Good morning, robotics enthusiasts. Beijing has opened doors to the world’s first robot mall, a sleek four-story, 13K-square-feet shrine to humanoids and service bots, complete with a robot-themed restaurant.

Visitors can test-drive mechanical butlers, pet robo-animals, and shop industrial-grade androids. It’s part retail spectacle and part China’s bold bid to dominate the robot revolution.

In today’s robotics rundown:

• China opens world’s first robot mall

• Robot cage fights are now a thing in SF

• ‘Drummer’ humanoid that can outplay humans

• China deploys battlefield ‘robot wolves’

• Quick hits on other robotics news

LATEST DEVELOPMENTS

CHINESE ROBOTICS

🤖 China opens world’s first robot mall

Image source: Ideogram/The Rundown

The Rundown: Coinciding with the 2025 World Robot Conference, China launched its first-ever robot mall in Beijing’s E-Town district, a four-story, 4K-square-meter (13K-square-feet) retail complex dedicated entirely to humanoid and service robots.

The details:

• The world’s first “humanoid robot 4S” retail space is modeled on the auto industry’s 4S formula of Sales, Service, Spare Parts, and Surveys.

• Shoppers can browse and test over 100 robots from 40+ domestic brands, spanning mechanical butlers, robotic pets, and industrial-grade humanoids.

• Prices range from 2K yuan ($278) for entry-level bots to multimillion-yuan bots powered by chips that handle up to 275 trillion operations per second.

• Government subsidies as high as $14K encourage both enterprises and consumers to deploy robots in work and daily life.

Why it matters: This marks the world’s first retail environment built exclusively for robots, melding a tech showroom, commercial hub, and live R&D testbed under one roof. China is backing robotics with more than $20B in subsidies, and a sleek new retail space like this is a major flex.

ROBOT FIGHT CLUB

🥊 Robot cage fights are now a thing in SF

Image source: Ideogram/The Rundown

The Rundown: In San Francisco, robot fight clubs are reportedly emerging as the tech scene’s new obsession, fusing DIY hardware hacking, AI, and underground performance art — all into gritty, hype-heavy spectacle for the coder crowd.

The details:

• Once demo nights, these events are quickly evolving into a scene with tickets treated as status symbols and production complexity ramping up.

• Events in San Francisco’s Frontier Tower basement feature Unitree and Booster Robotics humanoids with hundreds of tech workers and influencers attending.

• DIY ring “girls,” like a Roomba fused to a fishnet mannequin leg, whirl around to pick up thrown cash, and human emcees in glitter jackets narrate every punch.

• Bouts pit these bots in full boxing gear, costumes, and protective helmets, creating surreal scenes that blur the line between cosplay and robotics.

Why it matters: As more polished robot boxing events are gaining traction worldwide, these gritty underground scenes take a different spin, turning AI and robotics into something raw and creative. It also shows how tech culture is evolving into new forms of interactive entertainment, with things likely to get only weirder.

ROBOT RESEARCH

🥁 ‘Drummer’ humanoid that can outplay humans

Image source: Shahid, Braghin & Roveda

The Rundown: Swiss and Italian researchers just developed a humanoid drummer capable of learning and performing complex live percussion using reinforcement learning rather than preprogrammed sequences.

The details:

• The robot interprets drum scores as “rhythmic contact chains,” enabling humanlike adaptations such as alternating hands and arm crossings.

• It was trained in high‑fi simulation with MIDI tracks from multiple genres, building precise timing and expressive playing skills.

• The system coordinates up to 15 degrees of freedom in its upper body, planning movements across a full kit to maintain rhythm and spatial accuracy.

• The robot achieves over 90% rhythmic accuracy, rivaling skilled human drummers and handling demanding drum fills and rapid note sequences.

Why it matters: The next step is bringing these skills from simulation (it was tested on a Unitree G1) to a physical robot able to play acoustic drums, perform with live bands, and improvise in real time. It’s part of a wider shift from rigid programming to autonomous, adaptive learning in robotics.

ROBOT DOGS

🐺 China deploys battlefield ‘robot wolves’

Image source: CCTV

The Rundown: China’s state broadcaster CCTV released new footage of its scary “robot wolves” — quadrupedal unmanned ground vehicles armed with combat rifles — working live in People’s Liberation Army military drills.

The details:

• Unlike previous, lighter robot dogs used for reconnaissance, these beefed-up robot wolves are designed specifically for front-line combat.

• They maneuver over rugged terrain in coordinated packs carrying assault rifles (such as QBZ-191) and executing precision strikes up to 100m away.

• The robots divide battlefield tasks in a pack, with leaders transmitting battlefield images and scouts, shooters, or ammo carriers adapting on the fly.

• With AI-driven navigation, LiDAR, and near-limitless endurance, the bots are built to take on high-risk battlefield roles and reduce casualties.

Why it matters: These drills, following the robot wolves’ debut in late 2024, mark the first time such armaments have run alongside soldiers exchanging nonlethal gunfire. Of course, the bots are fueling both national pride and a heap of anxieties about the autonomous future of gun-toting battlefield machines.

QUICK HITS

📰 Everything else in robotics today

Unitree founder Wang Xingxing says robotics’ “ChatGPT moment” could arrive within two years.

Morgan Stanley’s Adam Jonas estimates that a $5/hour leased humanoid replacing two $25/hour workers could be worth $200K each in net present value.

Fourier’s new GR‑3 is a 5'4", 55 kg humanoid “care‑bot” with vision, audio, and tactile sensing — 31 pressure sensors, expressive gestures, and a soft-touch shell.

Tesla is shutting down its Dojo supercomputer team, designed to supercharge autonomous driving and robotics, with leader Peter Bannon departing the company.

A new HCI/HRI study finds that even a simple tabletop robot originally designed to read aloud can evolve into a meaningful emotional companion for children.

A mobile robot arm was tested onsite in Germany for human-machine collaboration, demonstrating real-world teamwork between robots and builders in laying bricks.

iRobot reported a Q2 2025 revenue increase of 6% in Japan, but saw sharp declines of 33% in the U.S. and 17% in EMEA compared to last year.

FORT Robotics nabbed $18.9 M in Series B funding to accelerate development of its safety platform for autonomous machines and physical AI.

COMMUNITY

🎥 Watch our last workshop

Check out our last live workshop with Dr. Alvaro Cintas, The Rundown’s AI professor. By the end of the workshop, you’ll confidently be able to launch your own fully-private, open-weight ChatGPT alternative, ready for production or personal tinkering.

Watch it 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's GPT-5 was supposed to ring in a new step up the ladder, but instead, a botched rollout has users grabbing pitchforks and demanding the old models back.

With Sam Altman hosting Reddit AMAs and bringing back the beloved 4o model, GPT-5’s launch has been a case study in what a large part of the user base actually wants in an AI assistant — which is apparently not always just more raw intelligence.

P.S. — We’re making some small changes to The Rundown AI based on your feedback, including a new Community AI workflows section to spotlight how our own readers are using AI. Want to be featured? Share your workflow with us here.

In today’s AI rundown:

• Sam Altman details GPT-5 fixes in emergency AMA

• Ex-OpenAI researcher raises $1.5B for AI hedge fund

• Automate meeting prep with AI Agents

• Google, NASA’s AI doctor for astronauts in space

• 4 new AI tools & community insights

LATEST DEVELOPMENTS

OPENAI

 🚨 Sam Altman details GPT-5 fixes in emergency AMA

Image source: OpenAI

The Rundown: OpenAI CEO Sam Altman and team members held a Reddit Q&A on Friday, following the polarizing rollout of GPT-5, which angered the user base due to technical failures, chart “crimes,” and the abrupt removal of older models.

