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
Meta AI’s App Downloads and Daily Users Spiked After Launch of ‘Vibes’ AI Video Feed
Meta’s AI mobile app saw a sharp uptick in users and downloads after introducing its Vibes short‑form AI video feed, with daily active users rising to 2.7 million as of Oct. 17 from about 775,000 four weeks earlier and installs climbing to roughly 300,000 per day versus under 200,000 weeks prior (compared with just 4,000 daily downloads a year ago). The surge, which began after the Sept. 25 Vibes launch, coincides with heightened interest in AI video following OpenAI’s Sora release, though Similarweb cautions its data doesn’t prove causation and may not capture Meta’s internal promotions. The growth suggests Meta is gaining traction in the AI content market and may be siphoning engagement from rivals or capturing users locked out of invite‑only competitors, with potential implications for market share and competition among AI chat and video generators.
GPTQ Brings Int4 Layerwise Quantization to Keras 3
GPTQ is a post-training, weights-only quantization method that compresses a model to int4 layer by layer. For each layer, it uses a second-order method to update weights while minimizing the error on a calibration dataset. It comes built-in in Keras 3...
Why Explainability Is Key to Trusting AI
Excited to announce the next session in my Smart SaaSy Webinar Series — “The Black Box Problem: Do We Need to Understand How AI Works to Trust It?” I’ll be joined by Dr. Eva-Marie Müller-Stüler — global AI & data...
The Friend AI Pendant’s Creator Publicized a ‘Friend Protest’ in NYC
Friend AI’s founder publicly highlighted a spontaneous “Friend protest” in New York after the company’s $129 chatbot-necklace and its >$1 million subway ad push drew widespread ridicule and vandalism. Founded in 2023 and only shipping this summer, the wearable has...
FTC Removes Lina Khan-Era Posts About AI Risks and Open Source
The Federal Trade Commission quietly removed three Lina Khan‑era blog posts on open‑source AI and consumer harms—published July 2024, October 2023 and January 2025—according to Wired, part of a broader purge that previously included roughly 300 AI‑related posts. The deletions...
Kohler’s New Toilet Camera Provides Health Insights Based on Your Bathroom Breaks
Kohler introduced the Dekoda, a clamp-on toilet sensor that uses optical spectroscopy to analyze waste for hydration and gut-health indicators and to discreetly alert users to blood; the device is available for preorder at $599 with shipments starting Oct. 21....
These Nonprofits Lobbied to Regulate OpenAI — Then the Subpoenas Came
OpenAI has issued broad subpoenas to at least seven nonprofits that criticized its shift from nonprofit to commercial governance as part of its legal battle with Elon Musk, demanding extensive financial records, communications and documents related to OpenAI’s restructuring. Recipients...
Zocdoc CEO: ‘Dr. Google Is Going to Be Replaced by Dr. AI’
Zocdoc CEO Oliver Kharraz told TechFutures that AI chatbots are poised to replace “Dr. Google,” but Zocdoc expects to remain the underlying infrastructure that powers AI-driven scheduling and access to care. The company touts that a plurality of bookings occur...
5 Sensitive Details to Never Share with ChatGPT
ChatGPT Danger: 5 Things You Should Never Tell The AI Bot Protect yourself and your privacy. Here are five sensitive things you should avoid sharing with ChatGPT or any conversational AI. Read more 👉 https://lnkd.in/eD2ARRhq #ChatGPT #Privacy #AIUsage #BernardMarr
Import AI 432: AI Malware; Frankencomputing; and Poolside's Big Cluster
A Dreadnode proof‑of‑concept demonstrates AI malware that runs locally on on‑device LLMs (Phi‑3‑mini via ONNX), autonomously exploiting misconfigured Windows services to escalate privileges—flagging a nascent threat limited today to high‑end workstations and CoPilot+ PCs but with serious security implications as...
