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

AIOZ AI is here! - Upload models, run personalized challenges, and unlock the potential for token rewards on the DePIN Ecosystem for AI.

Explore AIOZ AI (more about our partner later…)

📰 What We’re Covering This Week…

Welcome to another edition of the Meta-OpenAI Big Machines newsletter! This week, the AI player transfer window truly begins to heat up, as Meta continues its talent raid of OpenAI, all explained in a series of shit footy analogies.

It’s a story that shows no sign of slowing down, as the two tech heavyweights engage in their game of attack and defence; Meta is looking to bolster its ‘superintelligence division', while OpenAI would like to keep the scoreline respectable. Hard to do, really, when four of your most important researchers defect in the space of a week…

Further down, we examine theories that Musk is orchestrating a complex financial restructuring xAI to seize more influence, while researchers tell us “self-improving AI” is some way off. There’s also musings on Apple’s preference to power Siri with Claude, and a belter story about the Man Who Can, Soham Parekh.

Finally, one last note, the letter has been penned by McEvoy the elder this week, with McEvoy the younger remaining on the subs bench (presumably with a hangover). Normal first team action will returns in the next. Enjoy.

🚀 A quick look at this week’s edition

• 🤝 Meta vs OpenAI: Zuck continues to drain Altman. Err…

• 💸 Musk Means Money: Is Musk refinancing xAI to acquire Tesla shares?

• 🧘‍♀️ NamAIstae: AI "isn't self-improving" according to researchers… yet

• 💞 Apple4Claude: Could Claude be used to power Siri?

• 🤖 Amazon launches DeepFleet: The tech giant gets all iRobot

• 🏛️ Power Plays: U.S. senators approve AI regulation at state level

• 🪓 Microsoft wields axe: How problematic is their workforce policy?

• 🥷 Who is Soham Parekh?: Unmasking tech's Keyser Soze

• ⚡ Big Quick Hits: AI aces an exam, Apple MLX team think about jumping, Cursor launches a mobile version, and xAI raises $10bn in debt and equity

🔴 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

• “Who is Soham Parekh, really?” I hear you ask. “Justice for Soham! We all (want to) do it!” others cry. If you haven’t heard the name, it’s time to get to know. This absolute fucking legend blew up the internet when it was revealed he'd quietly been juggling up to EIGHT full-time software engineering jobs – several of which were for high-profile tech firms and Y Combinator-backed startups alike – without anyone noticing. The Keyser Soze of software dev was rumbled when disgruntled former bosses shed light on the story, with the internet meme-machine then doing its thing. On a more serious note, it does tap issues around the concept of ‘overemployment’ – where a white collar worker has more than one career simultaneously – something becoming more prevalent in the tech world. I for one rate these antics. I mean, who else is going to backfill the workload of the 9,000 people Microsoft just let go?

• Reports indicate Microsoft is set to cut 9,000 U.S.-based roles, after applying for 14,181 H‑1B visas since October in the last fiscal year. The tech giant’s seeming over-dependence on H-1B skilled worker visas – imported talent on a visa programme – looks set to play into populist hands, with critics saying its workforce planning and labour strategy is at odds with the current administration’s ‘America First’ economic viewpoint, and also that it's kinda shitty to axe roughly 2% of its workforce after one of its best quarters ever, Microsoft made $26 billion between January and March of this year. The Windows maker says it's necessary to fill ‘niche’ roles, many of which are concerned with the development of artificial intelligence products. This, therefore, begs the real question: “When can we expect Microsoft Clippy to gain true sentience?”

• Speculation swirls that Musk’s xAI unit may be orchestrating a complex financing strategy involving ex-AI investors, to acquire additional Tesla shares. The rumour ties into xAI’s reported US$4.3 billion equity raise and proposed $5 billion debt issuance, prompting observers to suggest the proceeds could fuel an accumulation of Tesla stock, and as a result, a backdoor route to a larger controlling stake for Musk. It’s just another conspiracy at this point, but looking at Musk’s past interlocking buy-up strategies and resource overlaps, such as Tesla GPUs being used in xAI’s infrastructure, suggests there could be legs to this one.
  

