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:
AI May Not Need Massive Training Data After All
Johns Hopkins researchers demonstrated that AI systems built with brain‑inspired architectures can exhibit human‑like neural activity without any pre‑training. By comparing untrained transformers, fully connected networks, and convolutional neural networks, they found that only the convolutional designs produced activity patterns matching human and primate brains. These untrained CNNs performed on par with conventional models that require millions of labeled images. The study suggests that architectural choices may be as critical as data volume for future AI development.
NVIDIA Brings Agents to Life with DGX Spark and Reachy Mini
At CES 2026 NVIDIA demonstrated how its DGX Spark platform can power a personal AI assistant built on the Reachy Mini robot. Using open‑source Nemotron 3 Nano for reasoning and Nemotron Nano 2 VL for vision, the demo combined NVIDIA’s NeMo Agent Toolkit with ElevenLabs TTS to...
When Execution Is Cheap, Judgment Becomes the Bottleneck
Execution is no longer the constraint. As building gets cheaper, judgment becomes the scarce resource. Models can generate options endlessly. What they don’t accumulate is lived consequence. They don’t carry the memory of failed bets, abandoned paths, or tradeoffs that only made...
Treat AI Kindly; Abuse Doesn't Improve Its Learning
AI is now learning. Like I showed you with @blevlabs’s cognitive architecture last year. It will remember when you are an asshole😜 Grok can already tell you when I was. But with continually learning systems they get smarter like a human learns. If you...
This Autonomous Solar-Powered Robot Costs 1p, Is Smaller than a Grain of Salt, and Communicates Through 'Wiggles of a Little...
Researchers at the University of Pennsylvania and the University of Michigan have unveiled the world’s smallest fully programmable autonomous robot, measuring roughly 200 µm × 300 µm × 50 µm—smaller than a grain of salt. The microrobot harvests solar energy via a 75‑nanowatt panel and propels itself...
DeepSeek OCR Cuts Token Usage 10×, Hits 97% Accuracy
ICYMI DeepSeek has just unveiled an OCR monster 🤯 DeepSeek-OCR is a 3B-parameter model that redefines document intelligence. It reaches 97% character-level accuracy with 10× input compression, preserving every detail. Most OCR systems require over 6,000 tokens per page. DeepSeek-OCR achieves the same...
Europe's New Autonomous Maritime Platform Ready by 2026
introducing a multi-use maritime system built to defend Europe: in just about a year, @polarmistai has gone from system design, data collection, and three scaled up platform builds for a high-speed autonomous maritime platform that can carry many payloads -...
SwitchBot Says Its Humanoid Household Robot Can Do Your Laundry
SwitchBot unveiled the Onero H1 at CES 2026, a humanoid‑style household robot that can perform chores such as loading a washing machine, folding clothes, and making coffee. The robot rides on a wheeled base, boasts 22 degrees of freedom, multiple cameras,...
Community Reachy Mini Apps: Wishlist and Gaps Identified
Today, I'm trying all the reachy mini apps already created by the community. Which one should I focus on or which ones are still to be built? Resharing my xmas wish list below: - "What is love" Reachy mini plays "what...
Data Buzzwords Mapped: Distinguish Mining, Analysis, Viz, Science
Data buzzwords ≠ same meaning. Here’s a clear map: 🔍 Data Mining = find patterns 📈 Data Analysis = interpret insights 📊 Data Viz = communicate visually 🧠 Data Science = knowledge + ML 🏢 Warehouse (structured) 🌊 Lake (raw at scale) 🏞️ Lakehouse (best of both) 🐊...
World ‘May Not Have Time’ to Prepare for AI Safety Risks, Says Leading Researcher
Leading AI‑safety expert David Dalrymple warned that the world may lack time to address safety risks from rapidly advancing AI systems. He highlighted a widening gap between government understanding and private‑sector capabilities, noting that within five years most valuable tasks...
