Good morning, AI enthusiasts. Anthropic just took the AI consciousness debate from science fiction to serious research — launching a new program to develop frameworks for assessing potential model welfare.

With their own researcher estimating a 15% chance that models are already conscious, are we nearing the existential debate on whether digital minds deserve ethical treatment?

In today’s AI rundown:

• Anthropic’s new research explores AI welfare

• Adobe’s new Firefly models, AI integrations

• Turn your terminal into an AI coding assistant

• Google DeepMind expands Music AI Sandbox

• 4 new AI tools & 4 job opportunities

LATEST DEVELOPMENTS

ANTHROPIC

🤖 Anthropic’s new research explores AI welfare

Image source: GPT-4o / The Rundown

The Rundown: Anthropic just launched a new research program dedicated to “model welfare,” exploring the complex ethical questions around whether future AI systems might gain consciousness or deserve moral consideration in the future.

The details:

• Research areas include developing frameworks to assess consciousness, studying indicators of AI preferences and distress, and exploring interventions.

• Anthropic hired its first AI welfare researcher, Kyle Fish, in 2024 to explore consciousness in AI — who estimates a 15% chance models are conscious.

• The initiative follows increasing AI capabilities and a recent report (co-authored by Fish) suggesting AI consciousness is a near-term possibility.

• Anthropic emphasized deep uncertainty around these questions, noting no scientific consensus on whether current or future systems could be conscious.

Why it matters: Sam Altman previously likened AI to a form of alien intelligence. Soon, these models may reach a level that changes how we understand consciousness and ethics about them. We’ll likely see a polarizing divide—especially since there’s no threshold for when an AI could be considered “conscious” or deserving of rights.

TOGETHER WITH INNOVATING WITH AI

💼 Start your career as an AI Consultant

The Rundown: Innovating with AI’s new program, AI Consultancy Project, equips AI enthusiasts with all the resources to capitalize on the rapidly growing AI consulting market – which is set to 8x to $54.7B by 2032.

The program offers:

• Tools and framework to find clients and deliver top-notch services

• A 6-month roadmap to build a 6-figure AI consulting business

• Student landing their first AI client in as little as 3 days

Click here to request early access to The AI Consultancy Project.

ADOBE

🎨 Adobe’s new Firefly models, AI integrations

Image source: Adobe

The Rundown: Adobe just launched a major expansion of its Firefly AI platform at its MAX London event, introducing two new powerful image generation models, third-party integrations, a new collaborative workspace, and an upcoming mobile app.

The details:

• The new Firefly Image Model 4 and 4 Ultra boost generation quality, realism, control, and speed, while supporting up to 2K resolution outputs.

• Firefly's web app now offers access to third-party models like OpenAI's GPT ImageGen, Google's Imagen 3 and Veo 2, and Black Forest Labs’ Flux 1.1 Pro.

• Firefly’s text-to-video capabilities are now out of beta, alongside the official release of its text-to-vector model.

• Adobe also launched Firefly Boards in beta for collaborative AI moodboarding and announced the upcoming release of a new Firefly mobile app.

• Adobe’s models are all commercially safe and IP-friendly, with a new Content Authenticity allowing users to easily apply AI-identifying metadata to work.

Why it matters: OpenAI’s recent image generator and other rivals have shaken up creative workflows, but Adobe’s IP-safe focus and the addition of competing models into Firefly allow professionals to remain in their established suite of tools — keeping users in the ecosystem while still having flexibility for other model strengths.

AI TRAINING

🤖 Turn your terminal into an AI coding assistant

The Rundown: In this tutorial, you will learn how to install and use OpenAI’s new Codex CLI coding agent that runs in your terminal, letting you explain, modify, and create code using natural language commands.

Step-by-step:

1. Make sure Node.js and npm are installed on your system.

2. Install Codex typing npm install -g @openai/codex in your terminal and set your API key using export OPENAI_API_KEY="your-key-here".

3. Start an interactive session with codex or run commands directly like codex "explain this function".

4. Choose your comfort level with any of the three approval modes, e.g., suggest, auto-edit, or full-auto.

Pro tip: Always run it in a Git-tracked directory so you can easily review and revert changes if needed. For more info, here is the GitHub repository.

PRESENTED BY IMAGINE AI LIVE

💨 Fast-track your enterprise’s AI journey

The Rundown: IMAGINE AI LIVE '25 gives your enterprise direct access to the AI pioneers most companies can't reach, with speakers like Bindu Reddy, Dan Siroker, and Nathan Labenz compressing years of learning into just three days.

Meet the AI experts on May 28-30 at the Fontainebleau Las Vegas and:

• Bypass months of costly trial-and-error with frameworks built for enterprise scale

• Connect with leaders who've successfully embedded AI across entire organizations

• Get actionable roadmaps that translate cutting-edge capabilities into business impact

Accelerate your AI transformation with code AISPEAKERS200 to save $200 when you register by April 25th —  limited VIP passes are still available.

GOOGLE DEEPMIND

🎹 Google DeepMind expands Music AI Sandbox

Image source: Google DeepMind

The Rundown: Google DeepMind just released new upgrades to its Music AI Sandbox, introducing its new Lyria 2 music generation model alongside new creation and editing features for professional musicians.

The details:

• The platform’s new “Create,” “Extend,” and “Edit” features allow musicians to generate tracks, continue musical ideas, and transform clips via text prompts.

• The tools are powered by the upgraded Lyria 2 model, which features higher-fidelity, professional-grade audio generation compared to previous versions.

• DeepMind also unveiled Lyria RealTime, a version of the model enabling interactive, real-time music creation and control by blending styles on the fly.

• Access to the experimental Music AI Sandbox is expanding to more musicians, songwriters, and producers in the U.S. for broader feedback and exploration.

Why it matters: Google is targeting professional musicians, positioning Lyria 2 and the Sandbox as co-creation partners rather than just novelty music generators. The creative landscape for musicians is being reshaped by AI, like every other medium, but these tools are a big step in normalizing its currently polarizing use in the industry.

QUICK HITS

🛠️ Trending AI Tools

• 🔍 Dropbox Dash – AI universal search and knowledge management that will find every doc, video, image, or teammate across apps and turn content into first drafts, fast*

• 🎨 gpt-image-1 — OpenAI’s advanced image generation, now available via API

• 🤖 Researcher & Analyst - Copilot agents for research and data science tasks

• 🎆 Seedream 3.0 - Dreamina’s new high-level text-to-image model

*Sponsored Listing

💼 AI Job Opportunities

• 🧠 Deepmind - Research Scientist

• 🛠️ OpenAI - NOC Technician

• 🌍 Scale AI - Strategic Projects Lead

• 📊 Perplexity AI - Revenue Operations Analyst

📰 Everything else in AI today

OpenAI reportedly plans to release an open-source reasoning model this summer that surpasses other open-source rivals on benchmarks and has a permissive usage license.

Tavus launched Hummingbird-0, a new SOTA lip-sync model that scores top marks in realism, accuracy, and identity preservation.

U.S. President Donald Trump signed an executive order establishing an AI Education Task Force and Presidential AI Challenge, aiming to integrate AI across K-12 classrooms.

Loveable unveiled Loveable 2.0, a new version of its app-building platform featuring
“multiplayer” workspaces, an upgraded chat mode agent, an updated UI, and more.

Grammy winner Imogen Heap released five AI "stylefilters" on the music platform, Jen, allowing users to generate new instrumental tracks inspired by her songs.