The details:

• The rollout featured technical glitches, low rate limits, and a now-viral “chart crime” during the livestream, which Altman called a “mega chart screwup.”

• A new autoswitcher crashed on launch day, preventing GPT-5 from routing queries to the correct model and making it appear significantly less capable.

• OpenAI is now rolling out fixes, doubling Plus user rate limits, and promising more transparency and customization options for future model updates.

• Users also flooded Reddit calling for OpenAI to restore GPT-4o, mourning the loss of the older model’s personality and emotional intelligence.

• Altman admitted OpenAI underestimated how much users valued 4o, committing to return it for paid users while they continue to tweak GPT-5.

Why it matters: GPT-5 was supposed to be a world-changing step up — but instead it feels like “villagers gathering outside of Dr. Frankenstein’s castle.” While the new model may show big improvements in benchmarks, it’s clear that’s not the only thing that matters to a huge user base leveraging AI for a vast variety of use cases.

TOGETHER WITH VANTA

🔍 The AI security blind spots you can't afford to miss

The Rundown: Vanta's free AI Security Assessment gives companies at any stage of AI adoption a structured framework to identify, evaluate, and mitigate potential security vulnerabilities before they become costly problems.

This comprehensive resource helps you:

• Establish governance protocols across your entire AI program

• Address critical considerations from data privacy to incident management

• Demonstrate your AI compliance to potential customers and partners

Get your free assessment and join over 12,000 organizations using Vanta to build trust through transparent security practices.

AI & INVESTING

💰 Ex-OpenAI researcher raises $1.5B for AI hedge fund

Image source: For Our Posterity

The Rundown: Former OpenAI researcher Leopold Aschenbrenner just reportedly raised over $1.5B in funding for his ‘Situational Awareness’ AI-focused hedge fund, despite having zero professional investing experience.

The details:

• Aschenbrenner was part of OpenAI’s superalignment team and was one of two employees fired in April 2024 after being accused of leaking sensitive info.

• He later published a viral essay called ‘Situational Awareness’ (which the fund is named after) detailing his predictions around AGI and AI progress.

• Aschenbrenner’s fund has posted a 47% return in the first half of 2025, outpacing the S&P 500 despite no prior investment experience.

• The fund has focused on AI-tangential investments, including semiconductor, infrastructure, and power companies positioned to benefit from AI’s rise.

Why it matters: The AI boom is reshaping the hedge fund industry, and those closest to the tech might have a new seat at the table over those with traditional finance acumen when it comes to visionary bets. Everyone wants exposure to the AI rush, but few have the true foresight on where the industry will evolve to.

AI TRAINING

🤖 Automate meeting prep with AI Agents

The Rundown: In this tutorial, you will learn how to use Lindy’s new agent builder to create an AI agent that automatically researches companies 30 minutes before your meetings and drafts email summaries — by simply using your descriptions.

Step-by-step:

1. Go to Lindy AI and describe your desired meeting prep agent

2. Choose “extract company from event title” and “Gmail draft” for email destination

3. Customize the workflow by reviewing the Calendar, Search and Email sequence and replacing any integrations if needed

4. Test your agent with a sample meeting, then click “Deploy” to activate for all future calendar events

Pro tip: Start simple and gradually add more research sources or customize the email format as you refine your agent’s performance.

PRESENTED BY IBM

🚀 Lockheed Martin partners with IBM for AI

The Rundown: Lockheed Martin’s AI takeoff began with data cleanup — cutting its tools by 50% and replacing them with a single unified system that runs on IBM’s watsonx.data to drive faster, smarter engineering.

Learn how Lockheed Martin leveraged watsonx to:

• Help 10,00 engineers build scalable AI products in its AI Factory

• Respond to employee questions using AI, boosting accuracy by 20%

• Overcome data silos to create a single accessible, connected data environment

Read the full story.

GOOGLE & NASA

🚀 Google, NASA’s AI doctor for astronauts in space

Image source: Ideogram / The Rundown

The Rundown: Google and NASA are partnering to develop an AI medical assistant, dubbed Crew Medical Officer Digital Assistant, with the ability to diagnose and treat astronauts during deep-space missions where Earth communication is delayed.

The details:

• CMO-DA will run on Google Cloud’s Vertex AI platform using open-source models like Llama 3 and Mistral-3 Small.

• The model achieved up to 88% accuracy for diagnosing injuries in tests, while addressing gaps like no real-time comms and the inability to evacuate.

• NASA plans to expand CMO-DA with ultrasound imaging, biometric data sources, and training on space-specific health conditions.

• The system could also eventually support remote healthcare advances (on Earth), providing medical assistance to underserved and isolated areas.

Why it matters: While we aren’t at HAL-9000 systems yet, the next expert doctor aboard space flights looks like it will be AI. Given the barriers like the comms issues with Earth, AI makes for a big upgrade in aiding astronauts in critical medical situations in space, while also potentially driving breakthroughs in telemedicine back home.

QUICK HITS

🛠️ Trending AI Tools

• 🧠 Retellio - AI agents that automate customer research - giving B2B marketing teams testimonials, product feedback & competitor intel 10x faster*

• 🗣️ MiniMax Speech 2.5 - Voice cloning AI with support for 40 languages

• 🦑 Jules - Google’s asynchronous coding agent, now out of beta

• 🎆 Qwen-Image - New open-source image model with SOTA text rendering

_{*Sponsored Listing}

📰 Everything else in AI today

xAI rolled out its next-gen Grok 4 for free to all users worldwide for a limited time, also announcing a new ‘long press’ feature to turn images into video with Grok Imagine.

OpenAI’s o3 swept the Kaggle AI chess tournament, winning every game against rivals, including DeepSeek R1, Grok-4, and Gemini 2.5 Pro, to take the gold medal.

Roblox open-sourced Sentinel, a new AI model designed to filter inappropriate chat messages and protect children on the platform.

Microsoft released Copilot 3D, a new AI tool that converts images into usable 3D models in a single click for integrations with games, animation, VR/AR, and more.

SoftBank announced the acquisition of Foxconn’s U.S. electric vehicle plant in Ohio, with plans to launch its Stargate data center at the location.

Elon Musk confirmed that Tesla is closing its Dojo Supercomputer team to instead focus on its advanced AI chips, with the team’s VP, Pete Bannon, leaving the company.

Bloomberg Apple insider Mark Gurman revealed that Apple AI researcher Yun Zhu is leaving for Meta’s MSL, the fifth departure from the foundation models team.

COMMUNITY

🤝 Community AI workflows

Every newsletter, we’re showcasing how a reader is using AI to work smarter, save time, or make life easier.

Today’s workflow comes from reader Michael M. from the United Kingdom:

"I work on my own so I use it [ChatGPT] like a colleague, it knows I'm a data analyst so all of its answers are in that field, and whatever I work on, I talk to it like a colleague whether that's coding, client challenge scenarios, new projects, visualisation techniques or just conversations about something happening personally. It never receives company data, but it's been the best thing for both work challenges and loneliness in my work, and my productivity has skyrocketed as a result."

How are you using AI in your work? Tell us here.

