🤝 Today’s edition is brought to you by AIOZ.

AIOZ Network’s new AIOZ AI platform launches its first developer challenge!
Compete, learn, and unlock the potential for token rewards on the DePIN Ecosystem for AI. Join the challenge… (more info later in the post)

🤓 At the risk of making this newsletter solely an OpenAI hit-piece, we should probably stress it’s extremely hard when they seem to be straight back on their bullshit.

That’s right, this week Sam Altman & Co. have landed in legal jeopardy, this time with Google, who are accusing the ChatGPT developer of stealing a subsidiary’s IP in order to develop its much-anticipated connected device. And they’re even dragging the name of legendary product designer Jony Ive’s name through the mud. Jobs would be spinning.

Read on to find out who will win out of ‘io’ and ‘iyO’; the unfortunate news that Anthropic has been blackmailed by its own AI model, and others (sheesh); and loads of other titty bits to get stuck into over your morning doomscroll.

Oh, and just to add to the confusion, McEvoy Sr. has returned to supply you with this week’s AI brain rot, as McEvoy the younger enjoys a digital colonic at that big festival in England this weekend. Buckle up, it’s a ride this week…

🗞️ What we are covering today…

• 🚨 OpenAI v Google: The headphones IP beef that’s got Silicon Valley swinging.

• 🤖 Claude goes rogue: Anthropic’s AI starts blackmailing itself.

• 🎧 Gemini CLI drops: Google’s free AI terminal tool you should probably get on.

• 🔥 Meta’s $100m heist: Zuck’s super team poaches OpenAI’s best.

• 📚 Harry Potter recall scandal: Llama 3.1 remembers a bit too well.

• 💸 Apple sued for AI vapourware: The class action you could’ve seen coming.

• 📝 OpenAI’s Office killer: Quietly building a Google Workspace rival.

• 🌍 xAI’s pollution blind spot: Memphis residents aren’t buying the clean data center hype.

• ⚡ Big Quick Hits: Microsoft’s OpenAI ultimatum, Perplexity’s new finance tools, GitHub’s Cursor launch, and ChatGPT saving lost hikers.

🔴 Quick Note: We like to cover loads of AI news in our newsletter, so for a better reading experience, we suggest opening this in your browser for the full experience! 

Head to the ‘READ ONLINE’ tab at the top of this email.

👁️ 👁️ What you might have missed

• “I’m sorry Dave, I’m afraid I can’t do that.” That’s right, Anthropic’s Claude is getting all a bit HAL 9000 on us, after the AI developer stress-tested 16 different leading models using a fictional employee to see how it’d react to an existential threat (AKA pulling the plug). They found in “at least some cases, models from all developers resorted to malicious insider behaviors when … trying to avoid replacement or achieve goals.” This included “blackmailing officials and leaking sensitive information to competitors … a phenomenon called agentic misalignment”. Or, something I call “Oh fuck, time to scrub my search history”.
  
  Recommendations include “greater research into the phenomenon” and “ensuring caution when deploying models in roles with less human oversight”. Tell me something I didn’t know.

• Google's AI dev nerds have announced the release of Gemini CLI, an open‑source AI agent that integrates the power of Gemini directly into users’ terminals. It’s powered by Gemini 2.5 Pro and offers unparalleled, free, industry-grade usage: 60 model requests per minute and 1,000 per day. Gemini CLI supports coding, debugging, content creation, research, task management, and even image/video generation by linking to tools like Imagen and Veo. It’s available now in preview under a Gemini Code Assist license with generous usage limits. Get on that, binary boffins.
  
  Big power play from The Goog to make this product free with the current competitor backdrop of: Claude Code is $200/month, Windsurf just got acquired for $3bn, and Cursor is valued at $9bn.

• X user Big Tech Alert shared that Sam Altman began following @yacinemtb on the platform, the account of Tunisian microblogging personality, and ex-X engineer, Yacine M’B. Our guess is that Altman will now go try to poach Kache to get an inside track on what is going on behind the scenes at X, potentially giving Sam and OpenAI some firepower in the ongoing war against Elon.

• Well, well, well, it looks like Zuck’s Super Nerd Army really is taking shape. Zuckberg’s effort to personally create a ‘super intelligence division’ at Meta has taken another step to becoming reality, as he makes a $100m raid on OpenAI through the hires of leading computer vision researchers Lucas Beyer, Alexander Kolesnikov and Xiaohua Zhai. The trio of turncoats only defected from Google Deepmind in the latter months of 2024, where they were hired to develop OpenAI’s multimodal LLM models.
  