2025: Massive Growth, New Courses, and Personal Milestones
My 2025 wrapped: - released our first ever course & product 🚀 - followed up with 3 more courses (and a 4th coming soon with a great friend, @pauliusztin_) - invited to NVIDIA GTC and briefly met Jensen + many amazing people - landed...
15-Step AI Learning Roadmap Simplifies Your Upskilling Journey
Learning AI can feel overwhelming — but when you break it down into clear, structured steps, the journey becomes achievable. This infographic captures a powerful 15-step roadmap, from foundations in math and programming to ML/DL fundamentals, NLP, RL, cloud deployment,...
16 Common ChatGPT Mistakes You’re Probably Making (And How to Fix Them)
The article outlines sixteen frequent mistakes users make with ChatGPT and offers practical fixes. It emphasizes that prompt clarity, context, constraints, and verification are essential for reliable outputs. It warns that data privacy and model selection are critical responsibilities for...
Work in the Age of Infinite Agents
The episode examines how AI agents, rather than just speeding up tasks, expand the scale of knowledge work, enabling organizations to operate beyond human rhythms, meetings, and bottlenecks. Drawing on essays by Ivan Zhao and Aaron Levie, the hosts argue...
Build Pillar Pages for Topical Authority and Rankings
If your traffic is dropping even though your SEO looks clean — this is probably why: Google’s AI isn’t just checking if you mention the keyword. It’s checking if you understand the topic. Here’s the fix: Build topical depth. Create...
Subtle Releases Ear Buds with Its Noise Cancelation Models
Voice AI startup Subtle unveiled $199 wireless earbuds that combine noise‑cancellation with real‑time transcription and AI chat capabilities. The devices ship with a year‑long iOS and Mac app subscription, allowing users to dictate in any application without touching a keyboard....
Anthropic President Daniela Amodei Says "the Exponential Continues Until It Doesn't"
Anthropic President Daniela Amodei told CNBC that AI’s rapid, exponential progress has repeatedly outpaced expectations, but she cautioned it may not continue indefinitely. She highlighted that even if model capabilities keep improving, corporate rollout faces “human reasons” such as lengthy...
Local Resistance Blocks $98 Billion in AI Data Center Projects Across Eleven US States
Tech firms planning AI data centers encountered significant pushback, with 20 projects valued at $98 billion blocked or delayed across eleven U.S. states between April and June. Residents cite higher electricity bills, water use, noise, and loss of farmland as primary...
The "Curse of Knowledge" Means Smarter AI Models Don't Understand Where Human Learners Struggle
A recent study highlights a fundamental blind spot in large language models: while they can solve exam questions that stump humans, they cannot identify why those questions are difficult. This "curse of knowledge" bias means AI systems lack insight into...
Open Data: Strategic Edge for Innovation and Competition
Beyond its public value, open data reveals itself as a strategic layer for modern enterprises, where its thoughtful integration fuels informed innovation and shapes a more collaborative, intelligence-driven approach to market competitiveness. #OpenData #BusinessIntelligence #AI
DeepSeek Reveals Signal Distortion Causes Deep Net Instability
1/11 One of the most important papers of 2025 is DeepSeek's "mHC: Manifold-Constrained Hyper-Connections" To understand it, let's start with a fundamental question. Why do very deep neural networks suddenly blow up during training? Short answer: signals get distorted as they...
Tech Giants Are Racing to Embed AI in Schools Around the Globe
Tech giants—including OpenAI, Microsoft, Google, and Elon Musk’s xAI—are rapidly deploying generative‑AI tools and training programs in schools across India, the UAE, Kazakhstan, El Salvador, the United States, Thailand, Estonia, Iceland and beyond. The initiatives aim to reach millions of students...
AI Accelerates Mediocre Ideas, Reveals True Judgment
AI changes how fast mediocre ideas get shipped and how quickly real judgment is exposed. There’s no hiding behind effort anymore.