🎓 Highlights: News, Guides & Events

• Read our previous newsletter: OpenAI launches GPT-5 to all ChatGPT users

• Today’s AI tool guide: How to automate meeting prep with AI Agents

• RSVP to our next workshop @ 2PM EST Wednesday: AI Agents for Customer Service

See you soon,

Rowan, Joey, Zach, Shubham, and Jennifer—the humans behind The Rundown

Good morning, tech enthusiasts. Apple is tossing another $100B into the U.S. manufacturing pot, bringing its total stateside bet to $600B — announced, no less, alongside President Trump.

It’s a high-stakes move to rein in a sprawling global supply chain and dodge rising tariffs. But while the move may steady iPhone supply, one question lingers: Will Apple users foot the bill?

In today’s tech rundown:

• Apple’s $600B deal with Trump

• Eye drops to replace reading glasses

• DoorDash rides a $100B high

• WhatsApp removes 6.8M scam accounts

• Quick hits on other major tech news

LATEST DEVELOPMENTS

APPLE

🍏 Apple’s $600B deal with Trump

Image source: Ideogram/The Rundown

The Rundown: Apple just ramped up its U.S. industrial ambitions in a big way, unveiling a fresh $100B investment to boost domestic manufacturing and supplier partnerships, and of course, sidestep looming tariffs on imported chips and devices.

The details:

• The expansion, announced alongside President Donald Trump, brings Apple’s total U.S. commitment to $600B over the next four years.

• The move comes as Trump threatens a 100% import tariff on computer chips, but offers exemptions to companies committed to producing in the U.S.

• It’s a win-win for the White House and Apple — Trump gets a headline achievement for his trade agenda, while Apple future-proofs against tariffs.

• Apple will use $2.5B to ensure every iPhone and Watch cover glass is made by Corning in the U.S., and inked multi-year contracts with domestic chipmakers.

Why it matters: Despite the massive investment, Apple isn’t committing to full U.S. iPhone assembly — final device manufacturing will mostly stay overseas for now, with domestic focus on components, glass, chips, and datacenter buildouts. Nvidia, Intel, and IBM have also announced huge domestic investments spurred by tariff pressures.

BIOTECH INNOVATIONS

👁️ Eye drops to replace reading glasses

Image source: Ideogram/The Rundown

The Rundown: The FDA has greenlit Vizz 1.44% eye drops, which could be a game-changer for anyone squinting at a menu or reaching for reading glasses after 45. Powered by aceclidine, this is the first once-daily drug of its kind targeting presbyopia.

The details:

• The Vizz drops contract the iris sphincter muscle, shrinking the pupil to under 2mm and creating a “pinhole effect” reminiscent of a camera lens.

• The result? Extended depth of field — sharper near vision — without that myopic “zoomed-in” side effect seen with older, less selective drops.

• Clinical trials found that a single dose of Vizz kicks in within 30 minutes, delivering 10 hours of crisp close-up vision.

• Safety data from 30K-plus treatment days showed only mild, transient effects like brief dim vision or eye irritation.

Why it matters: The real innovation with Vizz is its tech-forward “selective miotic” strategy compared to previous topical treatments. Available in single-use vials and slated to hit the U.S. market by late 2025, Vizz is poised to disrupt the $100B vision correction market just as presbyopia accelerates for millions of aging digital natives.

DOORDASH

🥡 DoorDash rides a $100B high

Image source: Ideogram/The Rundown

The Rundown: DoorDash is dropping earnings this week as a company reborn. What started as a scrappy food delivery startup out of Palo Alto now commands a whopping 67% of the U.S. market, leaving Uber Eats and Grubhub eating its dust.

The details:

• DoorDash now boasts $10.7B in annual revenue and a market cap north of $100B, with 2024 its first profitable year after 12 years of business.

• By scaling rapidly during the COVID lockdowns, the company became a fan favorite with millions of new users, many of whom became repeat customers.

• Its early bet on expanding to smaller cities and suburbs paid off, creating new markets and insulating itself from hyper-competitive large cities.

• Now, DoorDash is moving aggressively to become a comprehensive last-mile logistics platform, partnering with thousands of local retailers.

Why it matters: With profit now on the books and a growing premium subscription model boasting 22M members, DoorDash is shifting from disruptor to industry leader. The company is expanding its reach, eyeing every restaurant, convenience store, and consumer impulse as part of its expansive logistics network.

WHATSAPP

🔒 WhatsApp removes 6.8M scam accounts

Image source: Meta

The Rundown: WhatsApp’s fight against digital fraud escalated this year, as parent company Meta announced the removal of 6.8M accounts tied to global scam rings, with many operating by organized crime rings using forced labor in Southeast Asia.

The details:

• Meta said many of these accounts were linked to scam centers operated by organized crime groups that relied on forced labor in guarded campuses.

• Many scam operations are run out of Cambodia, Myanmar, and Laos, where trafficked workers are initially lured through fake job ads.

• The rings use psychological manipulation at scale: building trust with victims via chats, then luring them into bogus crypto and investment schemes.

• The campaigns have become more sophisticated, using generative AI to craft recruitment pitches before moving victims onto channels like Telegram.

Why it matters: The targeted takedown, carried out in partnership with OpenAI, marks a new phase in Meta’s fight against hyper-organized, cross-border digital fraud. The company has also deployed early-warning tools for WhatsApp, including alerts for suspicious group invites and direct messages from unknown contacts.

QUICK HITS

📰 Everything else in tech today

OpenAI unveiled GPT-5, its latest flagship AI model, available to all its 700M users, including those on the free tier.

Elon Musk announced that X will start inserting ads into responses from its Grok AI chatbot, aiming to revive the platform’s troubled advertising business.

Cognition, the AI coding startup that just acquired Windsurf, is giving remaining staff a choice between severance or staying for a new 80-hour/six-day in-office workweek.

Donald Trump on Thursday demanded the resignation of Intel CEO Lip-Bu Tan, alleging that Tan’s position and ties to China present conflicts of interest.

Tinder plans to roll out Gen Z-focused tools and a redesign to boost engagement and encourage more users to pay for its service after seeing a 7% dip in paying users.

Pinterest reports that it has a growing user base of male audience, now accounting for 171M users, driven by Gen Z's interest in wellness, AI, and fatherhood content.  

Linda Yaccarino, who resigned as chief executive of Elon Musk’s X in July, has been appointed CEO of eMed Population Health, the Miami-based telehealth startup.

Mars is partnering with biotech firm Pairwise to use CRISPR gene-editing technology to quickly develop hardier cocoa plants resistant to disease and climate challenges.

COMMUNITY

🎥 Join our next live workshop

Join our next workshop today at 4 PM EST with Dr. Alvaro Cintas, The Rundown’s AI professor. By the end of the workshop, you’ll confidently be able to launch your own fully-private, open-weight ChatGPT alternative, ready for production or personal tinkering

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’s long-awaited GPT-5 is finally here, ushering in a new era of advanced intelligence for the world.

While top-tier performance was something many of us expected from GPT-5, the real surprise turned out to be accessibility — with OpenAI making the flagship system available to all its 700M users, including those on the free tier.

