Today's AI Pulse
OpenAI launches ChatGPT Images 2.0 with multilingual text and agentic reasoning
OpenAI has introduced 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 and create floor plans, maps and UI mock‑ups. The rollout expands OpenAI’s generative‑AI capabilities beyond text‑only outputs.
Also developing:
Tava Health Secures $40 Million Series C to Scale AI‑Powered Behavioral Health Platform
Tava Health announced a $40 million Series C round led by Centana Growth Partners, bringing its total funding to $73 million since 2020. The capital will fund three new AI‑enabled products—Overture, Symphony, and Tempo—and deepen support for clinicians, clinics, and health plans. The raise underscores investor confidence in digital behavioral health platforms that combine AI with practice‑management tools.
Snowflake Boosts AI Control Plane with New Intelligence and Cortex Code Features
Snowflake announced enhancements to Snowflake Intelligence and Cortex Code, positioning the platform as the control plane for the "agentic enterprise." The upgrades let business users and developers connect more data sources, enterprise apps and AI models, while preserving governance and...
Seeking AI Grants for South Berkeley Community Law Pilot
Hi my loves, quick return from twitter break A friend of mine is looking for assistance in locating AI-related grants that can be applied for to fund a pilot program in South Berkeley that serves as a public advocate and community-based...
General Catalyst Publishes First Quarterly Review, Highlights $43B Portfolio and $5B India Bet
General Catalyst released its first‑ever quarterly review, laying out the health of its $43 billion ecosystem, a $5 billion commitment to India, and a $380 billion valuation for Anthropic’s Series G round. The review also flags a broken software‑buyout model, discusses a potential Summa...
Nebius Secures $46 B in AI Cloud Contracts, Accelerating SaaS Expansion
Nebius Group announced a $46 billion backlog of AI cloud contracts, including a $19.4 billion deal with Microsoft and a $27 billion expanded agreement with Meta. The contracts, backed by a $2 billion Nvidia equity investment, could drive the company’s revenue to $3‑$3.4 billion in...
OpenAI Previews GPT‑Image 2 Ahead of Noon Livestream, Hinting at Next‑gen Multimodal AI
OpenAI unveiled GPT‑Image 2 on Monday, previewing the model on X and Reddit before a 12 p.m. PT livestream. The teaser highlighted expected gains in semantic coherence and prompt adherence, while investors pushed Nvidia shares up 0.8% on anticipated compute demand.
Your AI Feature Is Quietly Destroying Your Gross Margin
SaaS companies that embed AI often treat the new inference and infrastructure expenses as ordinary COGS, which can sharply erode gross margins. While classic SaaS targets 70‑80% margins, the ICONIQ 2026 State of AI survey shows AI‑enabled products anticipate only...
Biology’s Data Gap: Making Lab Results AI-Ready
Biology generates enormous amounts of data. Almost none of it is ready for AI. #SynBioBeta2026 is May 4-7th in San Jose, California, you can learn more about the conference and get your tickets here: https://t.co/8abYWJ1GbK Most experimental infrastructure was built for individual...
GPT‑4V Creates Flawless Model Table in One Shot
damn... GPT Image 2 model one-shotted the entire Venice model list into a graphical table, very few mistakes https://t.co/bHadlPm75U
Who Is Kaushal Sheth?
Kaushan Sheth, a three‑decade veteran of banking technology, co‑founded Sophos Solutions, which was later acquired by GFT Technologies where he now serves as chief technology officer. At GFT he leads efforts to embed agentic artificial intelligence into core financial infrastructure,...
Parameter Allocation Breakthrough Beats Pure Model Scaling
the efficiency gap here is wild. 770M pulling 1.3B performance means someone cracked parameter allocation, not just scaling tricks. that's the actual story, not the model size
Limited AI APIs Boost Business, Not Security
APIs and limited releases for AI models are not a safety policy, they’re a business model (which is totally ok as long as you’re transparent about it). Especially on cyber-security, they give a false impression of control and safety whereas in...
AI Hallucinations in Filing by a Top Law Firm
Sullivan & Cromwell filed a bankruptcy motion that included AI‑generated hallucinations, such as fabricated case citations, prompting a formal apology and admission that internal AI safeguards were not followed. The firm disclosed that mandatory training and review protocols failed to...
Claude Pro Loses Code Access as Token Costs Rise
Just wow. Tokens are getting more expensive. Claude Code is no longer included in Claude Pro plan. https://claude.com/pricing
Unauthorized Group Has Gained Access to Anthropic’s Exclusive Cyber Tool Mythos, Report Claims
Anthropic’s newly announced AI cybersecurity tool, Mythos, was reportedly accessed by an unauthorized group through a third‑party vendor. The breach was uncovered after members of a Discord channel posted screenshots and a live demo of the tool. Anthropic says no...
