Today's AI Pulse
Anthropic shuts down Claude Fable 5 and Mythos 5 after US export‑control warning
Anthropic disabled access to its Claude Fable 5 and Claude Mythos 5 models after the Trump administration, via the Commerce Department, warned that foreign nationals could exploit the systems. The move followed alerts from Amazon and other firms about a potential jailbreak that could aid cyber‑security attacks. Anthropic declined the administration’s request for a voluntary shutdown but complied with the order.
Also developing:
By the numbers: AMD acquires AI memory‑optimization pioneer Mext
The Subject Matter Expert Advantage in the AI Era
A senior solutions architect at Red Hat tested AI across three core tasks—analysis, coding, and content creation—to see if current models could replace his role. In comparative analyses, AI produced flawed or incomplete evaluations, requiring expert correction. In code generation, AI accelerated prototype work but failed to deliver production‑ready architecture without strong CI/CD and security guardrails. For writing, AI helped draft and edit but lacked the strategic depth and originality only a subject‑matter expert can provide. The overall finding is that AI serves as a powerful assistant, not a substitute for human expertise.
![[FounderCoHo @Stanford Event] Beyond Episodes: Infrastructure, Evaluation, and Benchmarking for Long-Running Agents](/cdn-cgi/image/width=1200,quality=75,format=auto,fit=cover/https://substackcdn.com/image/fetch/$s_!Dow6!,w_256,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F8fbbd31e-dbc8-4760-a64f-85c09e742bc8_558x558.png)
[FounderCoHo @Stanford Event] Beyond Episodes: Infrastructure, Evaluation, and Benchmarking for Long-Running Agents
FounderCoHo and Daytona are co‑hosting a high‑signal event at Stanford on April 29, 2026 to examine the emerging challenges of long‑running reinforcement‑learning agents. The discussion will cover the infrastructure required to sustain days‑long, stateful training, new evaluation frameworks for safety‑critical...
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,...
Anthropic's New Automation Layer: A PM's Setup Guide
Anthropic has introduced a new automation layer built around its Claude AI model, aimed at streamlining product managers' workflows. The guide frames PM work as two intelligence layers—personal knowledge and the team’s system‑wide context—and shows how Claude can feed both....
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...
Grok, Kling, Runway: AI Video’s Future Has Hollywood on the Outside Looking In
The AI video market has fractured after OpenAI shut down its Sora platform, leaving no single dominant player. Grok now captures the most traffic on social feeds, Runway is gaining traction in professional editing suites, and Kling is automating short‑form...
Gemini Agent Platform Tackles Enterprise Deployment Challenges
Google unveiled the Gemini Enterprise Agent Platform at Cloud Next ’26, extending its Vertex AI services with dedicated tools for building, securing, and orchestrating AI agents. The launch also introduced the Agentic Data Cloud, a cross‑cloud lakehouse that lets enterprises...
Disappointed by Compliance to AI Rules, Meity Proposes 'Continuous' Watermarks
India’s Ministry of Electronics and Information Technology (MeitY) is proposing an amendment to the 2021 IT Rules that would require a continuous watermark on AI‑generated content, replacing the current “clear and conspicuous” label requirement. The draft also removes the earlier...
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...
WiredCo Joins OpenAI’s ChatGPT Ads Pilot in Australia
WiredCo, an award‑winning brandformance agency, has joined OpenAI’s ChatGPT Ads pilot in Australia, bringing Boost Mobile and THE ICONIC on board as the first local brands. The agency is creating conversational ad units that sit natively within ChatGPT, targeting the...
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/
AI Now Draws Realistic Fake Graphs with GPT‑2
My most popular AI post was a bunch of made-up "graphs" four years ago. Now, the new GPT-2 image generator does it for real (though not perfect) Here's the famous AI task horizons graph with a touch of Basquiat, haunted by...
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...
LinkedIn’s New Tool Lets You Test the Outputs of Various AI Models
LinkedIn has launched Crosscheck, a new feature for Premium members in the United States that lets users test prompts across leading AI models from OpenAI, Anthropic, Google, Microsoft and others. The tool presents two anonymous model responses for each prompt...
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...
Still Little AI Return for UK Businesses, Accenture Survey Finds, and Many Reasons to Worry
Accenture’s "Generating Impact" survey reveals that UK executives remain skeptical about artificial intelligence, with 31% saying their operations would be unchanged if AI were removed tomorrow. Nearly half (46%) report little to no bottom‑line benefit from AI investments, and 25%...