Higgsfield AI introduced a new Turbo model for faster and cheaper AI video generations, alongside seven new motion styles for additional camera control.

COMMUNITY

🎥 Join our next live workshop

Join our next workshop on Monday, April 28th at 3 PM EST with Ellie Jacobs and Noam Markose from LTX Studio. In this live session, you’ll learn how to bring your AI-generated storyboards to life using LTX Studio’s powerful new timeline editor — no editing experience needed.

RSVP here. Not a member? Join The Rundown University on a 14-day free trial.

🤝 Share The Rundown, get rewards

We’ll always keep this newsletter 100% free. To support our work, consider sharing The Rundown with your friends, and we’ll send you more free goodies.

See you soon,

Rowan, Joey, Zach, Alvaro, and Jason—The Rundown’s editorial team

Good morning, robotics enthusiasts. San Francisco-based Physical Intelligence just gave a glimpse of how its decentralized AI model can enable robots to clean homes they have never seen before.

Unlike most bots that rely on environments similar to their training data, Physical Intelligence is thinking outside the box—are we finally approaching adaptable robots that can handle real homes without rigid programming?

In today’s robotics rundown:

• Physical Intelligence’s novel robot brains

• U.S.-China trade war hits Tesla Optimus

• A rice-sized bot for precision brain surgery

• This worm-shaped bot jumps 10 feet high

• Quick hits on other robotics news

LATEST DEVELOPMENTS

PHYSICAL INTELLIGENCE

🧠 Physical Intelligence’s novel robot brains

Image source: Physical Intelligence/YouTube

The Rundown: San Francisco-based Physical Intelligence just introduced π-0.5, a robotic foundation model designed to perform household chores — like wiping up spills and making beds — in never-before-seen environments.

The details:

• Built on a VLA model, π-0.5 processes high-level instructions in natural language and translates them into precise low-level motor actions.

• The model can accept a range of commands, from general instructions like “clean the bedroom” to more specific ones like “put that dish in the sink.”

• It has been trained with a mix of data sources, which enables it to understand the physical execution of tasks as well as their semantic context.

• The task is executed with a network of 100 decentralized “π-nodes” distributed throughout the robots’ bodies, with each node functioning as a mini-brain.

Why it matters: Similar to Google DeepMind’s Gemini, π-0.5 uses embodied reasoning and generalization to enable robots to perform complex tasks in unknown, real-world environments. While the AI isn’t perfect, it is a big step toward robots performing useful work in real homes with minimal manual programming.

TESLA

🤖 U.S.-China trade war hits Tesla Optimus

Image source: Tesla

The Rundown: Elon Musk announced on Tuesday that Tesla’s production of Optimus humanoids has been disrupted by China’s recent export restrictions on rare earth magnets, a key component in the robots’ electric motors.

The details:

• China, which produces nearly all of the world’s heavy rare earth metals and 90% of the magnets made from them, halted exports in response to U.S. tariffs.

• Exporters must now obtain licenses from China’s Ministry of Commerce to ship rare earth magnets, which reportedly can take weeks or even months.

• According to Musk, China is seeking assurances from Tesla that the magnets will not be used for military purposes and only for humanoid robots.

• Last month, Tesla completed a new pilot production line for Optimus at the Fremont Factory, with “legions” of robots expected soon.

Why it matters: The ongoing trade war is fueling concerns about global shortages of critical minerals from China — posing risks for Tesla and other tech companies building EVs, jets, and electronics. Still, Elon Musk says Tesla remains on track to produce thousands of Optimus robots this year.

ROBEAUTÉ

 🍚 A rice-sized bot for precision brain surgery

Image source: Robeauté

The Rundown: French startup Robeauté just secured €27M ($30M) to advance the development of its neurosurgical microbot—a device, no larger than a grain of rice, engineered to navigate the intricate pathways of the human brain.

The details:

• The microbot is injected through a 4-mm opening in the skull and can navigate neural pathways using its own engine, propulsion, and steering system.

• Once in, it can perform pre-programmed tasks such as biopsies, drug delivery, electrode implantation, and real-time data collection.

• Surgeons can monitor and control the robot’s movement in real time using an ultrasonic GPS placed on the patient's head.

• Once the job is done, they can trigger the bot to retrace its path and exit the brain safely, minimizing trauma to healthy tissue.

Why it matters: The microbot’s first human trials involving advanced biopsies of brain tumors are planned for 2026; future uses include targeted therapy delivery. Alongside U.S. rivals like Bionaut Labs, Robeauté’s microbot promises to be a transformative leap in minimally invasive brain surgery and personalized neurological care.

GEORGIA TECH

🪱 This worm-shaped robot jumps 10 feet high

Image source: Georgia Tech

The Rundown: Georgia Tech engineers just developed a 5-inch soft robot, inspired by parasitic worms, that can leap up to 10 feet — or roughly the height of a basketball hoop — despite having no legs.

The details:

• Inspired by the remarkable jumping mechanics of a nematode worm, the robot is made from a silicone rod reinforced with a carbon-fiber spine.

• Nematode worms bend and kink their bodies to store and rapidly release elastic energy, allowing them to leap up to 20 times their body length.

• The robot mimics these movements, using reversible kink instability to accumulate and unleash energy in just a tenth of a millisecond.

• Detailed in Science Robotics, the robot has potential applications in search and rescue or environments where wheeled or legged robots might struggle.

Why it matters: The soft robot leverages "mechanical intelligence" to solve complex movement challenges without relying on heavy computation or rigid components. While still in early phases, this innovation can pave the way for more robust and versatile soft robots targeting a wide range of inventive use cases.

QUICK HITS

📰 Everything else in robotics today

Chinese robotics powerhouse Unitree opened a 107K-square-foot Hangzhou factory to build humanoids as part of its massive three to five-year expansion plan.

Chinese automaker XPeng unveiled its first AI humanoid, IRON, at the Shanghai Auto Show, where it walked fluidly and waved to the crowd.

U.S. engineers created a material that can morph into new shapes and follow electromagnetic commands like a remotely controlled robot, even without motors.

MIT developed a new stamping technique that enables engineered muscles to contract in multiple directions, like the iris of an eye, to power future biohybrid robots.

Ukraine is set to deploy 15,000 AI-powered drones and ground robots to the frontlines this year as the war with Russia intensifies.

San Francisco-based Reborn, which is focused on building an open ecosystem for AGI, announced a strategic partnership with China’s Booster Robotics.

Autonomous vehicle startup Scout AI emerged from stealth with $15M in funding and backing from the Pentagon to make “large robotic armies.”

MIT developed a new approach, dubbed “Relevance,” to enable robots to better understand the real world by tuning into the most relevant data for assisting humans.

Airbus UK won a $194M contract from the European Space Agency to build what will be the first European rover set to touch down on Mars in 2030.

Starship Technologies announced that its sidewalk delivery robots have completed more than 8M autonomous deliveries in some 150 global locations.

COMMUNITY

🎥 Join our next live workshop

Join our next workshop this Friday, April 25th at 4 PM EST with Dr. Alvaro Cintas, The Rundown’s AI professor. By the end of the workshop, you’ll be ready to bring your most complex ideas to life using Canva’s powerful creative tools.

RSVP here. Not a member? Join The Rundown University on a 14-day free trial.

See you soon,

Rowan, Jennifer, and Joey—The Rundown’s editorial team

Good morning, AI enthusiasts. OpenAI just handed developers the reins to its viral image model, opening up SOTA AI creativity to thousands of apps and services overnight.

