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:
Meta’s AI Glasses Can Now Help You Hear Conversations Better
Meta has released software version 21 for its Ray‑Ban and Oakley Meta smart glasses, adding a conversation‑focus mode that amplifies the voice of the person you’re speaking with. The update also introduces a Spotify integration that can cue music based on what the wearer is looking at, such as an album cover or holiday decorations. The features debut in the U.S. and Canada for the hearing boost, while the Spotify function launches in multiple English‑speaking markets. Availability starts with Meta’s Early Access Program before a wider rollout.
Define Clear Questions First, Let AI Amplify Insight
Most organizations are sprinting toward “faster answers,” but almost no one stops to ask whether they posed a meaningful question in the first place. AI makes answers cheap. Priorities make answers useful. When your team treats questions as strategic assets,...
First University Talk: Turning AI Work Into Teach‑Ready Insight
I gave my first university talk this week at the University of San Diego. I was genuinely stressed. Not because the content was hard, but because it was the first time I had to turn what we do at Towards AI into...
Announcing The Forrester Wave™: AI Infrastructure Solutions, Q4 2025
The episode highlights the shift in AI infrastructure from experimental pilots to production‑scale workloads, emphasizing the need for integrated solutions that balance compute, networking, storage, and operations. It introduces the Forrester Wave™: AI Infrastructure Solutions, Q4 2025, which evaluates vendors on...
AI Agents Wear Multiple Masks—Ensure Ethics and Transparency
You compartmentalize your emotions, why not your AI? Emotionally partitioned AI agents are single AI systems that switch between different personas for regulators, customers, and internal teams, adjusting tone, priorities, and emotional style to match each audience. This makes interactions smoother...
ChatGPT Images Now Offers Faster, Precise Editing
The new ChatGPT Images is here. More precise edits, stronger instruction following, and up to 4x faster generation, while preserving details like lighting, composition, and likeness across edits. From quick fixes to full reimaginings, it’s easier to get exactly what you have...
Bridging the Divide: How Sanket Sable Is Architecting the Future of Privacy-First AI in Advertising
Sanket Sable, a senior technical solutions consultant at Google, has pioneered the "Shapeshifters" project, which integrates Gemini large language models with ad platforms to enable intent‑driven, hyper‑personalized advertising at scale. He also designed a DMA‑compliant measurement framework for Ads Data...
SAM Audio Isolates Any Sound From Mixtures via Prompts
🔉 Introducing SAM Audio, the first unified model that isolates any sound from complex audio mixtures using text, visual, or span prompts. We’re sharing SAM Audio with the community, along with a perception encoder model, benchmarks and research papers, to empower...
AI Agents Becoming Full‑Fledged Employees, Not Just Tools
Microsoft’s AI agents are getting… Workday HR profiles. Not metaphorically. Not conceptually. Actual HR profiles... alongside company email addresses, Teams access, and a seat next to human coworkers in the org chart. It’s the clearest signal yet that enterprises love...
Jailbreaking AGI: Pliny the Liberator & John V on AI Red Teaming, BT6, and the Future of AI Security
In this episode, Pliny the Liberator and John V discuss their radical approach to AI red‑teaming, emphasizing universal jailbreaks—skeleton‑key prompts that bypass guardrails across modalities—and the shortcomings of RLHF‑based safety as mere security theater. They detail hard vs. soft jailbreak...
Weight Decay Softly Prunes Noisy Features, Not Just Simplicity
Weight decay is usually presented as “encouraging simpler solutions”, but I tend to think that the real benefit is the soft pruning of noisy / unhelpful features. Without decay, a weight can random-walk to a large value even if the...
Agentic AI Turns Workflow Bottlenecks Into Speed
Complex deals move at the speed of their slowest workflow. One approval pauses another. One review delays the next step. Momentum fades quietly. Agentic AI introduces a new execution tempo when each step becomes computationally accessible. Velocity shifts from constraint...
Creating Psychological Safety in the AI Era
Enterprise AI adoption hinges on psychological safety, enabling staff to experiment without fear of reprisal. A MIT Technology Review Insights survey of 500 leaders shows 83% believe safety boosts AI project success, yet only 39% rate their organizations as very...
Why Physical AI Needed a Completely New Data Stack
In this episode, Lukas Biewald interviews a lineup of AI leaders—from Arvind Jain of Glean, who details the shift from enterprise search to agentic AI that grasps internal workflows, to Jarek Kutylowski of DeepL, who explains how a small team...
Game‑Trained AI Agents Pave Way for Smarter Robots
AI Agents Playing Video Games Will Transform Future Robots What gaming AI learns today could pave the way for smarter, more adaptive robots tomorrow — bridging virtual skills and real world robotics. Read more 👉 https://lnkd.in/d64Zp_VD #AI #Robotics #FutureTech #BernardMarr
From AI Agent Orchestrators to Deepfakes. The New Tech Order
The episode revisits the team’s bold January 2025 AI roadmap, evaluating how predictions about AI agent orchestrators, reasoning data, and deepfake threats have unfolded. Jordan Wilson highlights the rise of AI agents as a new employee class, the shift toward...
