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:
Claude Suffered a 'Major Outage.' Anthropic Says It's Fixed.
Anthropic confirmed a major outage affecting its Claude chatbot and Claude Code tool on Tuesday morning, leaving thousands of users unable to access the services. The company’s status page listed the incident as a "major outage" and reported that a fix was applied roughly 90 minutes later, restoring normal success rates. This disruption follows a series of recent reliability issues for Claude, even as the AI assistant’s app briefly overtook ChatGPT in Apple’s store downloads. The episode underscores the growing pains of rapidly scaling AI platforms.
OpenAI's Newest Fellowship Includes up to $15,000 in AI Compute a Month
OpenAI announced a new AI safety fellowship that provides each fellow with roughly $15,000 worth of compute per month and a weekly stipend of $3,850, running from September 14, 2026 to February 5, 2027. The program aims to attract external...
Reinvent Buyer Personas And Journeys Using the Power Of AI
At Forrester’s B2B North America Summit, a roundtable will explore how AI can transform static buyer personas and journey maps into living, decision‑driving assets. The session will show how machine‑learning models detect behavior shifts, update assumptions in real time, and...
Why AI Shouldn’t Be Used Even to Decide ‘Simple’ Court Cases
Generative AI is rapidly entering courtrooms, with judges already using it for drafting summaries, translating documents, and locating precedents. Jurisdictions such as the UK have issued guidelines that restrict AI to preparatory work and forbid its use for core judicial...
Advanced Deep Learning Interview Questions #17 - The Per-Step Update Trap
In a DeepMind senior ML engineer interview, candidates are asked why a custom 1D convolutional layer fails to learn translation invariance despite correct forward and chain‑rule calculations. The hidden issue is neglecting to aggregate the gradients computed at each time...
I Tried Google Photos' New AI Enhance Tool: How It Crops, Relights, and Fixes Your Shots - Sometimes
Google Photos has launched AI Enhance, an one‑tap editing feature that automatically adjusts lighting, color, framing, and cropping. The tool, introduced last year, is now rolling out to every Android device worldwide. Users receive three AI‑generated versions of each photo and...
AI May Threaten Critical Thinking in the Workplace
A University of Bath report warns that AI adoption in the workplace threatens critical thinking and creativity, especially for knowledge that relies on hands‑on experience, culture, and analytical judgment. Researchers differentiate AI‑compatible encoded and embedded knowledge from AI‑incompatible embodied, encultured,...
Houston Methodist to Deploy Midstream Health’s Agentic AI Financial Action Platform Across Enterprise
Houston Methodist has partnered with Midstream Health to roll out its Agentic AI Financial Action Platform across the health system, beginning with supply‑chain operations. The AI agents ingest both structured ERP data and unstructured contract PDFs to automatically detect pricing...
Anthropic Locked In Google & Broadcom While Its Revenue Tripled
Anthropic announced a multi‑gigawatt agreement with Google and Broadcom to secure next‑generation TPU capacity for training its Claude models. The partnership comes as the company’s revenue run‑rate surged to $30 billion, up from $9 billion at the end of 2025. Anthropic’s enterprise‑focused...
Anthropic Plans $1B Venture with Major PE Firms to Sell and Deploy AI Tools Across Their Portfolio Companies
Anthropic is negotiating a $200 million investment as part of a broader $1 billion fundraising round with private‑equity giants General Atlantic, Blackstone and Hellman & Friedman. The capital will launch a new consulting arm that embeds the Claude AI suite into the...
Unified Workflows, Not More AI, Boost Speed
#AI is everywhere. So why does everything still feel slow? Because your data is scattered. Across tools. Across teams. More AI isn’t the fix—connection is. In this video, I break down why unified workflows are the real unlock. Thanks to @adapt for sponsoring this post—what...
Variance Bags $21.5m to Power AI Compliance Agents
Variance, a developer of agentic AI tools for risk and compliance, secured a $21.5 million Series A round led by Ten Eleven Ventures. The funding will accelerate its platform that automates KYC, KYB, AML, transaction monitoring and fraud detection for financial institutions...