• OpenAI researcher Jason Wei asserts that truly “self-improving” AI – capable of independently redesigning or upgrading itself across generations – does not exist today. Instead, current systems rely on human-led feedback loops and incremental design refinements. Wei also recommends setting bullshit-o-meters to wiffy when it comes to 'fast‑takeoff narratives', i.e. will we see it any fucking time soon. Self-modifying systems, he says, will unfold over years, or even a decade, due to their reliance on human intervention, trial‑and‑error, and the need for computer engineering to evolve.

• In this week's edition of Rotten to the Core, reports suggest it’s likely Apple will integrate Anthropic’s Claude model into Siri, instead of leveraging its own IP or open-source alternatives, such as ChatGPT or Gemini. The human tasked with making the decision is the Apple Vision Pro former boss, Mike Rockwell, who took control of Siri earlier in the year. It’s believed Rockwell set straight to work on selecting an LLM that could power the voice assistant and accelerate its capabilities. Apple’s own woeful underperformance in the AI space means Rockwell reportedly favours a third party over his own in-house team. Ouch.
  

• In a dramatic 99‑1 vote, the Senate removed a 10‑year moratorium provision that would’ve prohibited states from enacting AI regulations, and a clause initially tucked into Trump’s ‘One Big Beautiful Bill’. The repeal reflects the importance of state-level oversight to address issues like children’s safety, misinformation, and privacy while federal lawmakers continue to get their act together and put forth more uniform legislation. It’s also a decision that maintains state autonomy, ensuring local governments continue to pilot AI oversight initiatives. So there you have it – can’t spell the word bipartisan without the letters A and I.

• In news straight from an Isaac Asimov script, Amazon has activated its one-millionth warehouse robot in Japan and unveiled DeepFleet, a new generative AI foundation model that boosts fleet coordination efficiency by 10%. DeepFleet is trained via AWS SageMaker on internal logistics data and acts like an intelligent traffic controller, optimising robot routes to slash travel times and costs. While Amazon emphasises that these robots are to complement rather than replace its estimated 1.56 million-strong human workforce globally, the scale of deployment hints at looming shifts in operational labour dynamics.
  

• ByteDance quietly released results this week showing how AI platforms can ace a notoriously difficult Indian college exam. Google’s Gemini 2.5 Pro ranked first, outperforming every human candidate by a margin. Other AIs, such as Seed 1.6-Thinking and Claude Opus 4 performed on par with a human, demonstrating LLMs rapidly evolving academic capability. Sounds good for crafty students, but such models pose massive questions about how digitised admission selection processes can be administered fairly.

🗣️ Other Bits

• Rumors suggest key, discontented members of Apple’s MLX and open-source AI teams are considering leaving. The ramifications for Apple’s lagging efforts in the field are huge – as comp science academic Stavros Kassinos (@KassinosS) writes, “it would mark one of Apple’s worst strategic missteps at a critical moment”.

• Just when you thought you couldn’t escape AI’s growing influence on the way we work, Cursor, known for its AI-assisted coding features, has launched a mobile version, enabling users to edit and generate code on-the-go. Give its development tools for smartphones and tablets a spin. On the bus home from work. Loser.

• New breakthroughs have emerged in computational antibody design. The AI Startup Chai Discovery promises faster, more efficient tailored antibodies for therapeutic and diagnostic use, developed within a matter of weeks and not months. Science!

• A viral post from Tesla showcases autonomous robots constructing other robots, hinting at self-replicating manufacturing systems, with Musk hinting at “singularity we are not ready for”. Foreboding, but my uncle was put out of a job by a robot in a car factory, so I hope these manufacturing droids get one over on it. The bastard.