With cheap pricing and the power to render accurate text, edit images, and create diverse visual styles now embeddable in any app, OpenAI is muscling into workflows once dominated by its creative-focused rivals.

In today’s AI rundown:

• OpenAI unlocks powerful image creation via API

• Microsoft’s new AI agents, workplace AI research

• Prepare for meetings instantly with Claude

• Ex-staff, experts challenge OpenAI’s restructuring

• 4 new AI tools & 4 job opportunities

LATEST DEVELOPMENTS

OPENAI

🎨 OpenAI unlocks powerful image creation via API

Image source: OpenAI

The Rundown: OpenAI just launched its advanced image generation model, gpt-image-1, to developers via API — bringing the viral success of ChatGPT's image capabilities to third-party applications and platforms.

The details:

• The gpt-image-1 model powers ChatGPT's image generation feature, which produced over 700 million images in just one week after its launch in March.

• The model enables high-quality image creation with varied styles, accurate text rendering, enhanced image editing, and more.

• OpenAI revealed that major platforms, including Adobe, Figma, and Canva, are already integrating the technology for professional design workflows.

• Developers can also control the moderation level to tailor generated content safety, with standard "auto" filtering or less restrictive "low" moderation.

• Pricing is structured per token usage, with text prompts ($5 / 1M), input images ($10 / 1M), and output images ($40 / 1M), or ≈2-19c per image based on quality.

Why it matters: After ChatGPT's image generator went viral last month, OpenAI is positioning itself to power the next wave of AI-enhanced creative tools — and there will be a TON of them. The API release also deepens the broader creative ecosystem’s reliance on OpenAI, a position previously held by other rivals.

TOGETHER WITH WORKOS

🔑 Scale your AI app for enterprise

The Rundown: AI tools are reshaping how teams work—but enterprise adoption takes more than innovation. It demands security. WorkOS provides drop-in APIs to help you meet enterprise requirements fast—without months of custom development.

With WorkOS, you'll:

• Quickly integrate SSO, SCIM, audit logs, and more

• Get Enterprise Ready in days, not months

• Join OpenAI, Perplexity, and Cursor in scaling faster

Make your app Enterprise Ready with WorkOS today.

MICROSOFT

🤖 Microsoft’s new AI agents, workplace AI research

Image source: Microsoft

The Rundown: Microsoft just released two new Copilot agents, Researcher and Analyst, alongside its 2025 Work Trend Index report — which maps out the rise of AI-centric, human-led “Frontier Firms” set to reshape the workplace.

The details:

• Researcher and Analyst bring deep reasoning to M365 Copilot for complex research and data science tasks like forecasting.

• The agents are rolling out as part of Copilot’s “Frontier” early access program, alongside updates that let companies build autonomous multi-agent systems.

• Microsoft's research across 31,000 workers shows companies leading in AI adoption are seeing major results:

  • 71% report their company is thriving vs 37% globally

  • 55% say they can handle increased workloads vs 20% globally

  • Workers show higher optimism about career opportunities

• Microsoft also believes that every employee will become an “agent boss,” with all companies becoming AI-human “Frontier Firms” for operations in 2-5 years.

Why it matters: Microsoft’s agent release coincides nicely with how the company views the future of work: one dominated by AI agents, but managed by humans. We’re in the beginning stages of a complete shift in how work gets done — with human-agent teams becoming the norm and companies moving to fully hybrid, AI-infused structures.

AI TRAINING

🤝 Prepare for meetings instantly with Claude

The Rundown: In this tutorial, you will learn how to use Claude's integration with Calendar and Gmail to easily research meeting participants, analyze previous communications, and get detailed company information all in one place.

Step-by-step:

1. Head over to Claude and click the settings menu to toggle Gmail and Calendar search.

2. Ask Claude to check your calendar and research participants by using a prompt like: "Check my calendar for Thursday and provide a brief summary about the participants and company."

3. Review past communications by asking: "Check my email for previous conversations with [name] or someone from [company]."

4. Request to recommend talking points based on the combined insights.

Note: Only connect your accounts if you feel comfortable sharing your information with Claude. Also, you can toggle these features on or off at any time for privacy control.

PRESENTED BY QWIET

🛡️ Revolutionizing AppSec with agentic AI

The Rundown: Qwiet AI’s Multi-AI Agent Approach revolutionizes application security (AppSec) by using multiple AI agents to provide enhanced code security and significantly improve efficiency between security and developer teams.

In this webinar, you’ll discover:

• Why multiple AI agents outperform single-agent solutions

• How Agentic AI corrects code vulnerabilities even when standard AI copilots fail

• The critical partnership between humans and AI in effective AppSec

Learn more and experience smarter AppSec with Qwiet AI.

OPENAI VS. THE INDUSTRY

❌ Ex-staff, experts challenge OpenAI's restructuring

Image source: Not For Private Gain

The Rundown: More than 30 AI experts and ex-OpenAI staffers published an open letter urging the attorneys general of Delaware and California to block OpenAI’s restructuring, warning it would undermine its original mission to benefit humanity.

The details:

• 9 former OpenAI employees joined notable figures like AI ‘godfather’ Geoffrey Hinton in calling to block the startup’s transition from nonprofit to for-profit.

• They argue the move will remove vital nonprofit oversight and safeguards, and redirect AGI development from public benefit to shareholder returns.

• OpenAI needs transition approval from both state AGs by year-end to secure a pending $40B SoftBank investment contingent on the restructuring.

• The letter follows an earlier motion by 12 former employees seeking to weigh in on Elon Musk's lawsuit against the company and CEO Sam Altman.

Why it matters: With billions in investments hanging in the balance and both former employees and AI heavyweights lining up against the AI leader, the opposition continues to mount against OpenAI’s for-profit shift — but we’ve seen little movement to suggest the transition is in jeopardy, despite the red flags being raised.

QUICK HITS

🛠️ Trending AI Tools

• 🗣️ Dia 1.6B - Open-source text-to-speech AI with emotion controls and more

• 🎬 AvatarFX - Create long-form, coherent talking avatars from a reference photo

• 🤖 Kortix Suna - Open-source AI agent built to work like a human

• 📋 ShumerPrompt - Discover and share powerful AI prompts

💼 AI Job Opportunities

• 📱 Pika - Software Engineer

• 🤖 Snorkel - AI Solutions Engineer

• 🧾 Synthesia - Financial Accountant

• 💼 Glean - Strategic Finance Lead

📰 Everything else in AI today

Perplexity released its Perplexity Assistant app on iOS, allowing users to take agentic actions, access web browsing, and more on mobile using voice commands.

ByteDance’s Dreamina launched Seedream 3.0, a new text-to-image model that ranks No. 2 on Artificial Analysis’ Image Arena Leaderboard behind only GPT-4o.

OpenAI is reportedly forecasting sales of $125B in 2029 and $174B in 2030, powered by AI agents, “new products,” and API and user growth.

NVIDIA released its NeMo microservices suite, allowing enterprises to easily build AI agents with optimized company data flywheels for high-quality performance.

BMW announced plans to integrate Chinese startup DeepSeek’s AI models into its new vehicles in the region starting later this year.

Tempus AI is partnering with biotech giants AstraZeneca and Pathos to develop the industry’s largest multimodal foundation model for cancer treatment discovery.