Celebrating Excellence: Bank Of America Wins Forrester’s 2025 Data & AI Impact Award
Bank of America won Forrester’s inaugural 2025 Data & AI Impact Award for an enterprise‑wide AI strategy that blends scale, trust and enablement to drive measurable business outcomes. The bank’s AI footprint includes 50 million users of its Erica virtual...
Napster Returns as an AI Companion for Your MacBook – Seriously
Napster has relaunched as a hardware and software AI company with Napster View, a $99 holographic screen that mounts to MacBooks (M1+) and surfaces a library of some 15,000 agentic AIs for tasks from design and wellbeing to business planning....
AI SEO, Claude Gets New Skills, and the UK’s Big AI Training Push
In this episode Paul and Martin cut through AI hype for marketers, emphasizing that generative search demands high‑quality, well‑structured content and strong distribution rather than chasing fleeting "AI SEO" tricks. They review Anthropic’s new Haiku 4.5 and Claude Skills, note Claude’s...
From Fine‑tuned CNNs to ChatGPT’s Pixel‑perfect Descriptions
It do be like that these days... 😅 I remember when we had to fine-tune an ImageNet-pre-trained CNN and preprocess the image to then send it and get a top-5 classification that was somewhat accurate... Now ChatGPT describes every pixel...
Should an AI Copy of You Help Decide if You Live or Die?
Researchers at the University of Washington are piloting research into AI “surrogates” that could one day help doctors and families make end‑of‑life decisions for incapacitated patients, though no hospital has yet deployed such systems. The project, led by resident fellow...
AI Could Predict Who Will Have a Heart Attack
Startups including Bunkerhill Health, Nanox.AI and HeartLung Technologies are applying AI to routine chest CT scans to automatically detect and quantify coronary artery calcium (CAC), potentially flagging millions of patients at elevated risk of heart attack who are currently missed....
Energy, AI, and Policy Converge at ADIPEC 2025
How do we power a world demanding more energy, intelligence, AND resilience – simultaneously? Nov 3-6, Abu Dhabi becomes the epicenter of this transformation at #ADIPEC2025 🌍⚡ Energy. Intelligence. Impact. Excited to see prominent US leaders joining: 🏛️ Doug Burgum, Secretary of the Interior...
2026 Robotics Trends: AI Goes Physical, Transforming Work
The 5 Robotics Trends In 2026 You Must Get Ready For Now In 2026, #artificialintelligence is moving from software into the physical world through humanoid #robots , #autonomous vehicles, and collaborative machines, transforming workplaces and cities. From #robotaxis operating in...
Wikipedia Says Traffic Is Falling Due to AI Search Summaries and Social Video
The Wikimedia Foundation reported an 8% year‑over‑year decline in human pageviews after updating its bot‑detection systems, revealing that unusually high traffic in May–June came largely from evasive bots. Foundation chief Marshall Miller attributed the drop to search engines increasingly delivering...
One Human's System Beats Billionaire AI Teams
Spent the entire day with @blevlabs. My first prompt in his system completely outclassed the results I got from Grok, Gemini, Claude, or ChatGPT. Breakthroughs are possible. From a 20-year-old who has been working on his system for seven years. Gave me hope...
Meta Bans Third‑party AI Chatbots on WhatsApp
No more #ChatGPT and #Perplexity on WhatsApp! Meta quietly updates its policy on third-party AI chatbots https://t.co/511MLKZ83e #WhatsApp #META #aichatbot #AI #artificial_intelligence @AlbertoEMachado @Eli_Krumova @postoff25 @Khulood_Almani @anand_narang @NutritiousMind @baski_LA @TanyaSinha_ @devaang @AlAmadi1 @jeancayeux @enilev @efipm @mvollmer1 @Nicochan33 @RagusoSergio @FrRonconi @Shi4Tech @sallyeaves...
AI's Rapid Evolution Sparks Bubble Debate and Nvidia Dominance
got to chat about the @stateofaireport on @tbpn last fri w/@johncoogan @jordihays, incl. bits on financing dynamics, how today's ai is magical vs. 12 years ago when i started, ai bubble (yes, no?), nvidia's dominance, ai search/agent experience, industrial policy...
