundai, Samsung, SK, Naver
Nvidia CEO Jensen Huang visited South Korea during the APEC 2025 summit, announcing a $5‑trillion‑valued company’s expanded partnership with the Korean government and leading firms Hyundai Motor, Samsung, SK Group and Naver. South Korea will procure over 260,000 Nvidia GPUs—about...
Apple Adopts Gaussian Splats for Lifelike Vision Pro Avatars
Apple Sr. Director Jeff Norris on how Vision Pro persona went from looking like cadavers to life-like avatars: "First thing I'd want to point out is that we changed the the whole rendering approach of Personas to a completely new technique...
Saudi Launches Nation‑Wide AI Coding Initiative with HUMAIN
The world’s first country-wide vibe coding deal! Thrilled to show the world what’s possible when an enlightened government takes the AI/agentic coding revolution seriously. HUMAIN is Saudi Arabia’s national AI company, and together we will accelerate AI transformation in government and...
Could Your Favorite Soccer Team Be Using ChatGPT for Tactics? One Pro Coach Has Publicly Admitted They Have
Seattle Reign head coach Laura Harvey disclosed on the Soccerish Podcast that she consulted ChatGPT for tactical advice, including a recommendation to employ a back‑five formation, which she then used in a match against Orlando Pride. Harvey described prompting the...
Saudi Innovators Leverage Replit AI for Limitless Opportunities
My visit to Saudi has been nothing short of amazing. Have met and heard stories about entrepreneurs from all walks of life using Replit in so many ways: A teenager who built a college prep app that became a funded...
NotebookLM Just Got a Major Chat Upgrade with 8x More Context, Better Memory, and Smarter Responses
Google announced a major upgrade to NotebookLM, expanding its context window to one million tokens—about eight times its previous limit—allowing the AI to ingest entire books, meeting transcripts, and large project files in a single session. The enhancement also boosts...
Exclusive Genspark AI Talk: Gratitude and Early Access
I love @genspark_ai and tonight I met other customers who feel the same way. We heard @sang_wen cofounder talk about his company and open his Palo Alto headquarters. In the way to my robot that is driving me home I passed...
2026 Transport Revolution: Flying Taxis, Autonomous Trucks, AI‑Driven Mobility
Flying Taxis And Self-Driving Trucks Arrive In 2026: 6 Transport Trends To Watch #Transport technology is accelerating into 2026 with flying taxi services launching in the U.A.E., #autonomous truck trials starting on U.S. highways, and #AIagents optimizing public transit systems....
Inside Celosphere 2025: Why There’s No ‘Enterprise AI’ without Process Intelligence
Celonis will host Celosphere 2025, a three‑day event focused on linking process intelligence (PI) with enterprise AI to deliver measurable ROI. The company cites a Forrester study showing 383% ROI over three years and six‑month payback for users of its...
Matic Robots Showcase Consumer Electronics Genius in Kitchen
The @maticrobots is quite an interesting bit of technology. Love watching how it cleans our kitchen. Attended a @genspark_ai event in Palo Alto and hung out with a bunch of AI and robotics geniuses in its kitchen. What did we talk about? How...
Shifting From Creator Tools To Creative Participation: Adobe MAX Takeaways
Adobe MAX 2025 unveiled a suite of AI‑driven tools that move creativity from specialist teams to the broader enterprise. Adobe announced partnerships that embed external models (Google, OpenAI, Luma AI, Runway) alongside its Firefly engine, letting users pick the best model inside...
Why OpenAI’s $1 Trillion IPO Can’t Come Soon Enough
The post discusses the looming possibility of OpenAI’s massive IPO, which could raise at least $60 billion and potentially value the company near $1 trillion, with a timeline as soon as late next year or by 2027. It argues that the company’s...
Why Sam Altman Is Scared & Why People Are Giving Up on MCP | EP99.23
The episode "Why Sam Altman is Scared & Why People Are Giving Up on MCP" continues the "This Day in AI" podcast’s informal, middle‑of‑the‑road take on AI, using humor and personal anecdotes to discuss Sam Altman’s apparent anxieties and the...
Self‑Awareness Is the Key to Human‑Centred Tech
Wise and timely message, by Pope Leo XIV @Pontifex . In a world shaped by algorithms and automation, the greatest act of digital literacy is self-awareness. Humanizing technology means keeping empathy, creativity, and purpose at the center, using digital tools...