Claude Code Completes a Year's Work in One Hour
Wow. Principal Engineer at Google reveals Claude Code built a year’s worth of work in just one hour, matching months of engineering effort by her team. https://t.co/QhPgnYwwgt
Surfshark VPN: 2025’s Milestones and the Roadmap for 2026
In 2025 Surfshark overhauled its core infrastructure, debuting the patented Everlink self‑healing VPN system and deploying 100 Gbps servers paired with FastTrack software that reportedly lifts speeds by up to 70 %. The upgrades were driven by rising AI‑powered threats and a...
Non‑technical Users Share Practical AI Code Tool Uses
if you’re not technical but use droid/claude code etc what do you use it for? how (be specific)
Classical 3
If you have a specific vision in mind - classical 3d to exert control, and generative ai to take it all the way continues to be an unbeatable combination https://t.co/bmicGSxtXg
Claude Launches Agent Plugin for Chrome—Risk Warning
The episode examines Anthropic's new Claude Agent plugin for Chrome, highlighting its broad tab manipulation and storage access capabilities that enable powerful agentic browsing but also introduce significant security risks. The host explains how the plugin's design allows stealthy data...
Outsourcing Thought to AI Erodes Your Judgment
The more you outsource thinking, the more expensive it becomes to get it back. The AI model will keep going day by day, happily producing plausible answers and polished reasoning. It will never say when you’re slowly losing the ability...
MIT Breakthrough Lets Rescue Robots Map Instantly
This is where #AI + #IoT + #robotics become mission-critical. Robots that understand their environment don’t just move faster, they save lives. MIT just unveiled a breakthrough that lets search-and-rescue robots rapidly map large, unpredictable spaces in real time. @MITAeroAstro @MITEngineering #CES2026 🔗...
Why the Nvidia-Groq Deal Is Adding Fuel to the Debate on AI Valuations
Nvidia announced a $20 billion deal to license technology, acquire assets and hire talent from AI chip startup Groq, making it one of the sector's largest transactions. The agreement has reignited debate over inflated AI valuations and the rise of circular...
AI’s Rapid Growth Outpaces Data Center Energy Capacity
🔥 What we know about energy use at U.S. data centers amid the AI boom 🔥 AI is scaling fast. Energy systems aren’t. Why this matters (and what’s next). ✅ Read my article: https://t.co/xQJPxsWnIu #AI #Sustainability #EnergyTransition...
AI Expands Productivity Frontier; New Skills Required
AI is expanding the productivity frontier. Realizing its benefits requires new skills and rethinking how people work together with intelligent machines. Source @McKinsey Link https://t.co/Ej5vRN7zb4 via @antgrasso #AI #AgenticAI #Robots https://t.co/76sesFnorZ
Why “Which API Do I Call?” Is the Wrong Question in the LLM Era
The article argues that the traditional question "which API do I call?" is being replaced by "what outcome am I trying to achieve?" Modern large language models enable this shift through the Model Context Protocol (MCP), which translates natural‑language intent...
Beyond Silicon: These Shape-Shifting Molecules Could Be the Future of AI Hardware
Researchers at the Indian Institute of Science have created molecular devices that can switch between memory, logic, selector, analog processing, and synaptic functions by tweaking ruthenium‑based complexes. Seventeen tailored complexes were synthesized, and subtle changes in ligands and ionic surroundings...
Invite: Suggest Topics for Deep AI Podcast
Doing a long, super-technical podcast on the state-of-the-art in AI. Let me know if you have question, topic suggestions. Everything from details of LLM training pipeline & architectures, to coding, robotics, scaling, compute, business, geopolitics, etc. Besides topics & questions......
AI Agents Transform Chatbots Into Goal‑Oriented Executors
AI Agent = AI that reasons + plans + uses tools + acts to hit a goal. Loop: goal → sense → plan → tool use → act → eval → memory → improve. Core parts: system prompt, tools/APIs, short- & long-term...