B2B Buyers Prefer AI Search; Marketers Still Blind
94% of b2b buyers now use ai during their purchase process, per forrester's 2026 survey of 18,000 buyers. ai search traffic converts at 14.2% vs google organic at 2.8%. that is 5.1x, per averi's analysis of 680 million citations. 84% of b2b...
Titan's Banking AI Makes Compliance Auditable, Not Optional
Domain-Led AI for Banks: Titan's SLMs Aim to Make Compliance Auditable, Not Optional Titan launched banking-native small language models (SLMs) embedding banking logic + regulatory frameworks. Proprietary Banker Trust Index (BTI) + RAGAS benchmarks claim higher safety, reliability, supervisory alignment vs...
OpenAI and Spotify Leaders Back London-Based AI Agent Security Startup in $13M Seed Round
London‑based Trent AI emerged from stealth with a $13 million seed round led by LocalGlobe and Cambridge Innovation Capital, joined by executives from OpenAI, Spotify, Databricks and AWS. The startup offers the first multi‑agent security platform that continuously scans, assesses risk...
Enterprise AI Is Fragmented, Slowing Decision-Making
Every enterprise is “doing AI”… But almost none are actually operating with it. The real problem isn’t adoption. It’s fragmentation. AI is everywhere, yet decisions are still slow. Thanks to @Adapt for partnering with me on this, because this is exactly the problem they’re solving. Here’s...
Redefine Your Job as Outcomes, Not Tasks, to Leverage AI
Your job is not the tasks you do. This is a critical reframe that every knowledge worker must embrace in order to successfully adapt to AI. Especially true given that the current frontier models are mind-blowingly good and the next...
Tokenmaxxing: When Compute Activity Masquerades as Productivity
The article warns that many firms are treating raw token counts as a proxy for AI productivity, a practice the author dubs “tokenmaxxing.” By tracking every token generated by large language models, companies create a vanity metric that rewards higher...
Broadcom Signs Expanded Deals with Google and Anthropic, Giving the AI Startup Access to 3.5 Gigawatts of Compute Capacity
Broadcom will manufacture future generations of Google’s AI chips and has expanded its agreement with Anthropic, granting the startup access to roughly 3.5 gigawatts of Google Tensor Processing Unit compute, up from 1 GW. Anthropic disclosed annualized revenue now exceeds $30 billion,...
DBOS, Inc. Announces Technology Partnership with Databricks to Increase Agentic AI Reliability and Trust
DBOS, Inc. announced a technology partnership with Databricks to enhance the reliability and observability of agentic AI workflows. The collaboration integrates DBOS’s open‑source durable execution platform with Databricks’ Lakebase serverless Postgres, creating real‑time checkpoints that let AI agents resume automatically...
☕🤖Tutorial: Stop Scrolling Job Boards. Let Claude Find the Right Job For You
The post walks readers through building an AI‑powered Job Hunter using Claude, Anthropic’s large language model. In Phase 1 Claude evaluates a candidate’s résumé, highlights strengths, identifies gaps, and suggests optimal job titles and hidden‑gem roles. Phase 2 connects Claude to live...
The Imaginary of Informed Consent: Rethinking Approaches to Data Use for AI in Healthcare
The article examines how India’s Digital Personal Data Protection Act 2023 relies on informed consent to legitimize health data use for AI, but this model struggles with the complex, secondary purposes of AI training. It outlines three consent challenges: multiple...
Inside Intelligent Enterprises
Wipro and Intel have launched the WINGS.OTNxT.AI platform, an end‑to‑end managed service that unifies operational technology (OT) and Internet of Things (IoT) environments for manufacturers. The solution, already deployed by more than 40 customers, combines device inventory, secure networking, vulnerability...
Perplexity Launches Computer for Taxes to Draft, Review, and Optimize U.S. Federal Tax Returns Using Loadable AI Tax Modules
Perplexity has expanded its Computer AI platform with dedicated tax modules that can automatically draft U.S. federal tax returns on official IRS forms, review professionally prepared returns for errors, and create custom tools for complex scenarios such as rental portfolio...
Local AI Agent Delivers Real Work, Trust Over Brains
Meet QoderWork — a desktop AI agent that doesn't just chat, it actually does your work. Evan leads the QoderWork product at Qoder, a Singapore-based AI company. Which is a new generalized agentic service that lets you: ++ Open files on your...