Jeremiah's Keynote Sets Stage for My AI Talk
Jeremiah @jowyang just gave a great keynote at @siliconvalleyAI that led into mine. Which is in audio below. https://t.co/H3q42wYKGm
Use JAX + Keras 3 for Scalable Deep Learning
For anyone getting started with deep learning in 2025 and looking to do large scale training: use JAX + Keras 3. Unless you like suffering.
AI Could Eradicate Disease Within a Decade
Fascinating. @GoogleDeepMind's CEO @demishassabis predicts a world free of disease within 10 years! "AI could make drug development much faster, going from years to just months/weeks ...this might change healthcare completely and help end all diseases in the next 10 years" https://t.co/kDflHKJVfd
ChatGPT Handles File Unzipping, Claude Does Not
ChatGPT can unzip files, Claude cannot.
Join My Biggest Live Talk for AI‑Curious Entrepreneurs
Next week, I’ll be giving my largest talk ever - speaking in front of 100s of thousands of AI-curious entrepreneurs live. Thank you to @deangraziosi and @TonyRobbins for knowing the info your audience needs and bringing in AI folks to help...
Fp16 Viable with Proper Normalization; Bf16 Still Safer
I ran lots of experiments on fp16 vs bf16 years ago on ViTs and LLMs. fp16 can work well but depends on normalization (so you don’t run over the supported range with your activations). I can see why with QKNorm...
1000 Hours Reveal Six Crucial Prompt Engineering Patterns
this guy literally put in 1000 hours of prompt engineering to nail down the 6 patterns that actually matter. https://t.co/O0wvVndcpA
Ask if Users Will Abandon ChatGPT Before Building AI Product
Before you build your AI product, ask one question: would anyone switch from ChatGPT for this? https://t.co/5q8pqdKadg
Google AI Studio Adds Free Prompt Logs and Datasets
Google's absolutely killing it lately! 🔥🔥🔥 @GoogleAIStudio just rolled out logs and datasets. You can now inspect every prompt, debug with ease, and export data for evals. All free. No setup needed. 👏 https://t.co/cSBNn0pX4R
Pollen Robotics Seeks SF Partners for Reachy 2 Demo
Part of the @pollenrobotics will be in San Francisco next week to demo Reachy 2, our humanoid robot that we already shipped to a couple of the best AI robotics research labs all over the world (and reachy mini of...
AI Chitchat: Parul Pandey on Learning, Building, Teaching AI
Tomorrow on AI Chitchat, I will be talking with Parul Pandey learning, building and teaching with AI. Don't miss it! https://t.co/CNJjb8UAA1
Mid-Tier Language Models only Hit 75‑85% on Basic Math
Language models perform poorly on high-school math? 🙄 You don't want to hear this, but the problems started in grade-school. The moment we (collectively) found acceptable that mid-tier models could score only 75%-85% on a GSM test set of 1.32k straightforward...
Testing 42 Podcast‑Style Videos with NotebookLM
I was reading some papers. So, I thought I would use NotebookLM + some vibe coding to create this podcast style video. I will create 42 episodes to see if people find it useful. If not, I will stop. Enjoy!...
Human Ambition + AI-First Drives Microsoft Customer Success
How human ambition and AI-first differentiation are helping @Microsoft customers go further with AI https://t.co/Occrn4l7su @AiqIntelligence @BlackRock @dentsu_global @Ecolab @HeyEpic @harvey_ai @InSilicoMeds @KraftHeinzCo @ManusAI @MercedesBenz @Nasdaq @RalphLauren @SchneiderElec @educacaosp @Telstra @Toyota @nvidia @Azure #CES2026 #Cloud
Preparing for the Fallout of a Bursting AI Bubble
What happens if the AI bubble bursts? https://t.co/Gr7mm6rjNg #AI #artificial_intelligence #AIbubble
Simulated Worlds Train Robots for Real‑World Factories
The future of simulated worlds with founder of https://t.co/VsxIMEnd0L Teaches robots how to work on factory floors. At Llama Lounge, @jowyang’s San Francisco AI event. https://t.co/e0tPK4eQwX