More Compute Trumps All Other AI Strategies
Without checking, what is the message behind the "Bitter Lesson", in your opinion? (a) all other things being equal, using more compute is better. (b) more compute is better than all the other things put together.
Explore the Power of Mixture of Experts (MoE)
🔗 Mixture of Experts (MoE): https://t.co/3CGjgO4H9p https://t.co/QA12nBeW0i
Overreliance on AI Risks Undermining Critical Decision-Making
The danger of AI dependency https://t.co/HFtMCYQf9g #AI #GenAI #TechNews #Technology #healthcare #Innovation #Tech #ArtificialIntelligence
Internal AI May Replace Risky Open‑Source Contributions
Even though this particular example worked out as you'd expect today, Open Source dynamics will certainly change. Accepting and merging contributions is always a risk and has high cost, so trusting an internal AI system for minor codebase improvements may become...
Defining “Large” In Language Models: GPU Limits
how large the model should be to be called large language model? if it doesnt fit on home gpus, its large?
Avoid the RL Hammer: Choose the Right Tool
If you sub-optimally define any problem to require RL, when most can be solved with different approaches, then of course the RL hammer looks like the right solution! Rather than defending RL thru semantics, better would be to ask: how can...
Obsessed with the Fresh Claude Interface Design
I am so very obsessed with the new Claude layout. https://t.co/BdqmCg9BU2
Meta Gives Parents New Tools to Block Flirtatious AI
"Meta" Grants Parents Wider Powers to Protect Teens from "Flirtatious AI" https://t.co/LvkJaPspkc #ParentingTips #parenting #Meta #AI #artificial_intelligence @PawlowskiMario @chidambara09 @Ym78200 @CurieuxExplorer @efipm @bigfundu @sayedflah @Ronald_vanLoon @cyngn @belindabeibi @odisseiaalfa @DigitalColmer @MyCompanionsAI @KirkDBorne @patricegorissen @jeffkagan @EdwardKens50830 @enilev @insom_ai333 @ChrisCCrowley @sallyeaves @gurmeet_judge @VairagyaSadhana @andresvilarino...
Williamsburg Launches New AI Safety Company
the ai safety company of williamsburg https://t.co/EWwuDys8Ry
Rethinking the Meaning of Personal AI
Thinking about the definition of “personal ai.”
Skilled Trades Set to Profit From AI Boom
👀🚨🤷🏽♀️ AI isn’t just for coders. It’s creating a boom for builders. 🏗️ Nvidia’s Jensen Huang told CNBC that electricians, plumbers & carpenters are the big winners of the AI era, powering the data centres that fuel intelligence itself. Nvidia’s investing $100B+,...
Salesforce CEO: AI Lacks Soul, Just Cut 4,000 Jobs
After going for AI agents, laying off 4,000 employees; Salesforce CEO Marc Benioff says: AI doesn't have a soul, it is not... https://t.co/Pz3MRjV15F #Salesforce #artificial_intelligence #AI @lexfridman @KirkDBorne @Ronald_vanLoon @erikbryn @antgrasso @sallyeaves @Nicochan33 @HaroldSinnott @mvollmer1 @marcusborba @CatherineAdenle @vinod1975 @YuHelenYu @ingliguori @fogoros @ylecun...
AI Mimics Emotion, But True Empathy Remains Human
Why AI Will Never Truly Understand Your Feelings — And Why That Matters Empathy, nuance, and emotional intelligence remain deeply human traits. This article explores why AI can mimic emotion but can’t genuinely feel—and the implications for trust. Read more...
AI Shifts Power From FAANG to Emerging Players
From the #FAANG gang to #MANGO: how AI came to dominate US tech https://t.co/y2hDAPZx2o #AI #artificial_intelligence #UStech #technology #US @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...