OpenNebula Systems Launches OpenNebula 7.2 to Power Sovereign Clouds and AI Factories at Scale
OpenNebula Systems released OpenNebula 7.2, a minor update aimed at production‑scale sovereign clouds and AI factories. The version adds a gRPC‑based API, deeper NVIDIA integration, and hardware‑rooted security features such as confidential computing and virtual TPM. It also expands storage mobility...
Overwhelmed by Applications, Recruiters Turn to AI Pre-Screening Tools to Winnow Down Applications
Recruiters are drowning in applications as hiring slows, with 72% of global firms still reporting a talent shortage. Two‑thirds of talent‑acquisition professionals say they received more candidates per role last year, yet 46% say candidate quality has declined. In response,...
So… What Are We Doing with AI? Innovating in an Age of Caution
CEOs are under intense board pressure to justify AI spending, yet 56% report no meaningful financial returns and only 12% see both cost cuts and revenue gains. Confidence in revenue growth for 2026 has slipped to a five‑year low of...
Access, Not Model Limits, Bottlenecks AI Agents
The biggest misconception about AI agents right now is that they’re blocked by model capability. From what I’ve seen, the real bottleneck is access. Most real work lives behind logins, dashboards, and messy human interfaces, while agents are pushed through...
Asylon and Thrive Logic Bring Physical AI to Enterprise Perimeter Security
Asylon and Thrive Logic have partnered to embed physical AI into enterprise perimeter security, combining Asylon’s autonomous robotic patrols with Thrive Logic’s AI‑driven analytics and automated incident workflows. The integration streams video from mobile robots to the AI platform, which...
Sony to Acquire Machine Learning Company Cinemersive Labs
Sony Interactive Entertainment announced the acquisition of UK‑based machine‑learning firm Cinemersive Labs, a developer of six‑degrees‑of‑freedom 3D capture technology. The deal, disclosed on April 2, adds the startup’s portable rig and smartphone‑based generative‑AI capture system to Sony’s Visual Computing Group. Sony...
Human vs AI: Debates Shape RSAC 2026 Cybersecurity Trends
The RSAC 2026 conference opened with AI taking center stage, as vendors aggressively promote AI‑driven security solutions, including ambitious agentic AI that could augment or replace traditional security‑operations centers. Executives debated the scalability of the "human‑in‑the‑loop" model, with Vodafone’s CISO Emma Smith...
![[OFC 2026] Part 3 of 5: Packaging: Glass Substrates to 3D Hybrid Bonding](/cdn-cgi/image/width=1200,quality=75,format=auto,fit=cover/https://substackcdn.com/image/fetch/$s_!FG3y!,w_256,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F3201360a-f898-4593-8267-8cfc0766cc04_1280x1280.png)
[OFC 2026] Part 3 of 5: Packaging: Glass Substrates to 3D Hybrid Bonding
The OFC 2026 session highlighted that AI‑driven data centers are now limited by packaging rather than raw silicon speed. Co‑Packaged Optics (CPO) aims to overcome this bottleneck by integrating optical engines directly into switch packages, but the choice of substrate...
Data Dominion: How Zeta Global Cracked the AI Code for the Next Generation of Martech
Zeta Global, led by CEO David A. Steinberg, has positioned its AI‑first data platform as a core infrastructure for marketers, now serving 51% of the Fortune 100. The company launched Athena, a voice‑enabled AI copilot built with OpenAI, after proving that...
Bigeye Joins Snowflake-Led Open Semantic Interchange to Power Data and AI Interoperability
Bigeye announced its membership in Snowflake‑led Open Semantic Interchange (OSI), an open‑source effort to create a vendor‑neutral specification for semantic metadata. OSI seeks to unify fragmented data definitions so metrics stay consistent across dashboards, notebooks, and machine‑learning models. By joining,...
AlphaSense Scales Global Operations to Meet Surging Demand for Applied AI
AlphaSense is accelerating its global footprint by bolstering leadership in APAC and EMEA and expanding its multi‑language AI platform. The company added a SVP of International Corporate Sales for the UK and a VP of Corporate Sales for Singapore, targeting...