Even AI Has Trouble Figuring Out if Text Was Written by AI — Here's Why
AI‑generated text is spreading across education, advertising and other sectors, prompting a surge in detection tools. The article outlines three main approaches—learning‑based classifiers, statistical probability tests, and vendor‑provided watermarks—each with distinct trade‑offs. It highlights that detectors quickly become outdated as...
Accuracy Isn’t Enough: Prioritize Relevance, Grounding, Faithfulness
Accuracy is a terrible metric for LLMs. And it’s the reason many AI demos look great but fall apart in real usage. LLMs don’t usually fail by being wrong. They fail by being: irrelevant ungrounded confidently misleading An answer can be “accurate” in isolation and still be useless...
AI Is a Multi‑Layered Stack, Not Just ChatGPT
Most people see only the tip of the iceberg when it comes to AI — tools like ChatGPT, generative models, and digital assistants. But beneath the surface lies a far deeper ecosystem: Machine Learning, Deep Learning, Neural Networks, Computer Vision,...
ByteDance's StoryMem Gives AI Video Models a Memory so Characters Stop Shapeshifting Between Scenes
ByteDance and Nanyang Technological University introduced StoryMem, a system that stores selectively chosen key frames in a memory bank to keep AI‑generated videos visually consistent across multiple scenes. By feeding these reference frames into a LoRA‑adapted Wan2.2‑I2V model, the approach...
Proof of Authority Beats Great Content Every Time
Great content isn’t enough anymore. Average content from someone with authority will outperform amazing content from someone unknown almost every time. Humans trust authority and Google’s AI reflects that. It surfaces people who look credible because that reduces risk for...
Unmatched Cross‑platform UX Keeps Users Returning
ive cancelled my chatgpt subscription many time. unfortunately, there is no other app that matches the ux of chatgpt app on phone, web and mac. thats the only reason i keep coming back. google can do better. dunno why they...
AI Masters Human Tools Faster than We Build AI Tools
ai seems to have gotten better at using tools made for humans faster than we’ve been able to create tools specifically for them it’s unclear why you would build an MCP (helpful only for agents) instead of a CLI tool...
AI Transforms Energy Utilities with Optimized Operations
AI applications have the capacity to revolutionise the way we produce & consume energy Today AI is helping power utilities optimise operations like equipment maintenance & replacements, as well improving fault prevention & detection More 👉 https://t.co/j1qWzqgCEL @IEA #CES2026 #AI #IoT #5G
Confidential Computing Secures Data Across Cloud and AI
Confidential computing strengthens data protection by using secure enclaves, encryption in use, hardware isolation and trusted verification, supporting private processing across cloud and AI environments through shared standards and alignment. Microblog @antgrasso #CloudSecurity https://t.co/dM7JbmMHpL
LLMs Still Struggle with Reasoning, Knowledge, Understanding, Planning
Four important limitations of LLMs: 1. Reasoning 2. Knowledge or expertise 3. Understanding 4. Planning and execution Learn more: https://t.co/qs7PYuNc2I @mitsmr #CES2026 #AI #IoT #5G https://t.co/v9NlLkpAHe
Gemma AI Uncovers Promising Cancer Therapy Pathway
How a Gemma model helped discover a new potential cancer therapy pathway https://t.co/lMiONmkR5m @GoogleDeepMind #ai #iot #5g #agenticai #ces2026 #Healthcare https://t.co/Surif36HUQ
AI Demystified: From Neural Networks to Deep Learning
Peeling back the layers of AI! This infographic unravels the complex world from Neural Networks to Deep Learning. Dive deeper into AI with the insights from @ingliguori's 'The Digital Edge' 👉 https://t.co/Nrh6BBTRcF #ArtificialIntelligence #MachineLearning #DeepLearning https://t.co/2q6oTZuXqE