AI Model Beats General AI at Quantum Calibration
I fear the AI engineers are going to cook us, quantum. 13 institutions across industry, national labs, and academia, covering superconducting, neutral atom, and electron-on-helium hardware just released Ising Calibration 1 and QCalEval. It's a 35B parameter vision-language model (mixture of experts)...
Turning AI Ambition Into Results
A study by ST Telemedia Global Data Centres and Ecosystm surveyed over 600 professionals across nine Asian countries, revealing that only 16% of organisations have operational AI strategies backed by robust infrastructure and just 1% have fully embedded AI. Thailand...
Mozilla Fixes 271 Firefox Bugs Using Anthropic Mythos
Mozilla scanned the Firefox codebase with Anthropic Mythos has fixed 271 vulnerabilities identified during this initial evaluation. https://blog.mozilla.org/en/privacy-security/ai-security-zero-day-vulnerabilities/
OpenAI Launches ChatGPT Images 2.0, Codex Labs Developer Training Service
OpenAI unveiled ChatGPT Images 2.0, an upgraded image generator that can create visuals up to 2,000 pixels wide and supports new extreme aspect ratios. The model delivers sharper text rendering in multiple Asian scripts and improves small‑text, icon, and interface...
How AI-Powered Media Asset Management Is Reshaping Broadcast Workflows
Broadcast producers are grappling with unprecedented volumes of high‑resolution media, exposing the limits of legacy Media Asset Management (MAM) systems. AI‑driven tools now auto‑generate metadata, apply speech‑to‑text transcription, and recognize objects, shrinking asset search times from minutes to seconds. These...
Almost Half UK Firms Cite AI Skills Shortage as Barrier to Growth
A new Amazon Web Services study finds that 49% of UK organisations view a shortage of AI and digital skills as the primary obstacle to scaling artificial intelligence, even as adoption accelerates. The report highlights a widening gap between AI...
SpaceX Is Working with Cursor and Has an Option to Buy the Startup for $60B
SpaceX announced a partnership with Cursor to build a next‑generation coding and knowledge‑work AI, leveraging its Colossus supercomputer. The deal includes a $10 billion development payment option or a $60 billion acquisition right later this year. Cursor, valued at $29.3 billion after a...
More Foreign Visitors Meet Xi; Service Sector Boost; Countering Improper Foreign Conduct; Youth Unemployment; AI Chip Shortage; More Pressure on...
Chinese President Xi met Mozambique President Daniel Chapo and Lao envoy Saleumxay Kommasith, stressing China’s readiness to help Africa navigate the spillover of Middle East conflicts and upgrading bilateral ties with Mozambique to a community‑level partnership. Meanwhile, former U.S. President Donald Trump...
A Student's Path to Publishing
Alessa Carbo, a junior computer‑science student at Johns Hopkins, co‑authored a paper on AI‑powered translation of sign‑language videos that was accepted at the 2025 EMNLP conference. The work, developed during the Frederick Jelinek Summer Workshop, uses a custom vision‑language model...
Prepare for Agentic AI: When Agents Talk to Agents
The question isn't if you'll use Agentic AI—it’s how you’ll handle it when your customers’ AI agents start talking to your AI agents. 🤖💬 91% of leaders are already seeing AI-driven crime. The "manual" era is over. Join @twifintech x @Alloy on...
Chegg First to Crumble; Which Large‑Cap Faces AI?
Chegg. The official first company to get decimated by AI. What large cap stock do you think has the ability to have this happen due to AI?
UnitedHealth Group Investing Billions in AI
UnitedHealth Group announced a $1.5 billion investment in artificial intelligence to modernize its internal operations and launch commercial AI products. The spend targets member and patient engagement tools, workflow automation, and productivity gains across its health‑care and insurance divisions. In Q1...
The Oreo Cookie Method: How to Get the Best of Human Expertise and AI in Contract Review
The Oreo Cookie Method is a five‑step, tool‑agnostic workflow that places two attorney‑led reviews around AI‑driven redlining. By having lawyers set context, direct AI playbooks, run general reviews, and perform targeted clean‑ups, the process captures AI speed while preserving human...
Field Notes From Transform 2026: What Leaders Are Figuring Out About AI
At Transform 2026, HR leaders confronted the gap between AI hype and day‑to‑day execution, asking what concrete changes their teams should make this week. The conference highlighted that “AI fluency” is often a vague mandate without clear expectations, leaving managers...
Anthropic Likely a Short‑term Edge, Not Forever
Shame that these companies aren’t publicly traded. It would be a good voting machine for which AI companies are going to win. There is a hypothesis being tested: “Is Anthropic a temporary winner with a slightly better coding model?” Or “is...
Create Classroom Infographics in Minutes Using AI
📊 Have an infographic idea but not sure where to start? 🤖 In this new blog post, I’m sharing how AI can help you create classroom infographics in just a few minutes. 🔗 I also included examples and prompt ideas you can...
This AI Warning Is A Myth; The Danger Is Not...