• X is testing AI-generated Community Notes, where AI chatbots draft fact-checking annotations that humans review before publication. The move aims to scale context-adding efforts while maintaining human oversight to prevent hallucinations. To me, it sounds like a snake eating its own tail, but it could give poor old Grok a chance to put its feet up.

• A growing number of OnlyFans creators are using LLM-powered tools to write messages, scripts, and manage interactions. It may boost productivity for creators, but it does raise ethical concerns about authentic engagements on the platform. I for one am distraught to find out it’s a horny AI I’ve been tipping for feet picks every evening.

🤝 In Partnership with AIOZ

👷 Build, Share, and Monetize with AIOZ AI

AIOZ AI is the Web3 collaborative marketplace where developers can upload models, store datasets, run personalized challenges, and unlock the potential for token rewards for contributions. Built on AIOZ Network’s DePIN ecosystem, it empowers the AI community to scale real-world solutions.

Join the first AIOZ AI Challenge before July 8th and explore a new way to publish and monetize AI.

Register Now!

🔫 POACHING SEASON IS OPEN - The AI transfer window is open. Meta has just raped and pillaged the OpenAI village.

If, like me, you’re into football and support a team that is a piece of shit, you’ll understand that the drama of any transfer window can often be underwhelming at the least and totally pass you by at the most.

Only joking, I support European Champions Tottenham Hotspur, so the entire thing is as enraging as it is dumbfounding. For the supporters of an opposing team – or even an interested neutral – the whole spectacle can be extremely amusing.

That’s how you may feel observing the latest twists and turns in the escalating talent war between Meta and Open AI – one in which bears all the hallmarks of a sporting transfer saga, to which I have contrived this janky analogy.

Zuckerberg’s Meta, for instance, are a prime Galáctico-era Real Madrid, pillaging any and all lesser clubs in order to assemble a team of generation-defining talent. The Sam Altman-‘managed’ OpenAI is Manchester United, and Madrid want a whole team of David Beckhams.

In many ways, Meta’s efforts to build out a ‘superintelligence division’ has been 24-carat theatre for anyone with an inkling of interest in the world of tech, but has also underscored the serious, high-stake nature of Silicon Valley power dynamics, and how they shape not just a fledgling industry, but the development of technology as a whole.

So let’s get into it, right at the beginning(ish).

Back in June, Meta founder Zuckerberg announced he’d personally be assembling a “superintelligence division” team, which would enable the tech firm to level-up efforts to improve its own LLM, Llama 4, and take greater strides towards the realisation of artificial general intelligence (AGI).

Zuck so far has bet that a critical mass of elite researchers can close the gap with competitors OpenAI and Google – and perhaps leapfrog them altogether.

No shit, I guess, when you intend to poach them directly from the aforementioned competition. The company’s deep pockets and Zuck’s personal oversight has given the drama a unique flavour, as rarely do you see a high-profile talent war unfold in real time through the medium of social media, or as we prefer, memes.

Meta’s recruitment drive has not been subtle. Zuckerberg is offering up to $100 million in signing bonuses (questionable), targeting not only OpenAI staff, but also bods from Google DeepMind, Perplexity AI, and hot startups like Runway.

But the company’s biggest coup so far: acquiring a 49% stake in Scale AI for $14 billion back in late May, and bringing on its founder, Alexandr Wang to lead the new Meta Superintelligence Labs. Wang’s arrival, alongside former GitHub CEO Nat Friedman, signalled the genesis of this ongoing saga, and that Meta was certainly not here to test the waters.

In mid-June, Meta placed OpenAI in its sights, before pulling the trigger on the hire of three of its top researchers: Lucas Beyer, Alexander Kolesnikov, and Xiaohua Zhai. It was a short-lived stint for the trio, who had only defected from Google DeepMind to join Altman and co. back in December 2024.