Zoom Says It Aced AI’s Hardest Exam. Critics Say It Copied Off Its Neighbors.
Zoom announced that its federated AI system achieved a 48.1% score on the Humanity's Last Exam, surpassing Google’s Gemini 3 Pro benchmark. The approach routes queries to multiple external models and selects the best output via a proprietary Z‑scorer. Critics...
With 91% Accuracy, Open Source Hindsight Agentic Memory Provides 20/20 Vision for AI Agents Stuck on Failing RAG
Vectorize.io’s open‑source Hindsight memory architecture outperforms traditional retrieval‑augmented generation (RAG) by organizing agent knowledge into four specialized networks. The system achieved a record 91.4% accuracy on the LongMemEval benchmark, dramatically boosting multi‑session recall, temporal reasoning, and knowledge‑update scores. Hindsight’s TEMPR...
Zencoder Drops Zenflow, a Free AI Orchestration Tool that Pits Claude Against OpenAI’s Models to Catch Coding Errors
Zencoder unveiled Zenflow, a free desktop AI orchestration tool that coordinates multiple AI agents—such as Claude and OpenAI models—to plan, implement, test, and review code in structured workflows. The platform replaces ad‑hoc prompting with repeatable sequences, spec‑driven development, multi‑agent verification,...
Only 6% Trust AI Agents, Yet 86% Will Increase Use
6% of business decision-makers trust AI agents while 86% plan to increase their agentic exposure — is this gap a good thing or a bad thing? Use this exact link to scoot past the paywall on Medium: https://lnkd.in/eZVfkhjJ #ai #aiagents...
Vertical AI Systems and Open Source Flexibility
SymphonyAI CTO Raj Shukla argues that targeting vertical AI applications—especially in manufacturing and supply‑chain—accelerates enterprise adoption. He notes that vertical AI firms achieve deeper market penetration because they solve well‑understood industry problems. Shukla also highlights open‑source models as a flexible,...
Vodafone's New Smart Home Router Can Even Tell You when Your Family Gets Home
Vodafone has introduced the Ultra Hub 7 router with a built‑in “Who’s Home” feature that uses network‑level detection to identify when household devices join or leave the Wi‑Fi. The tool, accessible via Vodafone’s app, provides a privacy‑friendly dashboard showing who is...
How the Tech Industry’s Embrace of Trump May Be Paying Dividends
Donald Trump’s executive order barring states from regulating AI has been hailed as a victory for Silicon Valley, rewarding years of lobbying by OpenAI, Microsoft, Nvidia and other tech giants. The order signals the administration’s preference for rapid AI growth...
Latent Labs Launches Latent-X2 AI Drug Design Model
Latent Labs unveiled its new AI model, Latent‑X2, which can generate drug‑like antibodies and macrocyclic peptides in a single design cycle, eliminating traditional iterative wet‑lab optimization. In validation tests the system produced hits on nine of 18 difficult targets, including...
Mining Business Learnings for AI Deployment
BHP is moving AI from pilot projects to an operational capability, embedding machine‑learning models directly into daily decision points across its mining sites. By targeting repeatable decisions—such as predictive maintenance and resource‑efficiency—the company assigned owners, KPIs and real‑time alerts, turning...
Why It’s Time to Reset Our Expectations for AI
The MIT Technology Review’s new "Hype Correction" series urges a reset of AI expectations after years of overstated promises. It highlights the gap between lofty claims—solving climate change, achieving human‑level intelligence, eliminating jobs—and the current reality of limited, sometimes flawed,...
Anthropomorphism in AI: Why We Treat AI Like Humans, and What That Says About Us
The article examines why professionals instinctively treat AI systems as if they were human teammates, linking the tendency to deep‑seated psychological needs for control, social connection, and pattern recognition. It cites research showing that increased interaction with conversational models amplifies...
BNP Paribas Introduces AI Tool for Investment Banking
BNP Paribas has launched an internal AI‑driven tool called IB Portal to streamline investment‑banking pitch preparation. The system searches the bank’s historical pitch decks and uses smart prompts to surface relevant slides, charts and analysis, cutting research time by days. IB Portal sits...
🧑🚀 UBI in the AI Age, U.S. Tech Dominance, & NVIDIA's Nemotron
Today NVIDIA unveiled the Nemotron 3 family of open AI models, featuring a hybrid mixture‑of‑experts design that promises higher throughput and lower inference costs, with the Nano version already available and larger variants slated for 2026. The company also signaled a...
6 Charts That Show The Big AI Funding Trends Of 2025
AI dominated venture capital in 2025, capturing $202.3 billion—about 50 % of all global funding. Foundation‑model firms alone raised $80 billion, representing 40 % of AI capital, while the United States accounted for $159 billion (79 %) of the sector’s investment. Private‑equity investors led $63 billion in...
Google's Free Guide Redefines Agent Quality Evaluation
Google just dropped one of the most practical guides on agent quality in production 🔥 It's a great 50-page deep dive on why traditional QA fails for agents and what to replace it with. 100 % free. Download link in 🧵↓ → The...