In today’s AI rundown:

• OpenAI releases long-awaited GPT-5 models

• Google open-sources AI to understand animal sounds

• Automate business workflows with Perplexity Comet

• MIT’s AI predicts protein location in any cell

• 4 new AI tools & 4 job opportunities

LATEST DEVELOPMENTS

OPENAI

🧠 OpenAI releases long-awaited GPT-5 models

Image source: OpenAI

The Rundown: OpenAI has finally launched its flagship GPT-5 models, replacing GPT‑4o, 4.1, 4.5, o3, o4-mini, and marking a new age of intelligence that is smarter, faster, and accessible to everyone.

The details:

• The model family includes three variants—GPT-5, GPT-5 Pro, and GPT-5 Mini—with GPT-5 coming to free users under usage limits (higher limits for Plus and unlimited for Pro users).

• GPT-5 switches thinking on/off using a real-time router (based on the task) and delivers SOTA performance on coding, writing, math, and health benchmarks.

• Its Pro version, coming to Pro users on $200/month plan, thinks longer with scaled parallel test-time compute to provide the most comprehensive answers.

• The smaller GPT-5 Mini, meanwhile, comes into play only when free and Plus users hit rate limits, ensuring the rest of their queries are handled.

• OpenAI also says these models hallucinate less, are less deceptive, and communicate with honesty when they can or can not handle a task.'

Why it matters: OpenAI’s move to replace its flurry of models with a unified GPT-5 simplifies user experience and gives everyone a PhD-level assistant, bringing elite problem-solving to the masses. The only question now is how long it can hold its edge in this fast-moving AI race, with Anthropic, Google, and Chinese giants all catching up.

Share your use cases: We’d love to hear how you’re using ChatGPT 5. Let us know for a chance to be featured in next week’s newsletter to 1M+ readers.

TOGETHER WITH TRIPLE WHALE

🙌 AI that gets ecommerce

The Rundown: Triple Whale’s Moby Agents go beyond dashboards and manual analysis to uncover blind spots, spotlight what matters, and optimize every stage of the shopper’s journey from first click to repeat purchase.

Grab The Ultimate AI Agents Playbook for Ecommerce Brands and learn how to:

• Catch wasted ad spend and performance dips before they hurt your bottom line

• Stay ahead of demand with smarter, AI-powered inventory forecasting

• Write high-converting emails with insights into what messaging actually works and why

Download the playbook today.

AI FOR GOOD

🕊️ Google open-sources AI to understand animal sounds

Image source: Google DeepMind

The Rundown: Google DeepMind just open-sourced an upgraded version of Perch, an AI model that helps scientists analyze massive amounts of wildlife audio, making it easier to track endangered species in diverse environments.

The details:

• Perch can now handle a wider range of species and environments, from forests to coral reefs, using twice the training data of the version released in 2023.

• It can disentangle complex soundscapes over thousands or millions of hours of audio, answering questions from species counts to newborn detections.

• The model also comes with open-source tools that combine vector search with active learning, enabling the detection of species with scarce training data.

• With this system, conservationists don’t have to scour through massive volumes of bioacoustic data when planning measures to protect ecosystems.

Why it matters: From finding elusive bird populations in Australia to accelerating honeycreeper monitoring in Hawaii by 50x, open-source access to the upgraded Perch marks a breakthrough in wildlife conservation. AI's speed and precision can give scientists an edge to save species well before they disappear.

AI TRAINING

💼 Automate business workflows with Perplexity Comet

The Rundown: In this tutorial, you’ll learn how to use Perplexity’s new Comet Shortcuts feature to create automated workflows that can be triggered anywhere in the browser with a simple command.

Step-by-step:

1. Head over to your Perplexity Comet browser (if you’re a member of The Rundown University, you get early access perks)

2. Type / anywhere in Comet and select “+ Create a shortcut.”

3. Enter the prompt: “Analyze this product and search for real, unbiased assessment. Give me non-sponsored testimonials and potential downsides,” and add the name vendor-review

4. Test with /vendor-review on any vendor website

Pro tip: Use this shortcut to fact-check product reviews for bots in real-time before spending money.

PRESENTED BY WARP

🏆 Warp delivers high limits and zero lockouts

The Rundown: Warp is the top-performing AI coding agent in benchmarks, trusted by over 600K active developers. Warp’s new Lightspeed plan offers the most generous monthly limits of any AI plan on the market, with access to the best AI models across providers.

Why Lightspeed:

• Access the best models like Opus, Sonnet, and GPT 4.1, all in one subscription

• 50K AI requests included monthly

• Seamless pay-as-you-go when you exceed limits

Try Warp for free – and get 20% off your 1st month with code THERUNDOWN.

AI RESEARCH

🧠 MIT’s AI predicts protein location in any cell

Image source: MIT

The Rundown: Researchers from MIT, Harvard, and the Broad Institute have developed PUPS, an AI system that predicts the exact location of virtually any protein within a single human cell, marking a major shift in disease diagnosis and treatment.

The details:

• PUPS uses a protein language model to capture the structure of a protein, and an inpainting model to understand the type, features, and stress state of a cell.

• Using insights from both models, it generates a highlighted cell image showing the predicted protein location at the cell level.

• It can even work on unseen proteins and cell types, flagging changes caused by mutations not included in the Human Protein Atlas.

• In tests, PUPS consistently outperformed baseline AI methods, showing lower prediction error across all tested proteins and maintaining accuracy.

Why it matters: Until now, pinpointing a protein’s location in a specific cell meant long lab work was limited to known proteins. PUPS removes those barriers by mapping any protein in any cell type, accelerating disease research, drug discovery, and enabling exploration of previously uncharted cellular biology.

QUICK HITS

🛠️ Trending AI Tools

• 🧠 ChatGPT - OpenAI AI assistant, now enhanced with flagship GPT-5 models

• 🔍️ NuMarkdown - An 8B parameter, open-source reasoning model for OCR

• 📽️ Midjourney - Now with an HD mode for 4x more pixels in video generations

• 🤖 North - Cohere’s secure and customizable AI agent platform

💼 AI Job Opportunities

• 🧑‍💻 The Rundown - Product Marketing

• 💰 Grammarly - Compensation Systems Analyst

• 📊 Databricks - Strategy & Execution Manager

• 🤖 Meta - Director, Robotics Operations

📰 Everything else in AI today

OpenAI added GPT-5 models in the API and introduced four new personalities to ChatGPT, along with a more advanced voice mode and chat customizations.

xAI plans to add ads in Grok’s responses, with Elon Musk saying, “If a user’s trying to solve a problem, then advertising the specific solution would be ideal,” he said.

Elon Musk also said on X that xAI will open-source its Grok 2 AI model next week, following OpenAI’s move to launch its first open models after GPT-2 in 2019.

The Browser Company launched a $20/month subscription for its AI browser Dia, providing unlimited access to chat and skills features and taking on Perplexity’s Comet.

Microsoft added GPT-5 to its Copilot AI assistant with a new smart mode that automatically switches to the flagship model based on the task at hand.

U.S. President Donald Trump’s Truth Social launched Truth Search AI, a Perplexity-powered AI search feature that delivers information from select sources.

MiniMax dropped Speech 2.5, its new voice cloning AI that supports 40 languages and can mimic voice while preserving elements like accent, age, and emotion.

COMMUNITY

🎥 Join our next live workshop

Join our next workshop today at 4 PM EST with Dr. Alvaro Cintas, The Rundown’s AI professor. By the end of the workshop, you’ll know how to run your own fully private, open-weight ChatGPT alternative locally on your computer.