COMMUNITY

🎥 Join our next live workshop

Join our next workshop this Friday, April 25th, at 4 PM EST with Dr. Alvaro Cintas, The Rundown’s AI professor. By the end of the workshop, you’ll be ready to bring your most complex ideas to life using Canva’s powerful creative tools.

RSVP here. Not a member? Join The Rundown University on a 14-day free trial.

🤝 Share The Rundown, get rewards

We’ll always keep this newsletter 100% free. To support our work, consider sharing The Rundown with your friends, and we’ll send you more free goodies.

See you soon,

Rowan, Joey, Zach, Alvaro, and Jason—The Rundown’s editorial team

Good morning, AI enthusiasts. Two Korean undergrads with little AI experience took Sam Altman’s “you can just build things” quote to heart — building an open-source speech model that outperforms industry leaders — with zero funding.

With AI tools and resources becoming more accessible globally, the era of solo builders competing with billion-dollar labs has officially arrived.

P.S. — Our next workshop is today at 1 pm, where you’ll learn how to turn raw ideas into polished, on-brand presentations with Gamma. RSVP here.

In today’s AI rundown:

• Two undergrads unveil SOTA speech AI

• The Washington Post joins OpenAI's alliance

• Automate your sales with personalized emails

• Anthropic CISO: AI employees are coming

• 4 new AI tools & 4 job opportunities

LATEST DEVELOPMENTS

NARI LABS

🗣️ Two undergrads unveil SOTA speech AI

Image source: Nari Labs

The Rundown: Korean startup Nari Labs released Dia, an open-source text-to-speech model that claims to exceed the capabilities of leading commercial offerings like ElevenLabs and Sesame — developed by two undergraduate techies with zero funding.

The details:

• The 1.6B parameter model supports advanced features like emotional tones, multiple speaker tags, and nonverbal cues like laughter, coughing, and screams.

• The work was inspired by Google’s NotebookLM, with Nari also using Google’s TPU Research Cloud program for compute access.

• Side‑by‑side tests show Dia outshining ElevenLabs Studio and Sesame CSM‑1B in timing, expressiveness, and handling nonverbal scripts.

• Nari Labs founder Toby Kim said the startup plans to develop a consumer app focused on social content creation and remixing based on the model.

Why it matters: Dia is a living testament to Sam Altman’s ‘you can just do things’ tweet, with two inexperienced undergrads training an open-source model that competes with the top voice tech on the market. There’s never been a better time to try and build something, with AI unlocking new access to learning like never before.

TOGETHER WITH GAMMA

🚀 Create stunning content in minutes, not hours

The Rundown: For consultants, educators, marketers, and sales pros drowning in content creation, Gamma's all-in-one platform is the lifeline you need — helping you create beautiful presentations, websites, and social media content with minimal effort.

Gamma allows you to:

• Generate content from simple text prompts or import and improve existing work

• Edit images with a single click — change subjects, styles, or remove backgrounds

• Easily export content to Google Slides/PDFs/PowerPoint

• Experience an upgraded AI and UI for a powerful content creation experience

Click here to experience Gamma and join 50M other users creating 700k presentations a day.

OPENAI

📰 The Washington Post joins OpenAI's alliance

Image source: Washington Post

The Rundown: The Washington Post just announced a new partnership with OpenAI, allowing the AI leader to bring summaries and links from its reporting directly into ChatGPT answers.

The details:

• ChatGPT will now feature summaries, quotes, and direct links to relevant Washington Post articles in its responses to user questions.

• The deal adds the Jeff Bezos-owned Post to OpenAI's expanding roster of media partners, with over 20 major news publishers.

• It also comes amid ongoing legal battles between OpenAI and other major publishers, including the NYT, over training data and copyright issues.

• The Washington Post has been actively experimenting with AI, launching tools like Ask The Post AI and Climate Answers over the past year.

Why it matters: Another major news outlet is choosing partnership over litigation, with a bet that visibility via ChatGPT will be key to reaching audiences worldwide. OpenAI also gains more trustworthy content for its AI, and a giant like WaPo joining on further isolates publishers like the NYT, who are fighting back in court.

AI TRAINING

📊 Automate your sales with personalized emails

The Rundown: In this tutorial, you will learn how to use n8n to turn your static contact lists into dynamic sales outreach machines by automatically sending personalized emails to prospects based on their company, role, and interests.

Step-by-step:

1. Create a new n8n workflow and set up a Google Sheets trigger that monitors when new leads are added to your spreadsheet.

2. Add an AI Agent node and connect it to a language model to process your contact information.

3. Configure a Gmail node to create drafts of personalized emails instead of sending them directly.

4. Write detailed instructions in the AI Agent's system message telling it exactly how to craft sales emails.

Pro tip: Fine-tune your AI's writing style by testing different system message instructions. We did an extensive workshop showing how to create your own AI Agent to automate tasks and run local AI models with n8n here.

PRESENTED BY AXOLOTL

⚙️ Post‑train LLMs without limits

The Rundown: Axolotl v0.8.0 is an open‑source toolkit built by devs, for devs — giving you full transparency and the freedom to customize large language models through simple configuration.

With Axolotl, you can:

• Post‑train LLaMA, Gemma, Mistral, and other leading models from one workflow

• Scale effortlessly — from a single local GPU to multi‑node cloud clusters

• Get started fast thanks to an easy, plug‑and‑play setup

Dive into Axolotl v0.8.0 to start post‑training today.

ANTHROPIC

💼 Anthropic CISO: AI employees are coming

Image source: Reve / The Rundown

The Rundown: Anthropic's Chief Information Security Officer, Jason Clinton, just predicted that AI-powered virtual employees will begin operating on corporate networks within the next year, bringing major new challenges in security management.

The details:

• These AI employees would have their own corporate accounts, passwords, and "memories," a significant step up from current task-specific AI agents.

• Clinton said security challenges will include managing AI account privileges, monitoring access, and determining responsibility for autonomous actions.

• He sees virtual employees as the next “AI innovation hotbed,” with virtual employee security also emerging as an area of focus alongside it.

• Anthropic said it's focused on securing its own AI models against attacks and watching out for potential areas of misuse.

Why it matters: Work is about to change completely in the AI age, and so will the security measures needed for this brand-new type of employee. The question is if cybersecurity practices will update fast enough to keep up — or if it will take a major breach or exploit as a wake-up call for new threats that autonomous AI workers present.

QUICK HITS

🛠️ Trending AI Tools

• 🗣️ Agent-to-Agent Transfers - Hand off conversations between AI agents

• 📽️ AI Co-Editor - Descript’s new agentic video editor

• ⚡️ Genspark AI Slides - Agentic tool for quickly creating presentation slides

• 🎥 Edits - Meta’s new video creation app with AI features

💼 AI Job Opportunities

• 🔐 Runway - Member of Technical Staff

• 🎧 Soundhound AI - Customer Success Representative

• 📈 Abridge - Director, GTM Strategy + Pricing

• 🗣️ Captions - Community Manager

📰 Everything else in AI today

OpenAI’s head of product, Nick Turley, testified in Google’s antitrust trial that the AI leader would be interested in buying its Google Chrome browser if a sale were forced.

Apple removed "available now" claims from its Apple Intelligence marketing page following the National Advertising Division's concerns about misleading availability.

Character AI launched AvatarFX, an AI platform that allows users to create long-form, coherent talking avatars from a single reference photo and voice selection.

IBM and the European Space Agency released TerraMind, an open-source AI system that uses nine data modalities and satellites for real-time climate monitoring.