Vibe Coding’s Hype Collapses: AI’s Messy Reality
The rise and fall of Vibe Coding: The reality of AI slop https://t.co/LQPaM6b4Jg #vibecoding #artificial_intelligence #AI #codingforlife #Coding @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...
Thinking Less Can Outperform Chain‑of‑Thought Prompts
Last week, I had the chance to attend the Conference on Language Modeling (COLM 2025) here in Montreal - a must-attend conference for anyone starting out in AI/LLM I was lucky to get an invite, it was awesome. Almost every session circled...
Ruben Hassid Releases Free Comprehensive AI Guides
Ruben Hassid (@RubenHssd) is one of the biggest AI voices on LinkedIn. He just shared all his AI guides in a single doc. And made it free! 🔥 → Great guides on prompting, LLMs, workflows, image models → Custom GPTs → His fave AI courses Really...
Free 36‑page Guide Mastering N8n Workflows
By far the best n8n guide I’ve seen. Nate’s worked with 1,000s of users and just wrapped everything he’s learned into this 36-page guide! 🔥 → Clear lessons on JSON, nodes and debugging → Cloud vs self-host setup → AI integrations & LLM chains Totally...
Low‑Quality Data Causes LLM Cognitive Decline and Safety Risks
🧠 Can LLMs Get “Brain Rot”? A new study says… yes. And the implications for AI safety are massive. When large language models are continually pretrained on low-quality, high-engagement web data — think clickbait, meme threads, and influencer chatter — they begin...
Distributed AI Community’s Aesthetic Inspires Across Organizations
The aesthetics and language of the distributed AI training / homebrew AI community are fascinating and motivating, and this is true across multiple organizations ranging from Prime Intellect to Nous to Exo. It's cool!
Mystery “World Models” Spark VC Excitement
World models: No one knows what it means, but it’s provocative. It gets the VCs going.
Reevaluating BERT: Still an LLM?
do we still consider bert as an llm?
Google Launches Free Course on Real-Time Voice Agents
Google just dropped an ace new course on building real-time voice agents using ADK + Gemini. You’ll learn how to build agents that can search, call APIs, remember context, and talk naturally. 100% free (link in 🧵↓) https://t.co/enAatsRTjV
New Humanoid Robot Promises Real Human Connection at Home
Yogi: New #humanoid robot capable of ‘genuine human connection’ could take over homes! https://t.co/r2EL55yYv4 #humanoidrobot #Robotics #Robots @PawlowskiMario @chidambara09 @Ym78200 @CurieuxExplorer @efipm @bigfundu @sayedflah @Ronald_vanLoon @cyngn @belindabeibi @odisseiaalfa @DigitalColmer @MyCompanionsAI @KirkDBorne @patricegorissen @jeffkagan @EdwardKens50830 @enilev @insom_ai333 @ChrisCCrowley @sallyeaves @gurmeet_judge @VairagyaSadhana @andresvilarino...
Credibility Will Dictate Visibility in the New Grok Algorithm
When Grok takes over the algorithm in a few weeks it will reward people it knows better. Or penalize them, it knows all my sins deeply. Either way it already has everything about me. First tip. It will know whether you...
Figma Uses AI to Turn Designs Into Code Instantly
LIVE FROM DREAMFORCE: Kris Rasmussen (@kris_rasmussen), CTO of @figma, sat down with us at @salesforce’s biggest conference of the year to share how Figma is bridging design, code, and AI. “We’re always trying to meet users where they are — building...
AI Remote Workers Need Context, Policies, Values, and Goals
What does an AI actually need to be a useful remote worker? And for folks who like to follow along with my AI predictions - what already exists today and what is coming? 1) CONTEXT (docs, email, PPT, meetings, Slack, SFDC,...
RL Fine‑tuning LLMs Caps at 61% Success Rate
It's hard to overstate how devastating this paper is, not only for reinforcement learning. They spent $4m of compute to find out that RL on LLMs basically taps out at 61% "asymptotic pass rate" (exact rate depends on context), but they...