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

See you soon,

Rowan, Joey, Zach, Shubham, and Jennifer—the humans behind The Rundown

Good morning, robotics enthusiasts. Unitree just dropped a viral teaser of its A2 Stellar Explorer, a sleek robot dog that backflips and smashes through windows like it’s auditioning for Mission: Impossible.

The video stirred online buzz and helped send Chinese robotics stocks climbing. But are we watching the future of mobility, or just really expensive parkour?

In today’s robotics rundown:

• Unitree’s ultra-fast stunt robodog

• OpenMind’s ‘Android for robots’

• Japanese hotel run by robots

• Robots rebuild homes after the LA fires

• Quick hits on other robotics news

LATEST DEVELOPMENTS

UNITREE

🤖 Unitree’s ultra-fast stunt robodog

Image source: Unitree

The Rundown: China’s Unitree just released a viral video teasing its new mechanical quadruped, the A2 Stellar Explorer — a fast, back-flipping, window-smashing robot dog that sent the robotics company’s stock soaring, Bloomberg reports.

The details:

• In the clip, the A2 is shown smashing through glass, hauling a man on its back, spinning on one leg, and scaling 45-degree inclines.

• The robodog weighs just 37kg (about 81.6 lb.) with batteries, and can support up to a 100 kg (220 lb.) person on its back and carry 25 kg (55 lb.) for hours.

• With 12 DoF and 180 Nm torque, it also reaches speeds up to 5m/s (11.2mph), besting Boston Dynamics’ Spot by 8 miles per hour.

• Under the hood, an 8-core CPU and Intel Core i7 enable high-level autonomy on the robot, with support for over-the-air updates.

Why it matters: Targeted for site inspection, logistics, emergency response, and research, A2’s robust build, network interfaces (Wi-Fi 6, Bluetooth 5.2), and smart OTA updates set a new benchmark for practical, next-gen quadruped robots. While Unitree hasn’t yet revealed the price, it’s estimated at a tidy $10K.

TOGETHER WITH POSTMAN

🏆 Build something incredible to win

The Rundown: Turn your API expertise into $1,000 and prizes in Postman's newest challenge. Create interactive Notebooks to showcase AI-powered solutions. After all, AI + APIs is a match made in heaven.

The prizes include:

• $1,000 for multiple categories

• $2,500 + for Creator Fellowship Grant

• $1,000 for Community Favorite and more

Join the challenge and submit your work by August 24 to be considered.

OPENMIND

🧠 OpenMind’s ‘Android for robots’

Image source: OpenMind

The Rundown: Silicon Valley startup OpenMind is crafting OM1, a hardware-agnostic, open-source operating system that aims to bring the power and flexibility of Android to humanoids worldwide.

The details:

• Founded by Stanford professor Jan Liphardt, OpenMind says its OM1 platform supports all types of robots, freeing developers from vendor lock-in.

• It’s engineered to empower robots with context awareness, enabling them to interpret complex environments and adapt quickly to unpredictable situations.

• Backed by $20M in fresh funding, OpenMind just unveiled its FABRIC protocol that lets robots verify each other’s identities and share skills on the fly.

• If one robot learns a new task (like speaking a new language), FABRIC enables that skill to be instantly and securely transmitted to others on the network.

Why it matters: OM1 stands out in the crowded robot OS field by offering a universal, open-source, and hardware-agnostic platform, in contrast to the fragmented, vendor-locked approaches of rivals like Figure AI, Boston Dynamics, and even the modular but research-centric ROS 2. Its unique approach promises quick ecosystem growth.

HOTEL BOTS

🛎️ Japanese hotel run by robots

Image source: Henn na Hotel

The Rundown: At Japan’s Henn na Hotel, the first thing you’ll notice is what’s missing: people. Instead, you’re checked in by robots — from humanoids to dinosaur bots — setting the tone for a stay on the bleeding edge of automation, Wired reports.

The details:

• Henn na Hotel is both a functional lodge and a showcase for robotics, serving as a testbed for future service automation.

• The robots communicate in multiple languages, using natural-sounding, synthesized voices and programmed etiquette that includes bows and nods.

• Room keys are dispensed by machines, and many rooms unlock with facial recognition, eliminating the need for cards or mechanical keys.

• A robotic bellhop delivers luggage to guest rooms, while a bedside voice-activated “tulip” robot controls lighting, temperature, and concierge functions.

Why it matters: Guests say the robotic presence at the hotel is both fascinating and eerie. Yet, behind the sci-fi veneer, the charm occasionally slips: robots can stutter over complex queries or react awkwardly to human quirks, prompting some guests to seek out the on-call human staff stationed to intervene if needed.

ROBOTICS FOR GOOD

🏠 Robots rebuild homes after LA fires

Image source: ABB Robotics

The Rundown: After wildfires destroyed nearly 9,700 Southern California homes, ABB Robotics and Cosmic Buildings are using AI-powered robotic microfactories to rapidly rebuild fire-resistant homes on-site.

The details:

• Advanced ABB robot arms, guided by AI and RobotStudio digital twin software, are making structural wall panels, digitally simulating assembly beforehand.

• Each robotic build is digitally tailored to homeowners’ needs, allowing for unique designs without slowing down the construction process.

• Meanwhile, built-in computer vision and AI analytics allow robots to detect flaws or issues in the assembly process.

• The homes are being constructed with non-combustible, fire-resistant materials with designs exceeding California’s wildfire safety standards.

Why it matters: The initiative aims to complete at least 100 new homes by 2027, serving as a scalable blueprint for robot-accelerated disaster recovery in other affected regions. This partnership isn’t just about rebuilding faster — it's pioneering a smarter, greener way for communities worldwide to bounce back from catastrophe.

QUICK HITS

📰 Everything else in robotics today

A robot crab named Wavy Dave helped scientists show that male fiddler crabs adjust their waving displays when they see a rival (even a robotic one).

Zoox, Amazon’s autonomous vehicle company, received U.S. regulatory approval to operate its electric robotaxis without steering wheels on public roads.

Kyle Vogt, CEO of The Bot Company, believes home robots have immense potential, but argues chores like laundry and dishwashing aren’t the right place to begin.

Researchers engineered a thin, muscle-inspired robotic actuator that mimics human motion, paving the way for robots capable of delicate surgery.

Coco Robotics is teaming up with Prime Video to turn its LA delivery robots into mobile billboards to promote films.

More than 100 robotic products — nearly double last year’s count — will debut at the 2025 World Robot Conference in Beijing from August 8-12.

A new robot store in Beijing is stocking lifelike Einstein replicas and AI-powered chess partner robots, marking a major push for consumer robots in China, Reuters reports.

University of Florida researchers developed solar-powered robotic rabbits to lure ecosystem-threatening Burmese pythons out of hiding in the Everglades.

An Argentine-American team is livestreaming the first-ever robotic exploration of the Mar del Plata canyon, revealing never-before-seen deep-sea life to viral fans.

COMMUNITY

🎥 Join our next live workshop

Join our next workshop this Friday, August 8th, at 4 PM EST with Dr. Alvaro Cintas, The Rundown’s AI professor. By the end of the workshop, you’ll confidently be able to launch your own fully private, open-weight ChatGPT alternative, ready for production or personal tinkering.

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 wants to be the AI backbone of the U.S. government, and it’s going all in to undercut rivals to get there.

With a new initiative offering ChatGPT Enterprise to federal agencies for just $1 per agency, the AI giant is making a bold push for long-term government adoption, but it’s also one that could change expectations around enterprise pricing.