Cohere CEO Aidan Gomez joined the board of electric automaker Rivian, aiming to integrate AI tech more broadly into the company’s products and manufacturing.

Motorola debuted SVX, a new AI-powered device that combines a body camera, speakers, and an AI assistant to reduce emergency response times.

COMMUNITY

🎥 Join our next live workshop

Join our next workshop today at 1 PM EST with Mel, Creative Director at Gamma. By the end of the workshop, you’ll learn how to turn raw ideas into polished, on-brand presentations using Gamma’s powerful AI storytelling tools — no design skills needed.

RSVP here. Not a member? Join The Rundown University on a 14-day free trial.

🤝 Share The Rundown, get rewards

We’ll always keep this newsletter 100% free. To support our work, consider sharing The Rundown with your friends, and we’ll send you more free goodies.

See you soon,

Rowan, Joey, Zach, Alvaro, and Jason—The Rundown’s editorial team

Good morning, tech enthusiasts. Chungin “Roy” Lee, the 21-year-old co-founder of Cluely, just announced that his startup has raised $5.3M for its controversial AI tool to “cheat on everything.”

Cluely, which also dropped a clip showing Lee using the tech to impress a date, argues that AI cheating tools will become as accepted as calculators and spellcheckers. Is this the natural evolution of human-AI collaboration, or are we living in an episode of Black Mirror? 

In today’s tech rundown:

• Cluely raises $5.3M to ‘cheat on everything’

• Netflix eyes video podcasts in next big move

• FTC sues Uber for ‘deceptive’ practices

• Airbnb to display full prices up front

• Quick hits on other major news

LATEST DEVELOPMENTS

CLUELY

 🤥 Cluely raises $5.3M to ‘cheat on everything’

Image source: Cluely/X

The Rundown: On Sunday, 21-year-old Chungin “Roy” Lee announced that his startup, Cluely, has raised $5.3M in seed funding to expand its controversial AI tool designed to cheat on everything from job interviews to exams.

The details:

• The tool operates via a hidden, undetectable in-browser window that provides users with real-time answers and suggestions during interviews or calls.

• It “listens” to audio, “sees” the user’s screen, and delivers instant answers invisible to interviewers or anyone else on the other end.

• Lee, a former Columbia University student, was suspended after he co-created an earlier version of the tool for software engineer job interviews.

• The controversy around “AI cheating” has only fueled Cluely’s growth: the startup launched this month and already boasts $3M+ in annual revenue.

Why it matters: Investors have flocked to the startup despite concerns from recruiters and tech giants like Amazon. But Cluely isn’t the only polarizing new AI startup: Famed AI researcher Tamay Besiroglu has also launched his own startup, Mechanize, with a mission to replace all human workers everywhere with AI.

NETFLIX

🎤 Netflix eyes video podcasts in next big move

Image source: Ideogram/The Rundown

The Rundown: Netflix is reportedly gearing up to expand beyond TV and film, with video podcasts emerging as its next big content play—a strategic move to challenge YouTube’s dominance in the space.

The details:

• Co-CEO Ted Sarandos said last week that “as the popularity of video podcasts grows, I suspect you’ll see some of them find their way to Netflix.”

• According to Edison Podcast Metrics, YouTube attracts 31% of weekly podcast listeners, compared to Spotify’s 27% and Apple’s 15%.

• Nearly half of podcast listeners watch their shows on smart TVs, and in March, YouTube made up 9.7% of all TV viewing, compared to Netflix’s 8.1%.

• Netflix has already dabbled in podcast-adjacent and creator-driven content, licensing a few shows and producing podcasts tied to its original series.

Why it matters: Netflix’s push into video podcasts is fueled in part by their lower production costs compared to scripted shows and films. With Gen Z increasingly tuning into podcasts in video format, the company is now reportedly stepping up efforts to woo creators and stake its claim in this growing trend.

UBER

⚖️ FTC sues Uber for ‘deceptive’ practices

Image source: Ideogram/The Rundown

The Rundown: Uber is facing a lawsuit from the U.S. Federal Trade Commission (FTC) over allegations of deceptive billing and cancellation practices tied to its monthly Uber One subscription service.

The details:

• The FTC claims that Uber enrolled customers on Uber One without their explicit consent and then made it extremely difficult to end the subscription.

• According to the complaint, the ride-hailing giant promised savings of $25 a month, but this figure did not account for the $9.99 monthly subscription cost.

• Plus, it made cancelling the subscription “unreasonably” burdensome for users, requiring them to navigate up to 23 screens.

• The commission also alleges that users were billed before their free trial ended, with some asked to contact customer support with no further instructions.

Why it matters: Uber has denied all allegations, asserting that its sign-up and cancellation processes are streamlined, with app cancellations taking just 20 seconds. Nevertheless, the move comes amid a broader FTC crackdown on “dark patterns” that make it hard for users to cancel, targeting giants like Amazon, Adobe, and Match.

AIRBNB

🏠 Airbnb to display full prices up front

Image source: Ideogram/The Rundown

The Rundown: Airbnb has responded to major pushback from both consumers and regulators by introducing a major update to its pricing transparency, now displaying the total price, including all fees, by default to users during the booking process.

The details:

• Airbnb said it will now display the total price of a stay, including all service charges, cleaning, and additional fees, by default before taxes.

• The guest service fee can be up to 14.2% of the booking subtotal, while cleaning and other maintenance fees set by hosts will also go in the total price.

• The change comes ahead of new U.S. federal regulations targeting so-called “junk fees,” set to take effect on May 12.

• This move will help travelers make more informed decisions and avoid sticker shock at checkout.

Why it matters: Airbnb has been testing this feature since 2019 in various countries and now plans to bring it worldwide as U.S. regulations tighten. For Airbnb hosts, however, this change may require more strategic pricing, but Airbnb plans to show hosts how much their guests are paying before taxes to help make adjustments.

QUICK HITS

📰 Everything else in tech today

Nintendo announced that it opened Switch 2 pre-orders in the U.S. and that its planned price of $499.99 will remain intact despite tariff concerns.

Instagram is intensifying its efforts to clamp down on teenagers who misrepresent their age by using AI tools to scan accounts for clues indicating that the user is a minor.

Apple has reportedly been assembling its iPhone 16e in Brazil (since the phone launched) as it moves production out of China amid the ongoing tariff war.

Tesla investors are pushing for updates from CEO Elon Musk on the launch timelines of Tesla’s long-promised affordable EV and its robotaxi service in Austin.

The U.S. government has cancelled or downsized $8B worth of clean energy projects since January, according to a report from nonpartisan think tank E2.

China's CATL launched Naxtra, a new brand for its sodium-ion batteries, and the second generation of its fast-charging battery for electric cars.

Bluesky launched its own blue checkmark verification system similar to X, but one that allows “trusted verifiers” to distribute the checks

Car rental company Hertz partnered with AI vehicle-inspection service UVeye to scan airport returns for damage and maintenance issues.

Manychat, a startup that provides brands with AI tools for automated messaging and chats across social media, raised $140M in funding.

Jeff Bezos-backed Slate Auto reportedly teased its secretive EV concept on the streets of California with a “Transformer-like design.”

COMMUNITY

🎥 Join our next live workshop

Join our next workshop today at 3 PM EST with Matt Waters from Superhuman. By the end of this workshop, you’ll have a fully optimized email system powered by AI, so you can move 4x faster and eliminate inbox chaos.