Adobe Expands Agentic AI Ecosystem With Big Tech Partners
Adobe unveiled an agentic AI platform called Adobe C Enterprise, built on collaborations with OpenAI, Anthropic, Google, Nvidia, Amazon and other tech leaders. The solution promises end‑to‑end, one‑on‑one personalization at scale for both B2B and consumer users. Adobe estimates the initiative...
Exodigo Leadership Moves Fuel Engineering Expansion
Exodigo, an AI‑driven underground intelligence firm, announced three senior appointments this week. Rod Lacy, P.E., was promoted to Chief Engineering Officer for the United States, tasked with expanding the company’s engineering organization and AI‑powered design capabilities. Dillon Twombly joined as...
Microsoft Expands Partnership with NABTU to Deliver AI Training for Skilled Trades Workforce
Microsoft and North America’s Building Trades Unions (NABTU) have broadened their collaboration to deliver free AI literacy courses and an industry‑recognized credential to skilled‑trade workers across the United States and Canada. Building on a prior effort that trained 1,500 instructors,...
Gen AI Disruption Is Hitting Legal Research. Are Legacy Players Under Threat?
Generative AI tools are reshaping legal research, challenging entrenched providers such as Westlaw and LexisNexis. Startups like Casetext and Klarity are leveraging large language models to deliver instant case law summaries, citation analysis, and predictive outcomes. Legacy firms are responding...
ANU Spinout Raises $36 Million Series A to Make AI Chips
Australian National University spin‑out Syenta announced a AU$36 million (≈US$26 million) Series A round, led by Playground Global and the National Reconstruction Fund. The capital will accelerate commercialisation of its lithography‑free Localized Electrochemical Manufacturing (LEM) process, which promises 40% fewer steps and higher...
Roundtables: Unveiling The 10 Things That Matter in AI Right Now
MIT Technology Review’s Roundtables aired a special live session from EmTech AI, unveiling a curated list of the ten most consequential AI technologies, trends, and movements for 2026. Hosted by Grace Huckins with executive editors Amy Nordrum and Niall Firth, the...
Report: 65% of Pharma Supply Chain Leaders Have Limited Confidence in AI
A new WBR Insights LogiPharma Playbook surveyed 100 European pharma supply‑chain heads and found that while AI is moving from ambition to pilot projects, confidence remains low. Only 36% use AI beyond isolated experiments and 47% are still planning deployments....
Apple Has an Opportunity to Rediscover Humanity in Its Push Toward AI
Apple faces a pivotal moment as senior VP of hardware engineering John Ternus prepares to succeed Tim Cook as CEO. The transition offers a chance to reverse the company’s recent image erosion caused by aggressive App Store fees, privacy‑vs‑advertising contradictions,...
Unpatched AI Flaw Poses Risk to Banking Sector
Security firm OX Security uncovered a critical flaw in Anthropic's Model Context Protocol (MCP) that lets AI agents execute arbitrary host‑machine commands via the default "stdio" setup. Anthropic has declined to patch the underlying code, placing the onus on developers—particularly...
UF’s HiPerGator Drives FNN’s AI Wildfire Detection in Verizon Prize Recognition
A Gainesville startup founded by University of Florida alumni, FNN, secured a share of Verizon’s $1 million Disaster Resilience Prize. Leveraging UF’s HiPerGator supercomputer, FNN’s AI platform shrinks wildfire detection time from 24 hours to 40 seconds and adds real‑time smoke monitoring. The...
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...
KIME Review 2026: The AI Search Visibility Platform Built for Brands That Want to Win in LLM Search
KIME is a purpose‑built AI search visibility platform that monitors how Shopify brands appear in large‑language‑model responses such as ChatGPT and Perplexity. It provides prompt‑level share‑of‑voice, citation analytics, sentiment scores and an Action Centre that delivers weekly optimization recommendations. Pricing...
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?
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...
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
Anthropic Drops Claude Code, Targets $100+ Coding Users
Confirmed that Anthropic - as of now - has removed Claude Code from new Pro signups. This is what the pricing page looks like. Feels like Anthropic has the bet that those doing coding work will be willing and ready to...
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 Launches, Excels at People Photos
📸 GPT-Image-2 is now live on https://t.co/1vEawpI5vb And it's great for photos of people (too) https://t.co/SVezBonGIs
AI Centralization Threatens Traditional Professional Services Model
Decentralised AI use makes partner structure untenable. Centralised it, and it makes partners irrelevant. End could be nigh for the entire idea of ‘professional services’
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