AI and Investing: David Trainer Presents at Harvard Business School
On April 8, 2026, Harvard Business School hosted a panel on AI’s impact on investing, featuring New Constructs CEO David Trainer, Hudson Labs’ Kris Bennatti, and Alumni Ventures’ Ron Levin, moderated by Assistant Professor Yuan Zou. Trainer outlined how AI...
AI Dismantles Agency Middlemen in Creator Economy
The creator economy just had its Uber moment. 30% agency cuts on a $5,000 brand deal = $1,500 gone to someone who forwarded three emails. An AI that knows your niche, pitches brands, and handles your inbox for free for a month...
Wolters Kluwer Accelerates AI Leadership with AI Center of Excellence and “FAB” Platform
Wolters Kluwer announced that its AI Center of Excellence, paired with the proprietary Foundation and Beyond (FAB) platform, is accelerating the delivery of trustworthy AI solutions across its five divisions. FAB is a model‑agnostic, enterprise‑grade enablement layer that standardizes tracing,...
Tenant Turner Releases AI Virtual Agent Beta to Automate Lead Conversations and Schedule Showings Faster
Tenant Turner has launched a beta version of its AI Virtual Agent, a conversational tool that automates routine prospect inquiries and pre‑qualification for rental listings. The agent operates 24/7 via a web widget and SMS, instantly answering common questions and...
Human Judgment Still Beats Overhyped AI Investing
"Should we outsource investment judgment to a machine?" While this feels like a new question, "quantamental" stock selection processes were the cutting edge of investment process in the 2010s. The argument was incredibly compelling conceptually: we can blend the best...
Perplexity Switches Wake Word to “Hey Perplexity” On Galaxy S26
Perplexity is changing its wake word from 'Hey Plex' to 'Hey Perplexity' for Samsung Galaxy S26 Changes in the Perplexity Android app appear to confirm work on the new wake word. ✅ Details - https://t.co/eVFc2SQVvx https://t.co/OxmPQs3yyf
Three Pillars Unlock Business Value of Agentic AI
What's The #Business Value Of #AgenticAI? Three Key Pillars For Scaling #AIAgents by Preetpal Singh @Forbes Learn more: https://t.co/kJ8518LdEZ #LLM #GenerativeAI #ArtificialIntelligence #MachineLearning https://t.co/OM0XLWOOPd
Harnesses: Software Scaffolds Managing Agent Context and Flow
Is a harness an agent? How's that different from the LLM? I liked this post from @rasbt that explains each—"a harness is software scaffold around an agent that manages context, tool use, prompts, state and control flow"—and what's in a coding...
AI‑RAN Embeds Intelligence in Networks for Instant Decisions
What is AI-RAN? It’s when AI is built directly into the mobile network—not just running in the cloud. That means faster decisions, lower latency, and smarter systems. I explain it simply here: https://t.co/qH8Xe7m0wg #MWC26 @SoftBank @SoftBank_RandD @ericsson @techiemats
Inside McClatchy Newsroom: How AI Transforms Journalism
Check out @CorbinBolies deep dive inside McClatchy newsroom and its use of AI, with lots of exclusive reporting:
Jovida Moves AI From Reactive to Life‑Changing
Most AI apps wait for your input. Jovida helps you actually change how you live. Here’s why this matters 👇 https://t.co/doSteuDVPi
AI Delivers Personalized Tutoring, but Adds Hidden Costs
How will AI change education? I am seeing glimpses in the way my kids use @sage_teacher: Each kid gets a personalized teacher, the AI guide keeps them engaged and challenged, and the parents have to pay extra to avoid the manipulation...
Turn Top LinkedIn Posts Into Endless Scheduled Variations
Remixing your greatest hits has never been easier. https://t.co/kiPTom1dLi's Chrome Extension: 1) Pulls your successful posts from LinkedIn 2) Analyzes the bangers 3) Gives you 4-5 different alternatives in your voice 4) Lets you schedule them to go live in the future Create once. Distribute...
Nutanix Agentic AI: Hyperconverged Power Unveiled
MyPoV: what’s behind the @nutanix Agentic AI solution? In the ask me anything session #NextConf Manosiz Bhattacharyya Thomas Cornely Lee Caswell Hyperconverged in an AI world https://t.co/XtuWzmUh4Q