RSVP here. Not a member? Join The Rundown University on a 14-day free trial.

See you soon,

Rowan, Jennifer, and Joey—The Rundown’s editorial team

Good morning, AI enthusiasts. Anthropic just pulled back the curtain on AI morality — revealing the first-ever map of Claude's real-world values based on hundreds of thousands of actual conversations.

With AI systems increasingly shaping our decisions (and now even legislation), cracking the alignment code behind models’ moral compass has never been more important.

Reminder: Our next workshop is today at 3 PM EST with Superhuman, where you’ll learn to fully optimize your email system with AI and eliminate inbox chaos. RSVP here.

In today’s AI rundown:

• Anthropic charts Claude's values

• UAE plans to let AI write the laws

• Research with NotebookLM web discovery

• Hassabis: AI could end all disease

• 4 new AI tools & 4 job opportunities

LATEST DEVELOPMENTS

ANTHROPIC

🧭 Anthropic charts Claude's values

Image source: o3 / The Rundown

The Rundown: Anthropic just published a study analyzing hundreds of thousands of real AI conversations to understand how models like Claude make moral judgements — building the first large‑scale map of the model's values in day‑to‑day interactions.​

The details:

• Researchers analyzed over 300,000 real (but anonymous) conversations to find and categorize 3,307 unique values expressed by the AI.

• They found 5 types of values (Practical, Knowledge-related, Social, Protective, Personal), with Practical and Knowledge-related being the most common.

• Values like helpfulness and professionalism appeared most frequently, while ethical values were more common during resistance to harmful requests.

• Claude's values also shifted based on context, such as emphasizing "healthy boundaries" in relationship advice vs "human agency" in AI ethics discussions.

Why it matters: AI is increasingly shaping real-world decisions and relationships, making understanding their actual values more crucial than ever. This study also moves the alignment discussion toward more concrete observations, revealing that AI’s morals and values may be more contextual and situational than a static point of view.

TOGETHER WITH VANTA

🛡️ Scaling security in the age of AI

The Rundown: Join Vanta, Wiz, and Modo Labs on May 8 for a live fireside chat that demystifies how leading companies evolve GRC and cloud defenses — so you can scale security without stalling innovation.

This interactive discussion will include:

• Insights into top-of-mind issues for GRC and security pros

• Actionable steps to strengthen your existing program by actioning on risks

• Key strategies for integrating AI capabilities into your security operations

Register now to save your spot.

AI & GOVERNMENT

🇦🇪 UAE plans to let AI write the laws

Image source: Reve / The Rundown

The Rundown: The United Arab Emirates unveiled plans to become the first nation to integrate AI directly into its lawmaking process, establishing a new government unit to oversee the transformation of how laws are written, reviewed, and updated.

The details:

• A new Regulatory Intelligence Office will lead the initiative, which aims to cut legislative development time by 70% through AI-assisted drafting and analysis.

• The system will use a database combining federal and local laws, court decisions, and government data to suggest legislation and amendments.

• The plan builds on the UAE’s major investments in AI, including a dedicated $30B AI-focused infrastructure fund through its MGX investment platform.

• The move was met with mixed reactions, with experts warning of the tech’s reliability, bias, and interpretive issues present in training data.

Why it matters: While many governments have already begun integrating AI into their ranks, this is one of the first examples of giving it legislative power in some capacity. As systems reach superhuman levels of persuasion, reasoning, and more, their use in politics will raise existential questions about AI vs. human judgment in lawmaking.

AI TRAINING

🔎 Research with NotebookLM web discovery

The Rundown: In this tutorial, you will learn how to use NotebookLM's new Discover Sources feature to find and add relevant web sources to your notebook with just a few clicks, streamlining the entire research process.

Step-by-step:

1. Visit NotebookLM and create a new notebook.

2. Click the "Discover" button in the Sources panel and enter a specific topic.

3. Review the curated sources that appear and add the most relevant ones to your notebook with one click.

4. Use NotebookLM's features with your new sources: generate Briefing Docs, ask questions via chat, or create Audio Overviews.

Pro tip: The more specific your topic description, the more relevant your source recommendations will be. Try describing exactly what you need to learn rather than using broad terms.

PRESENTED BY INNOVATING WITH AI

🚀 Launch a six‑figure AI consultancy in six months

The Rundown: Innovating With AI’s ‘The AI Consultancy Project’ gives you the frameworks, playbooks, and client‑ready templates needed to turn “interesting AI ideas” into a revenue‑generating business – helping you ride the wave of an AI consulting boom expected to grow by 8x this decade.

In this program, you will:

• Gain the tools and frameworks to find clients and deliver top-notch services

• Follow a 6-month plan to build a 6-figure AI consulting business

• Join a 700‑strong cohort where some members landed their first AI client in 72 hours

Click here to request access to The AI Consultancy Project.

GOOGLE DEEPMIND

⏰ Hassabis: AI could end all disease

Image source: CBS News

The Rundown: Nobel laureate and Google DeepMind CEO Demis Hassabis was interviewed on 60 Minutes, where he provided insights into AGI timeline, progress, and AI’s potential in medicine, while demoing DeepMind’s “Project Astra” assistant.

The details:

• Hassabis said AI-driven drug discovery could compress medical timelines from years to weeks, potentially eliminating all disease within a decade.

• His Project Astra demo included ID’ing paintings, reading emotions, and even a glasses-embedded version showcasing live features with visual understanding.

• Hassabis said AGI will arrive in 5-10 years — and while he doesn’t believe today’s AI is conscious, he said it could emerge in the future in some form.

• Another demo previewed an experimental robotics system with reasoning, showing the ability to understand abstract concepts like color mixing.

Why it matters: Coming from DeepMind's Nobel-winning chief, Hassabis' commentary isn’t just hype, but a signal of intense conviction from a key player in the field. While lofty goals like the end of disease and “radical abundance” sound like a pipe dream, 5-10 years of exponential growth is a scale that is hard to comprehend.

QUICK HITS

🛠️ Trending AI Tools

• 🤖 Gemma 3 QAT - 27B model that can run on consumer GPUs

• 📋 Grok Studio - Collaborative workspace for docs, code, and more

• 🎬 Kling Multi Elements - Add, swap, or delete video elements with prompting

• 🗣️ Orpheus TTS - Open-source text-to-speech AI with natural emotion

💼 AI Job Opportunities

• 🚗 Waymo - Software Engineer

• ⚖️ Hive - Legal Operations Analyst

• 📝 Grammarly - Senior Marketing Operations Manager

• 🤝 UiPath - Enterprise Sales Executive

📰 Everything else in AI today

Chinese chipmaker Huawei is reportedly preparing shipments of a new AI chip, 910C, rivaling Nvidia’s H100 and aiming to fill the void left by U.S. export restrictions.

Amazon is facing customer pushback over Bedrock service limitations for Anthropic’s models, with users reporting using Anthropic’s API to bypass the capacity issues.

Elon Musk is reportedly looking to raise $25B+ in fresh capital for his new xAI-X combined venture, which would place the company at a valuation as high as $200B.

ElevenLabs released Agent-to-Agent Transfers, allowing for the ability to transfer conversations between specialized agents for multi-layer workflows.

The Academy of Motion Pictures Arts & Sciences officially allowed the use of AI in film production, saying its use will “neither help nor harm the chances” of a nomination.

Anthropic published a new best practices guide to its Claude Code platform, providing detailed tips and patterns for success with the agentic coding tool.