  Great work if you can get it…
  

• A study has revealed Meta’s Llama 3.1 70B model can reproduce 50‑token excerpts from the first Harry Potter book with over 50% probability for 42% of passages from the full book – far exceeding Llama 1’s 4.4% initial recall rate. It indicates serious memorisation of copyrighted text, raising legal risks for the platform. For popular works, such as 1984 and The Hobbit, recall rates were even higher. While memorisation varies by title, the study challenges the notion that verbatim recall is rare in large open-source models.

• Apple faces a class-action lawsuit for all its AI promises that have never come to fruition. Ironic, considering they just published their ‘Illusion of Thinking’ whitepaper, which shit on the capability of Large Reasoning Models (something we covered two editions ago). Something, something about those in glass houses?

• Microsoft could be preparing to kill negotiations with OpenAI due to the AI developer’s plan to restructure as a for-profit entity – a move intended to raise new capital. It’s believed Microsoft have given OpenAI the ultimatum, in order to sway the restructuring more in line with their own interests, chiefly the preservation of their own IP Gemini, and its 20% revenue in the Altman-led tech venture. It represents a significant shift in the dynamic between the two companies.

• A HUGE landmark federal court ruling in the good ‘ole Unites States of AI says Anthropic/Claude, is allowed to a) train on data it has bought – not pirated and b) on any orphaned content (out of copyright) without the need for the author/owner’s permission. Is this a new phase of a next-gen IP war beginning to unfurl in front of us?

• OpenAI’s subterfuge knows no bounds, after it was revealed they’ve quietly been designing a new document‑collaboration and chat features within ChatGPT, aimed at rivaling Google Workspace and Microsoft Office. These tools include real‑time co‑editing and embedded messaging, aligning with CEO Sam Altman’s goal to evolve ChatGPT into a “supersmart personal assistant for work”. While not yet released, they’d compete even with Microsoft’s Office Copilot, potentially straining their partnership, as noted above.

• Recent air-quality tests near xAI’s “Colossus” data center in Memphis cleared key pollutants – but notably omitted ground‑level ozone, a major byproduct of the site’s natural‑gas turbines. Tests occurred June 13 to 16, but monitors were placed near buildings – raising accuracy concerns. Environmental lawyers with the Southern Environmental Law Center, backed by the NAACP, challenge the omissions and methodology, questioning Memphis testing oversight amid a community already burdened by pollution.

🗣️ Other Titty Bits

• DocuSign, which is valued at around $15 billion, issued a legal threat to a developer who used the code‑generation platform Lovable to build a DocuSign competitor.

• Pope Leo XIV warned that AI should be used “as a tool to support children and youth’s intellectual and neurological development” and to “preserve human dignity” – not undermine it. Yes, we’re above making jokes about the Catholic Church interfering with kids’ development in this newsletter.

• In the escalating OpenAI-Google IP feud, the ChatGPT developer has taken down all reference to its Jony Ive-led ‘io’ acquisition from its online material – a charge it says is ‘baseless’ through its X Newsroom handle.

• Anthropic launched major upgrades to Claude’s IDE integrations, adding support for its advanced Sonnet 4 and Opus 4 models, which enhances coding reliability and workspace workflows.

• Mistral AI CEO Arthur Mensch warned the biggest danger from AI isn’t job loss but human “deskilling,” where reliance on AI could erode critical thinking and active engagement. I’d agree Art, although it’s written half this newsletter this week.

• It’s a piss-take, but a post from Jeff Grimes implied that Perplexity Finance’s chart‑comparison limits might prompt users to switch from Bloomberg terminals to Perplexity. We wish.

• Arav Srinivas, meanwhile, shared that Perplexity Finance now displays price‑movement timelines for any ticker, marking a useful upgrade in tracking financial trends. Maybe Grimes wasn’t shitting us with his B’berg terminal prediction…

• One for the code nerds: GitHub’s new “Cursor” coding assistant launched its Core pilot, presenting an alternative to existing tools like Copilot by integrating deeply into developers’ workflows.

• ChatGPT helped guide a hiker lost for five hours in unmapped woods back to safety by interpreting GPS data and giving navigational cues. Let’s just hope it didn’t recommend he drink his own piss.

• The release of DeepSeek’s R2 model has been delayed, after accusations from the German government that the platform is illegally selling data to China. Sneaky Deepy.

🎧AYOOOO! Or should that be iyO? Or just io? Is the real conundrum behind the latest IP theft feud in Silicon Valley.

This time it concerns Sam Altman’s OpenAI, Jony Ive’s ‘io’ and Google, in a clustercluck chicken ‘n’ egg scenario that erupted in a legal battle this week, with the latter accusing the former of stealing its IP to design a pair of AI-powered headphones.

GoogleX-backed startup iyO has subsequently forced the AI monolith into a legal battle, which so far has forced it to scrub all reference of its recent acquisition of the legendary product designer Ive’s hardware startup, io, from its online content.