Zuckerberg’s recruitment of OpenAI staff since has been unrelenting: On June 26 it was announced influential researcher Trapit Bansal would become the next OpenAIer to switch teams, while less than two days later, star colleagues Shengjia Zhao, Shuchao Bi, Jiahui Yu, and Hongyu Ren, would join him.

These weren’t just any engineers: Zhao was a deep learning expert and GPT-4 contributor; Bi managed OpenAI’s multimodal models; Yu brought experience from Google DeepMind; and Ren led post-training for OpenAI’s mini models.

Altman has responded to the raid, by arguing Meta’s efforts are “mercenary” and that their mission will likely backfire, as you cannot buy the requisite culture needed to make a team like this tick (I dunno, tbh; Real Madrid won two LaLiga’s and a Champion’s League within the first three years of assembling the first Galácticos…).

The OpenAI CEO then took a more dismissive tone in public, saying Zuck has tried to poach the best-of-the-best, but “none of our best people have decided to take him up on that,” but did concede he has managed to take “a few excellent individuals”. In private, however, leaked memos and Slack logs show both Altman and the company’s Chief Research Officer, Mark Chen, are worried, likening the raid to “someone breaking into our home and stealing something”.

Industry insiders suggest the threat of a talent vacuum in a very specialist space could also pose a existential threat to an extremely lucrative, albeit young industry. Huge salaries might impact talent pipelines and ultimately narrow the career funnels of extremely bright people at the top of their game. For others, the allure of big-money wages could kill off any incentive to stick at rivals or even smaller tech startups – those who are driven to innovate and evolve based on purpose rather than cold, hard cash. Maybe Altman has a point?

Also, we need to ask ourselves, if a company like OpenAI is being priced out the talent market, what chance do other businesses stand?

To stem the bleeding, OpenAI has scrambled to talk down employees with Meta offers, considered company-wide compensation adjustments, and given staff a mandatory week off to regroup. Altman insists that OpenAI’s mission-driven culture and equity upside will ultimately outcompete Meta’s cash, declaring, “missionaries will outpace mercenaries”.

Zuckerberg’s shithousery is raising eyebrows across the tech world, however. Critics argue that Meta’s willingness to pay astronomical sums for talent – while its own Llama models lag behind rivals in code-writing benchmarks – reflects a desperation to catch up, not a coherent vision, while other tech commentators warn of the “extreme downsides” of building a team motivated solely by money.

Ultimately, when this talent window officially closes with the completion of Meta’s superintelligence team, it may well be remembered as more than a corporate raid on a rival, but a referendum on what drives progress in AI: mission or money; culture or compensation.

📈 Trending tools, models & apps this week

📋 LLM Leaderboard

Gemini 2.5 Pro still in the driving seat although I do have a sneaking feeling more and more users are dabbling with o3… maybe this upsets the leaderboard in the coming weeks.

📲 Trending tools & apps

🫵 Our Picks

What caught our eye this week…

• Clonify – Clone any website in seconds – Found a slick landing page? Spin up your own pixel-perfect copy in seconds. Just don’t get arrested.

• RaidingAI – Get anyone’s email instantly – Trying to reach that elusive founder? This tool pulls their email like it’s nothing. Again, try to not get arrested.

• Shopify Responses API – Shopify’s new Responses API – Build snappy product search, carts, and checkout flows without needing user logins or tokens. The future of shopping is LLMs.

🤓 Educational

Want to actually understand this stuff? Start here.

• UC Berkeley LL.M. Agents – Free AI Agent Courses – Want to actually understand agents? Berkeley’s got two free courses, one for beginners, one advanced, taught by some of the sharpest minds from DeepMind and top universities.

• MCP Toolkit – Redefining AI Automation – 95% of builders are sleeping on this. MCP lets you build agents that can run your calendar, organize files, and manage tasks—fully autonomous. Loïc breaks it all down with code and real examples.