COMMUNITY

🎥 Join our next live workshop

Join our next workshop today at 3 PM EST with Matt Waters from Superhuman. By the end of this workshop, you’ll have a fully optimized email system powered by AI, so you can move 4x faster and eliminate inbox chaos.

RSVP here. Not a member? Join The Rundown University on a 14-day free trial.

🤝 Share The Rundown, get rewards

We’ll always keep this newsletter 100% free. To support our work, consider sharing The Rundown with your friends, and we’ll send you more free goodies.

See you soon,

Rowan, Joey, Zach, Alvaro, and Jason—The Rundown’s editorial team

Good morning, robotics enthusiasts. As U.S. police roll out a range of robots—from Boston Dynamics’ Spot to drones—Thailand is taking it further with its first humanoid police officer, a dystopian-looking machine clad in a full police uniform.

Robots are expected to play a key role in law enforcement as more police departments around the globe embrace physical AI surveillance and response—but can we keep the boundaries of privacy and control in check?

In today’s robotics rundown:

• Thailand deploys humanoid police

• Robots go the distance at Beijing marathon

• Ex-Figure CTO’s ‘Persona’ humanoid

• Robot fingers touch and feel more like humans

• Quick hits on other robotics news

LATEST DEVELOPMENTS

ROYAL THAI POLICE

👮🏼‍♂️ Thailand deploys humanoid police

Image source: Royal Thai Police/Facebook

The Rundown: Thailand just introduced its first AI-powered police robot, AI Police Cyborg 1.0, during the bustling Songkran festival in Nakhon Pathom province—marking a strategic leap in how humanoids are being used worldwide.

The details:

• The robot was deployed to manage the massive crowds that gather for the water festival, where traditional policing resources can be stretched thin.

• It is equipped with a 360-degree AI camera system that integrates live feed with CCTV and drone footage, feeding info to the province’s command center.

• The cameras enable facial recognition, behavioral analysis, and weapon detection, allowing the robot to identify potential threats.

• The system is also reportedly context-aware—with the ability to distinguish between harmless festival items like water guns and actual threats.

Why it matters: Humanoids are still learning to walk on their own, so deploying fully active robo-police feels more like a symbolic gesture than a strategic move. Still, with the global security robot market projected to hit $44 billion by 2030, concerns around privacy are only set to grow.

BEIJING ROBOT MARATHON

🏃🏻 Robots go the distance at Beijing marathon

Image source: CGTN/YouTube

The Rundown: The world’s first humanoid half-marathon took place in Beijing this weekend, where 21 bipedal robots joined 12,000 human runners on a 21 km course—though only four robots made it to the finish line.

The details:

• The winning robot, UBTech’s Tiangong Ultra, completed the race in 2 hours and 40 minutes, after three battery swaps and a fall during the race.

• Many robots struggled with the physical demands of the race: some fell at the starting line, others overheated, and one even lost its head.

• Bloomberg reports that it took more than three hours for the other three bots to finish the course, while the fastest human finished in 1 hour and 2 minutes.

• The robots received technical support throughout the race, including battery swaps, cooling sprays, and navigation assistance from their engineering teams.

Why it matters: Even though only a few robots made it to the finish line and Unitree, one of China’s most hyped robotics companies, addressed the viral clip of its G1 collapsing by clarifying that it wasn’t an official participant, the marathon was surely a success—at capturing public attention and showcasing China’s future in embodied AI.

PERSONA AI

🤖 Ex-Figure CTO’s ‘Persona’ humanoid

Image source: Persona AI/X

The Rundown: Robotics newcomer Persona—a Houston startup founded by three robotics veterans, including former Figure CTO Jerry Pratt—just teased an image of its upcoming humanoid on X.

The details:

• Persona, which launched mid-2024, is developing a rugged, five-foot-eight-inch humanoid designed for heavy-duty industrial environments. 

• The other co-founders include CEO Nic Radford, who co-founded Nauticus Robotics and led NASA’s Valkyrie team, and robotics expert Jide Akinyode.

• Persona has already attracted significant attention and raised over $10M in pre-seed funding from investors in the U.S., London, and Singapore.

• The company is operating in “anti-stealth mode,” openly sharing its progress and aiming to deliver its first units to customers within 18 to 24 months.

Why it matters: Persona’s founders see their robot as the “Ford F-150 pickup truck” of humanoids—prioritizing durability and real-world utility over the sleeker but less robust designs of many competitors. While the space is crowded, it does seem to be smartly positioned for a slice of a market projected to reach $7T by 2050.

CHINESE ACADEMY OF SCIENCES

👌🏽 Robot fingers touch and feel more like humans

Image source: Ideogram/The Rundown

The Rundown: Chinese researchers are moving closer to human-like touch sensitivity with a new finger-shaped multimodal tactile sensor, detailed in Advanced Materials, that overcomes challenges of existing sensor tech.

The details:

• This sensor can detect not only the magnitude but also the direction of applied forces—a challenge that has limited previous sensor designs.

• Its structure consists of an internal force-sensing section featuring silicone microneedle arrays, a silicone bump, and five silver electrodes.

• The sensor uses the triboelectric effect for signal generation, enabling it to distinguish 12 common real-world materials with an accuracy of 98.33%.

• Interlocking microstructures within the sensor mimic the nuanced touch of human skin, allowing for more sensitive detection of subtle force directions.

Why it matters: Compared to rivals FingerVision and Ras Labs, researchers say their CAS sensor stands out for its ability to discern 12 materials in real time. The team envisions future enhancements, such as expanding the range of detectable materials and tactile cues, for use in smart prosthetics and intelligent manufacturing.

QUICK HITS

📰 Everything else in robotics today

NYC researchers unveiled RUKA, an open-source, tendon-driven robotic hand (with 15 degrees of freedom) that can operate for 20 hours without performance loss.

Chinese automaker Chery Auto updated its first humanoid, Mornine, with autonomous navigation and walking (while sporting a long blond wig) capabilities.

AI security robot maker Knightscope said that it has secured a 33K-square-foot headquarters in Silicon Valley to “power” its next phase of growth.

Chinese researchers are crafting biodegradable robots and robot parts, made from pork gelatin and plant cellulose, that can decompose within weeks.

U.S. robotics startup Foundation Robotics released a clip on X showing the evolution of its first humanoid Phantom’s legs.

Strawberry-harvesting robots in Japan, which use lidar tech for 3D mapping, are transforming agriculture while eliminating the need for human labor.

EU-funded researchers developed autonomous robots that can sort and remove batteries from electronics in e-waste recycling centers.

Chinese researchers developed the “world’s smallest and lightest” untethered terrestrial-aerial microbot that is capable of transforming into different shapes.

UK-based Open Bionics unveiled a wireless bionic arm, Hero, that the company claims to be the lightest hand available and fully waterproof.

COMMUNITY

🎥 Join our next live workshop

Join our next workshop on Wednesday, April 23rd, at 1 PM EST with Mel, Creative Director at Gamma. By the end of the workshop, you’ll know how to turn raw ideas into polished, on-brand presentations using Gamma’s powerful AI storytelling tools.

RSVP here. Not a member? Join The Rundown University on a 14-day free trial

See you soon,

Rowan, Jennifer, and Joey—The Rundown’s editorial team

Good morning, AI enthusiasts. The AI automation conversation just got uncomfortably direct — with Epoch co-founder Tamay Besiroglu launching Mechanize, a startup with the explicit goal of replacing human workers entirely.