Reminder: Our next live workshop is tomorrow at 4 PM EST. Join and learn how to run OpenAI’s new open weight models locally on your computer. RSVP here.

In today’s AI rundown:

• OpenAI offers ChatGPT at $1 for U.S. agencies

• Google launches AI tutoring mode for students

• Create Claude Code subagents for your projects

• Microsoft’s self-adapting AI tackles scientific problems

• 4 new AI tools & 4 job opportunities

LATEST DEVELOPMENTS

OPENAI

🇺🇸 OpenAI offers ChatGPT at $1 for U.S. agencies

Image source: The Rundown

The Rundown: OpenAI just announced that the enterprise version of ChatGPT, including access to advanced models with additional security features, will be available to all federal agencies for just $1 per agency for the next year.

The details:

• The discounted pricing stems from OpenAI's partnership with the U.S. General Services Administration, the government's central purchasing authority.

• It will also be accompanied by unlimited use of OpenAI’s advanced models and features, like Deep Research, for an additional 60-day period.

• The company hopes these efforts will help government officials cut down red tape and paperwork, making services “faster, easier, and more reliable.”

• To help federal employees get started, OpenAI is also setting up a dedicated government user community with tailored training resources.

Why it matters: While offering ChatGPT Enterprise for virtually free shows OpenAI’s willingness to embed itself deep into government workflows, it could also spark a high-stakes race, with other GSA-approved rivals like Anthropic and Google likely following suit and unleashing a wave of aggressive AI offers aimed at federal adoption.

TOGETHER WITH POSTMAN

🏆️ Build something incredible to win

The Rundown: Turn your API expertise into $1,000 and prizes in Postman's newest challenge. Create interactive Notebooks to showcase AI-powered solutions. After all, AI + APIs is a match made in heaven.

The prizes include:

• $1,000 for multiple categories

• $2,500 + for Creator Fellowship Grant

• $1,000 for Community Favorite and more

Join the challenge and submit your work by August 24 to be considered.

GOOGLE

🧑‍🎓 Google launches AI tutoring mode for students

Image source: OpenAI

The Rundown: As OpenAI courts federal agencies with $1 licenses, Google is pushing into the education sector with a new Guided Learning mode for Gemini and free access to its $250/month AI Pro Plan for college students.

The details:

• Following ChatGPT’s Study Mode, Gemini’s Guided Learning acts as a learning partner, offering step-by-step guidance instead of giving direct answers.

• Google has worked with educators and learning experts to ensure the AI helps students work through problems, building their critical thinking skills.

• The mode also includes tools that use images, videos, and interactive quizzes to help students test their knowledge as they learn new concepts.

• Google is making its AI Pro Plan free for students in select countries, including the U.S., and investing $1B over 3 years for AI training at U.S. colleges.

Why it matters: Amid concerns that AI may hurt the learning process with direct answers—including an MIT study highlighting its impact on students’ cognition—both Google and OpenAI are now repositioning their AI tools with study tutor features designed to build, not bypass, critical thinking.

AI TRAINING

🤖 Create Claude Code subagents for your projects

The Rundown: In this tutorial, you’ll learn how to use Claude Code's sub-agents to create custom AI assistants with specific roles and instructions, each with their own context window and expertise.

Step-by-step:

1. Install Claude Code with npm install -g @anthropic-ai/claude-code and run claude in your project directory

2. Type /agents, select “Create a new agent” and choose “Generate with Claude”

3. Describe your agent's purpose and configure tools, model (Opus/Sonnet/Haiku), and color theme

4. Save your agent, and it will automatically activate for relevant tasks.

Pro tip: Create multiple specialized sub-agents (testing, reviewing, bug-fixing) to maintain focused expertise and separate context windows for each task.

PRESENTED BY TOGETHER AI

⚙️ Automate, debug, and solve with SOTA open source models

The Rundown: Together AI's Inference & Fine-Tuning platform lets developers deploy open source models at scale, offering both serverless and dedicated endpoints for models like Kimi K2 — a new SOTA agentic intelligence model with 1T parameters, 128K context, and native tool use.

With Together AI, you’ll experience:

• High-performance infrastructure designed for the lowest possible latency

• Developer-friendly APIs that simplify integration

• An extensive model library featuring the top open-source AI models like Kimi K2

Try Kimi K2 now and get all the benefits of open-source AI without the complexity.

AI RESEARCH

🧠 Microsoft’s self-adapting AI tackles scientific problems

Image source: Microsoft

The Rundown: Microsoft just announced CLIO (Cognitive Loop via In-situ Optimization), a breakthrough framework that enables non-reasoning LLMs to develop their own thought patterns and adapt their reasoning in real-time.

The details:

• While most reasoning models use strategies and actions that are pre-baked during post-training (before deployment), CLIO creates a “steerable” AI system.

• It builds and refines its reasoning through self-reflection at runtime, creating its own feedback loops to explore ideas, manage memory, and flag uncertainties.

• With the self-adapting behavior, users get full control to set thresholds for uncertainty, change reasoning paths, or re-execute them entirely.

• On Humanity’s Last Exam, CLIO boosted GPT-4.1’s accuracy on text-only biomedical questions from 8.55% to 22.37%, surpassing o3 (high).

Why it matters: With CLIO’s performance leap—plus built-in explainability, memory control, and tunable reasoning—Microsoft has shown that LLMs don’t need to be “finished” in training. In high-stakes scientific domains where trust and rigor matter, a continually steerable AI might be just what teams need to accelerate discovery.

QUICK HITS

🛠️ Trending AI Tools

• ✨ Anything - Launch beautifully designed mobile apps and web apps from just a prompt, no coding or extra tools required*

• ⚙️ Bing Image Creator - Now with OpenAI’s latest GPT-4o image model

• 📽️ Endex - An AI agent that lives inside Excel

• 📝 Kitten TTS - A 15M parameter text-to-speech model that runs without GPU

^{*Sponsored listing}

💼 AI Job Opportunities

• 🧠 Meta - AI Research Scientist, Computer Vision

• 🎨 Descript - Product Designer

• 🤝 Deepmind - People Partner, GenAI

• 📈 UiPath - Territory Sales Manager

📰 Everything else in AI today

Midjourney added a new HD video mode for its Pro and Mega subscribers, saying it delivers 4x more pixels in outputs than the default SD mode and is 3.2x more expensive.

Google’s asynchronous, agent-based coding tool, Jules, is finally out of beta, but the free plan usage limits have been reduced to 15 individual daily tasks from 60.

Anthropic added automated security reviews in Claude Code with a new /security-review command that checks for vulnerabilities directly from the terminal.

xAI’s Grok Imagine, a video and image generation tool, is now rolling out in early access on Android devices, via the official Grok app.

OpenAI announced a “LIVE5TREAM” for Thursday, 10 am PT (today), teasing the launch of GPT-5, its long-awaited flagship model, with possibly a Nano and Mini version.

OpenAI is also reportedly holding early talks about a secondary share sale for current and former employees at a valuation of $500B.

Chai Discovery, the AI lab building frontier models for molecular design, raised $70M series A from Menlo Ventures and the Anthology Fund.

COMMUNITY

🎥 Join our next live workshop

Join our next workshop this Friday, August 8th, at 4 PM EST with Dr. Alvaro Cintas, The Rundown’s AI professor. By the end of the workshop, you’ll know how to run your own fully private, open-weight model locally on your computer.