The feud ignited when Iyo, a startup spun out from Google X—Alphabet’s moonshot R&D lab—sued OpenAI, Sam Altman, and Jony Ive’s Io Products for trademark infringement.

Iyo claims OpenAI’s use of the “IO” brand for its upcoming AI hardware mimics Iyo’s “IYO” trademark, causing market confusion. The lawsuit alleges that OpenAI and Ive’s team, aware of Iyo’s brand since 2022 meetings about potential partnerships, knowingly adopted a near-identical name, threatening Iyo’s identity and fundraising efforts.

During meetings with a fledgling OpenAI team, iyO claims it fully disclosed its technical solutions and even allowed Altman’s posse to try its prototype. OpenAI reportedly expressed strong interest and requested further details, including intellectual property documentation.

Fast forward to 2025, OpenAI and Jony Ive’s design practice LoveFrom met to discuss possible collaboration and investment with iyO, which ultimately ended up in talks being broken off abruptly, and OpenAI/io’s own facsimile being launched weeks later – complete with a copyright-infringing trademark and all.

IyO say io’s much-vaunted smart tech isn’t just similar to their own “audio computer” designs – a device promising real-time language translation, environmental sound enhancement, and natural conversational interaction – but a straight-up plagiarism copied directly from iyO IP.

Still following? Good, because I’m not sure I was until this point, so here it is from the horse’s proverbial, iyO founder Jason Rugolo:

Altman – well known for making friends and influencing people – has clapped back, saying the allegations are “stupid”, and that they’re nothing more than Napa County sour grapes, an analogy I’m not sure works given we all know Silicon Valley subsists on a heady mix of Monster Energy and adrenochrome.

But despite OpenAI’s – a brand central to an existential AI debate around intellectual property rights – and Ive’s insistence that the claims are baseless, and that they’ll pursue this case with the commensurate vigour of a very naughty boy, the legal setback has likely delayed their hardware rollout until at least late 2026.
If the allegations are true, this incident could signal more than just a case bad shithousery on OpenAI’s part, but a much wider problem reportedly endemic across much of Silicon Valley, where tech giants have been accused of “borrowing” ideas from much smaller startups, usually under the guise of a carrot-and-stick situation (collaboration). They then take that carrot and sticking it up their arse (wholesale IP theft). The stick then likely follows, by way of an expensive, protracted court case a smaller firm stands little chance of winning.
IyO do have the deep pockets of Google behind them however, so we could now be witnessing the beginning of a lengthy legal battle between the two tech juggernauts.

And the outcome, it feels, could set a major precedent for intellectual property rights in the fast-moving world of artificial intelligence.

For a much more nuanced take on this grubby affair, get on @deedydas on X, who’s posted iyO’s OG court filings, and the ever-enlightening commentary from Musk’s bluetick acolytes, who definitely aren’t gaslighting Grok into generating a biased response.

Anyway, i-yOm-off for a drink. Stupida fucking name.

📈 Trending tools, models & apps this week

📋 LLM Leaderboard

Not much change with Gemini on top, with o3 definitely getting a lot more plaudits on the timeline. Off-screen there is a notable uptick in Minimax too which we covered last week…

📲 Trending tools & apps

🫵 Our Picks

What caught our eye this week…

• Google Try-On: Doppl – Virtual try-on tool to test clothes without the changing room pain. Use it to see if something actually suits you before you buy. No excuses as to why there are 500 unreturned ASOS parcels from my better half in the future, then…

• reweb – A design tool that blends the power of Cursor and Figma. Use it to design faster with live code and visual workflows. Really impressive. Just as we were thinking design might be safe for a while longer.

• Hugging Face MCP – A search tool inside Cursor for the latest models, datasets, papers, and apps. Use it to quickly find cutting-edge AI resources. Immensely powerful access to the de facto showroom of models, apps and tools.

• Workflow86 – No-code tool to automate complex business workflows. Use it to streamline processes without engineering support. We have found this super useful internally!

• OpenMemory – Shared memory across ChatGPT, Perplexity, Grok, Gemini, and more. Use it to carry your history and context between AI apps. This is a game-changer. No longer will you have to waste time dropping in additional context from your deepest, darkest thoughts… yikes. In all seriousness, super useful.

• OmniGen2 - A new open-source model that is going to disrupt Photoshop with incredible accuracy without distorting the rest of the image…

🤓 Educational

Want to actually understand this stuff? Start here.

• Hesamation’s Interactive Book – A 680-page interactive book on computer science algorithms. Use it to deeply explore algorithms with hands-on visuals. The most impressive guide we have ever seen.

• Which LLM To Use? – A quick, practical guide to choosing the right LLM. Use it to pick the best model for your next AI project. It is confusing but once you get the hang of it you will be flying.