But with job losses already mounting and user trust in AI still lacking, getting the masses on board with the utopian dream of abundance from mass automation may be tougher than expected.

In today’s AI rundown:

• AI startup to automate the entire workforce

• Cursor AI’s hallucinated policy sparks cancellations

• Create full-stack web apps without coding

• DeepMind’s shift to 'experiential' AI learning

• 4 new AI tools & 4 job opportunities

LATEST DEVELOPMENTS

MECHANIZE AI

⚙️ AI startup to automate the entire workforce

Image source: GPT-4o / The Rundown

The Rundown: Epoch co-founder Tamay Besiroglu just launched Mechanize, a new startup developing virtual environments and training data to enable AI agents that can replace human workers for the “full automation of all work”.

The details:

• The company plans to create simulations of workplace scenarios to train AI agents in handling complex, long-term tasks currently performed by humans.

• Mechanize will initially focus on automating white-collar jobs, with systems that can manage computer tasks, handle interruptions, and coordinate with others.

• Backed by tech leaders including Jeff Dean and Nat Friedman, the startup estimates its potential market at $60T globally.

• The announcement drew criticism for both the economic implications and potential conflicts with Besiroglu's role at AI research firm Epoch.

Why it matters: Besiroglu and co. likely aren’t the only researchers that think AI is set to automate every aspect of work — but with tensions already high over both negative views of AI and mounting job losses, this goal might be saying the quiet part a bit too loudly. The age of automation is coming, and not everyone will be happy about it.

TOGETHER WITH SALESFORCE

💡Power enterprise AI with unified data

The Rundown: AI transformation is no longer optional — and companies need unified data to drive informed action across the enterprise. Data Cloud is the intelligent activation layer of the Salesforce Platform, bridging data silos and harmonizing info from various sources into a single customer view.

Data Cloud grounds AI agents built with Agentforce enabling:

• More relevant, accurate output and actions

• Automation through data-triggered workflows

• Analytics in the workflow

• AI-powered applications

Learn more and explore how Data Cloud powers AI success.

CURSOR

🤖 Cursor AI’s hallucinated policy sparks cancellations

Image source: Reve / The Rundown

The Rundown: Agentic coding platform Cursor faced backlash after its AI support agent, Sam, hallucinated a fake policy that caused user outrage and subscription cancellations.

The details:

• A Reddit user experienced unexpected logouts when switching between devices, leading to a support inquiry answered by an AI agent.

• The AI hallucinated a policy claiming single-device restrictions were an intentional security feature, with the post sparking backlash and cancellations.

• Cursor's co-founder acknowledged the error, explaining a security update caused login issues, with the policy completely fabricated by the AI.

• He added that the company is implementing clear AI labeling for support responses going forward and refunding the affected users.

Why it matters: The hype surrounding AI agents has never been stronger, but cautionary tales like this one show that hallucinations are still a major issue to consider when deploying customer-facing bots. Despite companies rushing to automate customer service, it may still be too early in the AI boom for complete automation.

AI TRAINING

🚀 Create full-stack web apps without coding

The Rundown: In this tutorial, you will learn how to use Google’s new Firebase Studio to build and deploy complex web applications — without writing a single line of code —through AI-powered prototyping.

Step-by-step:

1. Visit Firebase Studio and log in with your Google account.

2. Describe your application in detail in the "Prototype an app with AI" section.

3. Review and customize the AI-generated app blueprint (name, features, colors).

4. Test your prototype, make adjustments if needed, and click "Publish" to deploy.

Pro tip: Upload sketches or images of your app design to help the AI better understand your vision. Also, for more advanced customization, click the "Switch to code" button in the top right corner.

PRESENTED BY TAVUS

🗣️ AI that speaks — and shows up

The Rundown: A Tavus avatar appeared in a NY courtroom, sparking a national debate — and showcasing a groundbreaking future where AI doesn't just speak, but visually represents and communicates effectively at scale.

With Tavus, you can:

• Build real-time video agents and generate realistic videos instantly via simple APIs

• Create agents with 30+ languages, natural expressions, and tool-calling capabilities

• Deploy versatile video agents anywhere human interaction occurs

Start building video agents today.

GOOGLE DEEPMIND

🧠 DeepMind’s shift to 'experiential' AI learning

Image source: Google DeepMind

The Rundown: DeepMind researchers published “Welcome to the Era of Experience”, proposing AI development that moves beyond human-generated training data with “streams” that let AI learn from real-world interactions and environmental feedback.

The details:

• Authored by RL legends David Silver and Richard Sutton, the paper argues that human data training caps AI's potential and prevents truly new discoveries.

• Streams would allow AI to learn continuously with extended interactions rather than brief Q&A exchanges, enabling adaptation and improvement over time.

• AI agents would use real-world signals like health metrics, exam scores, and environmental data as feedback, rather than relying on human evaluations.

• The approach builds on techniques that helped systems like AlphaZero master games, expanding them to handle open-ended real-world scenarios.

• The researchers suggest this shift could enable AI to discover solutions beyond current human knowledge while still maintaining adaptable safety measures.

Why it matters: As AI approaches the limits of what it can learn from human-generated data, this pivot to experience-based learning could unlock more advanced capabilities — moving beyond imitating human knowledge towards truly autonomous (and superhuman) learning and discovery.

QUICK HITS

🛠️ Trending AI Tools

• 🧠 Gemini 2.5 Flash - Fast, cost-efficient model with controllable reasoning

• 🎞️ Google Whisk Animate - Turn images into eight-second animated clips

• 🎥 Wan 2.1-FLF2V - Alibaba’s open-source first and last frame video model

• 📽️ Camera Angle Concepts - Control the video POV with Luma’s Ray2 model

💼 AI Job Opportunities

• 🧵 Figure AI - Softgoods Developer

• 🛠️ Shield AI - Prototype Technician

• 💸 Palantir Technologies - Financial Analyst

• 📊 Databricks - Product Designer

📰 Everything else in AI today

Third-party testing and internal evaluations revealed that OpenAI’s new o3 and o4-mini models hallucinate significantly more than older models.

Google launched a new version of Gemma 3 with ‘Quantization-Aware Training’, enabling the 27B version to run on consumer GPUs with maintained performance.

OpenAI CEO Sam Altman revealed that the company has spent “tens of millions of dollars” in compute on users saying “please” and “thank you” to its AI models.

Wikipedia's parent, Wikimedia Foundation, partnered with Google’s Kaggle to publish a dataset for AI developers to discourage scraping of the company’s platform.

MIT published a “sequential Monte Carlo” approach that generates AI code efficiently, allowing small models to outperform larger ones by axing unpromising outputs early.

OpenAI introduced a new Flex processing option, halving API costs for o3 and o4-mini models in exchange for slower responses.

COMMUNITY

🎥 Join our next live workshop

Join our next workshop on Wednesday, April 23rd, at 1 PM EST with Mel, Creative Director at Gamma. By the end of the workshop, you’ll know how to turn raw ideas into polished, on-brand presentations using Gamma’s powerful AI storytelling tools.

RSVP here. Not a member? Join The Rundown University on a 14-day free trial.

🤝 Share The Rundown, get rewards

We’ll always keep this newsletter 100% free. To support our work, consider sharing The Rundown with your friends, and we’ll send you more free goodies.

See you soon,

Rowan, Joey, Zach, Alvaro, and Jason—The Rundown’s editorial team