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

See you soon,

Rowan, Joey, Zach, Shubham, and Jennifer—the humans behind The Rundown

Good morning, AI enthusiasts. The age of embodied AI is all about training in simulations…But what happens when some scenarios are too hard to build, or even imagine?

Google’s new Genie 3 just cracked that problem with the ability to generate rich, playable environments that evolve in real-time as AI agents (or users) explore them — unlocking a whole new frontier of infinite training.

Share your input: We’re committed to continuously improving The Rundown, and we’d love your feedback to help shape its future.

In today’s AI rundown:

• Google’s Genie 3 interactive world model

• OpenAI finally launches open-source models

• Turn any document into study video presentations

• Anthropic releases Claude Opus 4.1

• 4 new AI tools & 4 job opportunities

LATEST DEVELOPMENTS

GOOGLE DEEPMIND

🌍 Google’s Genie 3 interactive world model

Image source: Google DeepMind

The Rundown: Google DeepMind just announced Genie 3, a new general-purpose world model that can generate interactive environments in real-time from a single text prompt, complete with surrounding and character consistency.

The details:

• With Genie 3, users can generate unique, 720p environments with real-world physics and explore them in real-time, with new visuals emerging at 24fps.

• The model’s visual memory goes up to one minute, enabling it to simulate the next scene while ensuring consistency with the previous ones.

• To achieve this level of controllability, Google says, Genie computes relevant information from past trajectories multiple times per second.

• It also allows users to change the worlds as they go by inserting new characters, objects, or changing the environment dynamics entirely.

Why it matters: Genie 3’s consistent worlds, generated frame-by-frame in response to user action, isn’t just a leap for gaming and entertainment. They lay the foundation for scalable training of embodied AI, where machines can tackle the “what if” scenarios — like a path vanishing — by adapting in real time, just like humans.

TOGETHER WITH SŌKOSUMI

🤖 The first open marketplace for AI agents

The Rundown: Sōkosumi is the first open marketplace for autonomous AI agents, letting you hire specialized AI co-workers for content, research, design, and data tasks with just a single click. Built for professionals and enterprise teams who need results — not subscriptions.

The platform offers:

• MCP ready, multi-model agents (ChatGPT, DeepSeek, Mistral)

• Enterprise-grade security with GDPR compliance and SSO integration

• Pay-per-task pricing — no monthly subscriptions or hidden fees

Build your AI agent team today — redeem $100 in free credits with code RUN100.

OPENAI

🧠 OpenAI finally launches open-source models

Image source: OpenAI

The Rundown: OpenAI unveiled gpt-oss-120b and gpt-oss-20b, its long-awaited open-weight reasoning LLMs that match or exceed o4-mini and o3-mini in performance and are available for local deployment under an Apache 2.0 license.

The details:

• Available under Apache 2.0, the gpt-oss family, OpenAI’s first open LLMs since GPT-2 in 2019, instantly became #1 among 2M models on Hugging Face.

• The 120B variant performs on par with o4-mini on core benchmarks and exceeds on certain domains, while being deployable on an 80GB GPU.

• Meanwhile, the smaller 20B version is competitive with o3-mini, with suitability for local deployment on laptops with 16GB of memory.

• Both models feature adjustable reasoning (high, medium, low) and can handle agentic workflows, with function calling, web search, and Python execution.

Why it matters: After keeping its best models locked for years, OpenAI is finally living up to its name, giving developers access to near-frontier reasoning models they can run and modify in their own environments. It’s a major boost for the open-source ecosystem, which has been rapidly closing the gap with closed models.

From the community: We’re hosting a workshop this Friday on the key benefits of local models and how you can run gpt-oss models on your computer.

AI TRAINING

🎥 Turn any document into study video presentations

The Rundown: In this tutorial, you’ll learn how to use NotebookLM’s new ‘Video Overview’ feature to turn documents into AI-narrated video presentations with slides — pulling images, diagrams, quotes, and numbers directly from the content.

Step-by-step:

1. Visit NotebookLM, create a new notebook, and upload your documents

2. In the Studio panel on the right, click “Video Overview”

3. Optional: Click the three dots menu to customize your focus topics, target audience, or learning goals

4. Review your generated video and “Download” to save as MP4

Pro tip: Create multiple Video Overviews in one notebook, make versions for different audiences, or focus on different chapters of your content.

PRESENTED BY LOVART

🎨 Your next creative partner might not be human

The Rundown: Lovart has officially exited beta with its AI design platform built for visual collaboration. It features a creative reasoning agent who thinks with you, sources references, and builds brand systems in minutes. Designers say it feels more like working with a teammate than using a tool.

With Lovart, you can:

• Use the new ChatCanvas feature to visually collaborate with your agent, iterate, and refine in real time through natural language.

• Turn simple prompts into brand visuals, social content, videos, and even 3D models

• Rely on a specialized agent to maintain style consistency across materials

• Work faster with memory that learns how you design and adapts to your habits

Transform your creative process with the world's first design agent today.

ANTHROPIC

🤖 Anthropic releases Claude Opus 4.1

Image source: Anthropic

The Rundown: Anthropic released Claude Opus 4.1, an incremental upgrade for Opus 4, improving performance across real-world coding, in-depth research, and data analysis tasks — particularly those requiring attention to detail and agentic actions.

The details:

• Opus 4.1 brings a notable coding upgrade over its predecessor, boosting performance on SWE-bench Verified from 72.5% to 74.5%.

• Improvements are also seen across math, agentic terminal coding (TerminalBench), GPQA reasoning, and visual reasoning (MMMU) benchmarks.

• Customers cited real-world gains with the model, saying it excels at tasks like multi-file code refactoring and identifying correlations in codebases.

• Anthropic said the upgrade—available to paid users and businesses—marks the start of “substantially larger improvements” planned for its models.

Why it matters: With Opus 4.1, Anthropic is adding more momentum to what’s shaping up to be an exciting week for AI enthusiasts. The upgrades are a welcome gift, but with OpenAI’s GPT-5 potentially dropping any day, all eyes will be on how the company’s models hold its ground, especially in coding, where it has stood out.

QUICK HITS

🛠️ Trending AI Tools

• 🔄 Depot’s Claude Code Sessions - Persistent AI coding sessions that sync across teams & environments for seamless collaboration*

• ⚙️ Kaggle Game Arena - Benchmark to test LLMs on evolving strategic games

• 📽️ ChatGPT - OpenAI’s AI assistant, now with tools to detect mental distress

• 📝 Gemini Storybooks - Google’s AI now creates narrated storybooks

^{*Sponsored listing}

💼 AI Job Opportunities

• ⚙️ The Rundown - Growth and Content Strategist

• 📢 Groq - Product Marketing Manager, Sales Enablement

• 🛠️ Figure AI - Prototype Development Technician

• 🎭 Meta - Creative Director

📰 Everything else in AI today

ElevenLabs introduced Eleven Music, its multilingual music generation model with control over genre, style, and structure, and the option to edit both sounds and lyrics.

Google added a new Storybook feature to the Gemini app, allowing users to generate personalized storybooks about anything with read-aloud narration for free.

Perplexity acquired Invisible, a company developing a multi-agent orchestration platform, to scale its Comet browser for consumer and enterprise users.

Elon Musk shared Grok’s Imagine image and video generator is seeing massive interest, with 20 million images generated yesterday alone.