A new study by Carnegie Mellon, Oxford, MIT and UCLA finds that brief AI assistance erodes people’s willingness to tackle problems, even after the tool is removed. Participants using AI skipped nearly twice as many questions as a control group,...
GenAI Cuts Data Creation Cost Like Internet Cut Transport
the internet collapsed the cost to move the bits. GenAI collapsed the cost to make the bits. -@ShapiroDoug
AI Model Evolution Surpasses LLMs in Cambrian Explosion
We're seeing a massive Cambrian explosion of AI model designs and architectures. In the same way when I started building agents a year + ago, the evolution of models is even more profound. Looking under the hood, MANY new models aren't LLMs,...
Plaud NotePin: Smart Wearable AI Boosts Meeting Notes
I review the Plaud NotePin Wearable AI Note Taker device and find that there's a lot to like with this smart little gizmo. It's particularly helpful for meetings and lectures, as I demonstrate in this detailed video: https://t.co/Q45YtD9jZc #ai #aiWearable...
GPT Image 2.0 Outshines 1.5 in Visual Tests
GPT image 1.5 VS GPT image 2.0 Provider = @AskVenice Platform = CardLab by @Cardify_Club Same exact prompt (i have 5 templates for A/B testing) which do you prefer? https://t.co/2HEvIbEi6K
2026 Advances: Agent Harnesses, Sub‑Agents, LLM Fine‑Tuning
Seroter Daily Reading List – April 21, 2026 (#768): Today’s links look at agent harness engineering, working with sub agents, and how to fine-tune LLMs in 2026. https://t.co/TYPQE7sm4z
Tight Controls Needed to Prevent Mythos Abuse
A lab leak on their hands here. We don't want adversaries using Mythos to expedite the drop of 0days and other malicious AI-powered attacks. Without tight controls, we're may sadly see Mythos used for evil before we get to see everything...
BlackRock Launches RockAI, No‑code AI for 5,000 Developers
JUST IN: BlackRock rolls out "RockAI" to 5,000 in-house developers, a no-code platform for building custom AI agents
AI Transforms Governance of Enterprise Unstructured Data
"Unstructured data now makes up the vast majority of enterprise information, and AI is redefining how organizations bring control, accessibility, and security to it." #DataGovernance #AI https://t.co/PYomJYHDkY
More AI Usage Beats Perfect AI Strategy
It doesn’t matter what you do with AI, just do more of it. 😅 https://t.co/eZc4HEI1d8
Adobe AI Consumption Jumps 45% Quarter‑over‑Quarter
$ADBE shares its AI growth metrics. The 45% increase in generative AI consumption Q/Q signals enterprises are using more Adobe AI, not less, despite the Claude kills everything narrative. 👀 https://t.co/BDWlQPDn8z
AI-Driven Pushpaganda Tricks Users with Malicious Notifications
Pushpaganda uses AI-generated content and deceptive tactics to trick users into enabling notifications that deliver scams and evade traditional security defenses. https://t.co/x8sjalWKPR
SAP's AI Teams Blend Support with Value‑Engineering Sales
SAP's AI adoption teams: genuine support or a sales tactic in disguise? For complex products like SAP, this 'value engineering' approach helps customers navigate features and unlock more value, especially for underutilized legacy systems. #SAP #AI #CustomerSuccess https://t.co/nfoo4U8z3w
Replit CTO Explains Trusted Platform for Vibecoding Risks
Okay so we all now know vibecoding is a massive opportunity AND a massive risk. Now the question is what platform can you trust and why? Replit CTO lays it out ⬇️
AI Integrates Omics and Lifestyle to Forecast Health Outcomes
A multimodal AI agentic model that integrates electronic medical records, lifestyle, layers of biologic omics data to predict health outcomes and, with perturbations, "what if" scenarios a person improved lifestyle or took a medication @Cell_Metabolism https://t.co/qgVomjkipS https://t.co/JDaeYv0JGu
LLMs Need a Full Operating System, Not QBasic
We really are still in the QBasic era of using LLMs. They really need a true Operating System. The beginnings of this is the LLM Harness, but there's so much more to be done.
MIT Study Reveals Vast Ecosystem of Niche AI Agents
A handful of agent platforms hog the headlines, but many function-specific agents are available to developers and users. MIT's latest study explores the broader agentic ecosystem. https://t.co/7bRibPFsRA via @ZDNET
95% of Financial Firms Struggle to Build Effective AI
From FICO, not MIT: Nearly 95% of organizations are having a difficult time developing AI systems that serve the needs they were designed to address. The study is based on 254 C-suite leaders in financial enterprises around the world. via...
AI Can Enable Technofascism, Already Under Construction
AI does not inevitably lead to technofascism, "but it can enable it". #Palantir’s post "offers us a glimpse of the technofascist trajectory: not as a distant possibility, but as a world already under construction". Great piece by @MCoeckelbergh ⬇️ https://t.co/NZs6TkA5Gf