• Grokking Deep Learning – A Free Visual Intro – One of the best free intros to deep learning out there—visually stunning, easy to follow, and perfect if you just want to get your head around the basics.

 🔥 Top Trending

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

• Shortcut – Superhuman Excel Agent – Shortcut one-shots most Excel tasks and smashes Excel World Championship Cases 10x faster than humans. If Excel is your battlefield, this is your cheat code.

• StereoDrift Arpeggiator – AI‑Powered Music Maker – Control beats and melodies using just your hands via webcam, launching drum hits with one, arpeggios with the other. It’s like conducting your own groove in mid-air. I mean, it is going to look weird if someone walks in in you doing this but it would be fucking hilarious so…

• AI Comic Factory – Turn text into comics in seconds – Have a story idea? Just type it, pick your style (American, Franco‑Belgian, Japanese), and watch it unfold into a full comic, no drawing skills needed. Don’t make it weird with Hentai.

💸 Financials

• Figma has filed for an IPO on the New York Stock Exchange (NYSE) with the SEC, under the exchange symbol ‘FIG’. Last valued at $19 billion, it marks not just a major moment for the design software company but potentially one of the biggest initial public offerings on the NYSE this year.

• xAI has raised $10 billion in a mix of debt and equity, fueling speculation about Elon Musk's broader ambitions across Tesla and X. The proceeds “will support xAI’s continued development of cutting-edge AI solutions, including one of the world’s largest data center [sic] and its flagship Grok platform,” Morgan Stanley wrote.

• Grammarly has acquired AI-powered email client Superhuman, signalling a push to build out productivity tools beyond writing assistance. Founded in 2014, Superhuman has raised $114 million in funding from backers and its last valuation was $825 million. Neither firm has commented on financials.

• “Time to get with the vibe”: Swedish ‘vibe-coding’ (a more ‘natural-language approach’ to AI coding) startup Loveable has been valued at more than $2 billion after recently raising $150 million in investment.

• Nvidia insiders reportedly sold over $1 billion in stock during a recent market surge, which has raised eyebrows amid the company’s explosive growth. The firm’s stock has rallied, shooting up by 44% in the last three months, and more than 17% this year despite concerns over curbs limiting AI chip sales overseas.

• Genesis AI launched with $105 million in seed funding. Founded last year, the startup wants to build a general-purpose model that will “enable robots to automate a wide range of repetitive tasks, from lab work to housekeeping”.

• AI procurement startup levelPATH raised $55 million in Series B funding, positioning itself as a next-gen alternative to legacy enterprise software. They would know: founders Stan Garber’s and Alex Yakubovich’s previous startup, Scout RFP, was acquired by Workday for $540 million back in 2019.

• Campfire, a lightweight AI-powered ERP for startups, secured $35 million in Series A funding, challenging accountancy software incumbents like NetSuite with an LLM-powered alternative.

🕵️‍♂️ 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

Well, this has started to turn into an AI gossip magazine, and we’re fine with that. It is hilarious to think that there is such a high demand for AI engineers whilst companies are also raising billions (collectively) for code-generation platforms.

Now, it wouldn’t be us to not finish on our favourite memes of the week that we have found. When chaos ensues, the hilarity is also in equal proportion.

Too young to buy Bitcoin at $3, but born at the perfect time to be a turncoat in the largest technological revolution since the printing press.

This is the most life-like, realistic robot we have seen yet. They have obviously been watching some of the professionals and learning how to dive as a consequence.

• I swear I've seen this bloke in one of those funky German nightclubs before… maybe it is for if you need some assistance.

Next time you get one of those annoying recruiter messages on LinkedIn, here’s the perfect response. They may even be stupid enough to offer you a Meta-level salary.

And finally… this week in a nutshell.

This is a wild old time, and we will be here each week to cover it. So, do us a favour and share it with anyone you think would find it useful!

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

Follow us on Twitter.