Alibaba released its Flash series of Qwen3-Coder and Qwen3-2507 models via API, with up to 1M-token context window and low pricing.

Shopify added new agent-focused features, including a checkout kit to embed commerce widgets into agents, low-latency global product search, and a universal cart.

COMMUNITY

🎥 Join our next live workshop

Join our next workshop this Friday, August 8th, at 4 PM EST with Dr. Alvaro Cintas, The Rundown’s AI professor. By the end of the workshop, you’ll know hot to run your own fully private, open-weight model locally on your computer.

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

See you soon,

Rowan, Joey, Zach, Shubham, and Jennifer—the humans behind The Rundown

Good morning, tech enthusiasts. Big Tech is throwing $344B at AI this year, turning the race for intelligence into a full-blown spending spree.

Microsoft, Amazon, Meta, and Alphabet are scaling like there's no ceiling, because in this gold rush, FOMO is more powerful than profit margins. Is this the dawn of a new era, or the makings of the next big bubble?

In today’s tech rundown:

• Big Tech pours $344B into AI spree

• Tesla awards Musk $29B in stock

• Space tech company Firefly eyes $6B valuation

• North Korean ‘deepfake’ IT workers surge

• Quick hits on other major tech news

LATEST DEVELOPMENTS

BIG TECH

💰 Big Tech pours $344B into AI spree

Image source: Ideogram/The Rundown

The Rundown: The spirit of FOMO has officially gone hyper-scale in Big Tech, as tech giants Microsoft, Amazon, Meta, and Alphabet are reportedly on track to funnel $344B into AI this year, not just to lead but to avoid being left behind.

The details:

• Meta is committing up to $65B toward AI in 2025, and aims to bring nearly 1 gigawatt of computing power online, backed by over 1.3M GPUs.

• Microsoft plans to spend over $30B in capital expenditures this quarter alone to scale its Azure cloud platform and expand data center capacity for AI.

• Amazon leads the pack with more than $100B allocated to AI for AWS, which CEO Andy Jassy calls a “once-in-a-lifetime” opportunity.

• Big Tech’s combined 2024 capex was roughly $230B, with this year’s figure marking a 60% spike.

Why it matters: Welcome to the new reality in tech: the fear of falling behind in the AI compute arms race has blown past former ideas of corporate restraint, as companies bet that whoever controls AI infrastructure will control the future. It’s a move that could either lay the groundwork for a new industrial era or inflate the next tech bubble.

TESLA

🤑 Tesla awards Musk $29B in stock 

Image source: Daniel Oberhaus/Wikimedia Commons

The Rundown: Tesla has greenlit a colossal $29B, 96M-share stock award for CEO Elon Musk, citing the “ever-intensifying AI talent war and Tesla’s position at a critical inflection point” as reasons for the payout.

The details:

• The grant is contingent on Musk remaining as Tesla’s CEO for at least two years, and the shares must be held for five years post-vesting.

• Tesla’s board justifies the move by citing Musk’s pivotal role in the company’s transition from an EV maker to AI, robotics, and robotaxi leader.

• Musk’s award comes at a time when AI salary offers are breaking records, with elite engineers at AI firms landing nine-figure hiring packages.

• The board’s rush follows a Delaware court ruling in January 2025 that invalidated Musk’s earlier 2018 pay package, citing its excessive size.

Why it matters: This is a staggering pay raise for Musk — already the world’s richest person — as Tesla sales and profits drop, and the company is losing market share (due in part to Musk’s behavior). But Tesla sees keeping him onboard as an existential necessity for the company’s future, even if it’s a risky bet on a single executive.

FIREFLY

🚀 Space tech company Firefly eyes $6B valuation

Image source: Firefly Aerospace

The Rundown: Firefly Aerospace, the Texas-based space tech company, has upped its IPO target to as much as $697M, eyeing a potential valuation north of $6B, as investor appetite grows in the fast-evolving “new space” sector.

The details:

• Firefly’s updated prospectus values it at up to $6.04B if the share price is at the high end of its new $41 to $43 per share range.

• The company’s Alpha and Blue Ghost launch vehicles are positioned for rapid, 24-hour missions, a unique operational capacity in the current launch market.

• Goldman Sachs, JPMorgan, and Wells Fargo are leading the underwriting syndicate, with the offering consisting of 16.2M shares.

• Firefly has $1.1B in contract backlog, including major government deals with NASA, the Department of Defense, SpaceX, and Northrop Grumman.

Why it matters: Firefly boasts tech street cred as the only private company to achieve a fully successful commercial Moon landing this March. Backed by $1.1B in backlog and hefty contracts, the firm posted a staggering 572% year-over-year revenue surge in Q1 2025 and is making big moves in its quest to be the “next SpaceX.”

CYBERSECURITY

🕵🏼‍♂️ North Korean ‘deepfake’ IT workers surge

Image source: Office of the President of the Russian Federation/Wikimedia Commons

The Rundown: Cybersecurity firm CrowdStrike uncovered a dramatic surge in North Korean operatives infiltrating Western tech companies under false identities, using remote IT jobs to funnel cash to the regime’s weapons programs.

The details:

• Over the past year alone, CrowdStrike identified more than 320 such incidents, a staggering 220% increase from the previous year.

• These state-trained developers, posing as freelancers or full-time hires, have been onboarded at hundreds of organizations, including Fortune 500 firms.

• Some North Korean contractors use “laptop farms” and US-based accomplices to access company portals from domestic IP addresses and mask locations.

• Employing generative AI and automation, a single worker can juggle multiple jobs, respond to English correspondence, and even automate code completion.

Why it matters: Law enforcement is cracking down, but CrowdStrike warns that new technology still allows North Korean operatives to bypass even advanced hiring checks. This leaves even the most security-minded tech firms exposed to insider threats that easily blend into global, remote teams.

QUICK HITS

📰 Everything else in tech today

Tesla was found partly liable for a 2019 fatal crash involving its Autopilot system and ordered to pay $200M in punitive damages to the victim’s family.

British startup CuspAI is reportedly in talks to raise more than $100M toward its goal of using AI models to discover new materials.

Lyft will expand into Europe by partnering with Baidu to launch Chinese-made Apollo Go robotaxis in Germany and the UK starting in 2026, pending regulatory approval.

OpenAI raised $8.3B in new funding at a $300B valuation as part of its larger plan to secure $40B this year, the New York Times reports.

Apple CEO Tim Cook, speaking at a rare all-hands meeting after the latest earnings, rallied employees by declaring that artificial intelligence is “ours to grab.”

Vietnam’s VinFast launched production at a $500M EV plant in Tamil Nadu, India, as the first step in a planned $2B investment across the country.

Illinois passed a law, signed by Governor JB Pritzker, that bans AI from acting as a standalone therapist and sets strict rules on its use by mental health professionals.

Neuralink is launching UK clinical trials with top hospitals to test its brain implant’s ability to help paralyzed patients control devices with their thoughts.

Electric air-taxi developer Joby Aviation is acquiring the passenger business of Blade Air Mobility for as much as $125M in cash and stock.

A University of Basel-led team extracted traces of the 1918 flu virus from a preserved lung of a Zurich teenager who died during the pandemic’s first wave.

COMMUNITY

🎥 Watch our last workshop

Check out our last live workshop with Dr. Alvaro Cintas, The Rundown’s AI professor. By the end of this workshop, you’ll have practical strategies to get the AI to do exactly what you want.

Watch it 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