JPMorgan Chase AI Strategy: US$18B Bet Paying Off
JPMorgan Chase’s $18 billion annual tech budget is delivering 30‑40% year‑over‑year AI‑driven ROI growth, backed by a model‑agnostic LLM Suite used daily by 200,000 employees across 450+ production use cases. The platform, which won American Banker’s 2025 Innovation of the Year,...
What Is SpamGPT and How Does This Dark Web AI Tool Power Mass Phishing Operations?
SpamGPT is a dark‑web subscription service that provides AI‑driven tools for automating mass phishing campaigns. Priced at roughly $5,000 per year, it offers a full‑stack phishing management platform with dashboards, AI‑generated email content, SMTP/IMAP testing, and evasion features that can...
Stay Ahead: Real‑World Data Science Community & Insights
Data science doesn’t stand still and neither should your thinking. If you’re working in data science today, you’re navigating far more than models and code. You’re dealing with real decisions, AI governance, evolving techniques, and growing expectations from the business....
Strong Contractor Belief in AI for Industry-Wide Transformation
Research by Dodge Construction Network and CMiC shows 87% of contractors believe AI will meaningfully transform their businesses, yet current adoption remains low, with only about a third aware of AI tools. Automated proposal generation and progress‑tracking from site photos...
Six 2026 Education Trends: AI, New Literacies, Human Skills
The 6 Education Trends That Will Shape Learning And Skills In 2026 The year 2026 will bring major shifts in #education and #workforce #skills as #AI transforms how people learn, teach and prepare for fast-changing careers. This article explores the...
Create Fun Images Instantly with ChatGPT
Also a very fun way to use it to easily get fun images in ChatGPT: https://t.co/sWSedwAaII
OpenAI's 'Code Red' And the AIO Best-Of Listicle Hack
The episode examines OpenAI's rushed launch of GPT‑5.2, noting its strong benchmark scores but mixed real‑world performance and urging cautious testing against GPT‑5.1. It reveals how classic "best‑of" listicles are now a rapid shortcut to AI search visibility, though this...
Create Your Own Custom Board AI Today
Build your own @boardyai.
Benchmark Humans Against the Best Alternative, Not Averages
You should measure human capability on a task not in terms of "average human" or "random human", but in terms of your best alternative (to AI) if you were to hire a human to solve the task. Which isn't average...
Boost for Artists in AI Copyright Battle as only 3% Back UK Active Opt-Out Plan
A UK government consultation on AI‑trained data revealed that 95% of over 10,000 respondents want stronger copyright protection and mandatory licensing, while only 3% supported the government’s original active opt‑out scheme. High‑profile artists such as Elton John, Dua Lipa and Paul McCartney have...
AI's Potential to Create Original Scientific Discoveries
new on @airstreetpress: can ai generate new science? link below https://t.co/5egWvv3OPd
Open‑source Chatterbox Turbo Beats TTS Leaders on Latency
If you’re building voice agents, latency matters more than anything. For years, proprietary TTS dominated benchmarks. That’s changing. @ResembleAI’s Chatterbox Turbo is reshaping benchmarks: → Open-source → Crushes ElevenLabs & Cartesia on quality → Far lower latency 🧵↓ https://t.co/7xvyL9q65M
OpenAI's GPT Image 1.5 Challenges Google at Enterprise-Grade Visuals
OpenAI released GPT Image 1.5, upgrading ChatGPT’s image generation and API with the GPT‑5.2 engine. The update delivers sharper editing, consistent lighting, composition, and more reliable instruction following. It also improves text rendering and the ability to generate smaller faces...
Seeking Frontier AI Evaluation Expert for London Role
i'm looking for someone who has experience building evals for frontier ai systems for an amazing opportunity at a well-funded @airstreet company with strong traction ideally in london or wanting to move there ping here or dm if you're interested /...
Google's AI Loop Fuels Endless Startup Creation
Google is building an RL environment for infinite startups Use Gemini to ideate and build the product/service Use Google ad platform to sell it Repeat
DoorDash Rolls Out Zesty AI Social App
The episode breaks down DoorDash's new AI-powered social app, Zesty, which helps users discover restaurants and share dining experiences. It explains how Zesty's AI curates personalized recommendations, leverages community content, and integrates social features to make food discovery more interactive....
20 Essential AI Tools for 2025 Across All Domains
🔥 20 must-try AI tools for 2025. This stack covers everything: 🤖 chat & reasoning 🎨 image gen 🎬 video gen 🗣️ voice/music 🔎 AI search 🛠️ automation 🧠 AI agents My top 3 right now: ChatGPT, Perplexity, Midjourney. What are yours? 👇 #AI #AITools #GenerativeAI #Automation #AgenticAI
Voice Agents Require Far More than Simple Speech-to-Text
Most people think speech to text is just turn audio into words. Anyone who has built a real voice agent knows... that's the easy part. https://t.co/vNSn0Tum1y
OpenAI Slashes Microsoft Revenue Share, Adds $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...