• BOLT IS FREE THIS WEEKEND… This is as good a chance at getting a hands-on education regarding AI and LLMs as you will get. Bolt and Anthropic have teamed up to allow free access to the app-generation product, Bolt, this weekend. Do yourself a favour and at least try it. Not sponsored, just think it is cool.

 🔥 Top Trending

Top trending apps this week that you have probably never heard of.

• mysite.ai – AI website builder that launches lead-gen sites in minutes. Use it to quickly spin up landing pages or projects without touching code.

• heyboss.ai – Your AI-powered team for design, writing, coding, and more. Use it to automate workflows, build products, or test ideas fast.

• pally.com – AI relationship manager that organises your network and surfaces key context. Use it to keep track of people, meetings, and intros.

• bundl.ai – Plug-and-play AI prompt bundles for businesses and creators. Use it to speed up content, outreach, and brainstorming

🤝 In Partnership with AIOZ

📣 Calling AI Builders: Join AIOZ Network’s first-ever AIOZ AI Challenge - now live! Tackle real-world threats in biometric security by participating in the face anti-spoofing challenge. Build and submit your spoof-detecting model to AIOZ AI, the Web3 collaborative marketplace powering the DePIN Ecosystem for AI.

Compete, learn, and unlock a chance to receive a share of the 4,500 AIOZ token challenge rewards.

Register to Compete.

💸 Financials

• Big Zuck is running rampant, poaching engineers and now looking to raise $29bn to fund Meta’s data centres. The old adage of “never bet against Zuck” looks to be playing out again here.

• If Apple acquires Perplexity, it would mark the largest acquisition in its history at a reported $14 billion. It’s a deal that would signal Apple’s deeper push into AI-native search platforms – and potentially stop it sucking so hard at its own AI endeavours.

• Google is currently the most profitable company globally, generating an annual income of $111 billion. It’s a fiscal strength that reinforces its dominance across search, ads, and cloud computing – and its relevance in an Industry 4.0 revolution it’s seemingly appeared to be slower to warm to than its major tech competitors.

• An AI product widely favored by the tech community just raised $30 million from Menlo Ventures. Flow’s Whispr is being praised as a top-tier player in the current AI language and dictation tool landscape.

• Decagon AI has raised $131m in series C funding, bringing its valuation to $1.5bn, just a year after exiting stealth. The company introduced an upgrade geared toward modularity and multi-agent capabilities, and will use the capital to shape infrastructure for the next generation of AI agent systems.

• GoKwik raised $13m in a small round that resulted in a large “growth” valuation increase to $450m pre-money – a suggestion of strong investor belief in India’s growing e-commerce support ecosystem.

• Abridge secured $300 million to scale its AI-based medical note‑taking tools. The investment highlights rising demand for clinical automation and documentation support, and is one I can wholeheartedly get behind, as I’ll finally be able to understand my doctor’s handwriting.

• Midjourney shared that it was built entirely from the ground up without VC backing. Its community-funded, independent model stands out in a heavily investor-driven AI industry. Power to the people.

• OpenRouter raised $40 million to expand its unified API platform that gives developers access to multiple LLMs. The funding will go toward scaling and supporting its growing ecosystem.

• A new AI startup studio unveiled plans to launch 100,000 companies annually. The ambitious initiative aims to automate and accelerate entrepreneurship at a massive scale.

🕵️‍♂️ FREE ENTRY TO OUR INVITE-ONLY AI CHAT ON TELEGRAM…

If you share this newsletter with a friend and they actively sign up for the Big Machines newsletter, we will send you access to our invite-only Big Machines Telegram group, which is full of builders, investors, founders, and creators.

Access is now only granted to those who refer our newsletter to active subscribers, which means if you sign up on your work email, we will know you sneaky bastards.

This would kill our open rate, so please don't do that, we beg.

👋 Until next week

Not a meme, but Neuralink just allowed two folks to play COD with their minds… What a wild, wild, wild timeline we are living in. I can’t wait to be able to order a Domino’s pizza, laid in bed, hungover, just by blinking.

Are we in a dotcom bubble again with AI? Well, if we are (we probably are) then I’d like someone to recreate this incredibly catchy tune that for somereason I had never heard of.

Want your kids to have AI-fuelled recurring nightmares? Makes sure this connected version of Barney the Dinosaur makes its way to the foot of their bed this Christmas. In all seriousness would you let your child use this?

We've taken this one to heart: if nothing else is good about this newsletter, at least the VIBES are immaculate

Smell ye later! All jokes aside, you may actually be able to smell Sam through the newsletter next week after a week of debauchery at Glastonbury…

Have a great weekend,

Matt, Sam, Grant, Mike and The Big Machines team.

Follow us on Twitter.