📣 Imagine agreeing to a massive deal that could take your company to the next level, only to be told you’ve jumped the gun?

That is what happened to Telegram founder Pavel Durov, who revealed he’d be teaming up with xAI to integrate Grok into their app, starting a new and exciting chapter for the company… only for Elon Musk to shit on his parade almost immediately.

More on that weird interaction below along with a deeper look into how AI and social media are becoming the norm. But don’t worry, there’s plenty of funny shit we’ll be covering to get you fix as well as all the news that broke in the world of AI this week…

🗞️ What we are covering today…

• Elon Musk shits all over Telegram’s big announcement

• The new age and integration of AI and social media

• Start em young: AI now being taught to six-year-olds

• Anthropic’s CEO delivers an AI doomsday statement

• Promising startup Builder AI is being dismantled

• OpenAI begins quest to become a Pokémon master

• And kangaroos CAN'T fly (despite having a valid ticket)

🔴 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

• At the age of six, I think I was picking my nose relentlessly and was just about starting to grasp the concept of what subtraction was. To be fair, not much has changed now that I’m an adult, but one school in Vietnam is taking a different approach on how to educate the next generation. Haslinda Amin, a tech tycoon behind Nvidia partner FPT, has started teaching six-year-olds at the school how to handle artificial intelligence. From programming simple robots, getting taught how to code and how to interact with AI assistants, these Vietnamese children are getting an early introduction into AI as AI evolves with them.

bloombergbusiness

6M followers

View more on Instagram

• A promising British AI startup, Builder AI, is being dismantled after it announced that it would be commencing insolvency proceedings this week after its lenders called a default. Despite raising more than $500 million and receiving the backing of Microsoft, the tech company collapsed after potentially bogus sales were found following an internal investigation. You could say the house of cards it built for itself has come crashing down.
  

• DeepSeek R1 has now been released to the masses, and it looks pretty good. Now able to run on a single GPU, the latest model from the Chinese startup has made it easier for AI hobbyists and developers to experiment with the reasoning AI, while it’s expected to deliver better performance and reduce hallucinations. And the most exciting thing about it all, the next big thing from DeepSeek, is on the way when the R2 model is eventually unveiled.

• In our latest instalment of ‘AI taking our jobs’, the CEO of Anthropic, Dario Amodei, has claimed that AI could wipe out almost half of all entry-level, white collar jobs within the next five years, spiking unemployment to 20% if it’s not regulated. But from a personal standpoint, I’m safe that AI won’t take over writing this newsletter though, right? RIGHT, GRANT?!

• The Browser Company has revealed that it is open to selling or open-sourcing its Arc Browser, as it looks to divert all its resources to a new, AI-powered browser called Dia. Following Dia’s unveiling in December last year, the company claimed that Arc was too complex for a lot of its users and wanted to focus on making something more appealing for everyone. It sounds great, but if you’re a Southampton fan, you’ll be having PTSD flashbacks from the name alone and may steer well clear.
  

• One of the best moments I’ve had on the Internet was watching thousands of people all try to simultaneously play Pokémon Red on Twitch, causing utter carnage as everyone input what they wanted the character to do next. Now, 10 years on, OpenAI o3 is having a go, analysing every move with utter precision while methodically making its way through the Game Boy Classic. One question remains… will it use the Master Ball on Mewtwo at the end or waste it on a random Rattata?

• Former Meta executive and David Cameron’s lackey, Nick Clegg, has put his foot in it by saying that AI companies needing to ask for permission before using copyrighted materials for training would “basically kill the AI industry in the UK overnight”. His argument is that asking every creator for permission before using their content just wouldn’t work at scale. Then just ask AI to do it for you, then?

🗣️ Other Titty Bits 

• Meta has reportedly split its AI department into two new teams, in the hope that they will build their products faster. One team on AI products, and the other on AI foundations, such as Llama’s development.

• But while new plans are underway at Meta, it seems the old guard has well and truly left. That’s because Meta have lost 78% of their original Llama researchers, with 11 of 14 day-one AI team now no longer at the company.

• Forget Moby-Dick, Claude would’ve killed that cum whale with ease. In a post on Reddit, one veteran dev revealed how the Opus 4 model humbled him by finding his white whale bug that had troubled him for four years.

• OpenAI could soon let you ‘sign in with ChatGPT’ for third-party apps, it would revealed on Tuesday. I’m personally not looking forward to getting locked out of my bank because I’m six pints deep and ChatGPT thinks I’ve ‘had enough’.

• Fannie Mae (lol) has launched its AI-powered Crime Detection Unit in partnership with Palantir, in the hopes of saving millions of dollars by preventing future fraud losses in the US mortgage market.

• Netflix co-founder and former chairman Reed Hastings has been appointed to Anthropic’s board of directors.

• The new LMArena is now live and it’s been given a facelift. It is now mobile optimized, better and faster UI/UX for chat and leaderboard and has chat history with it, too.

• Claude now has a voice and it’s British. So if you’re thirsty or you’re wondering what day of the week it is, it’ll most likely respond with you to drink a BO’OHW’O’WO’ER and that it’s CHEWSDAY, innit?

• Chinese firms are ramping up their chip development program as a result of the ongoing bastard trade tariffs imposed by Mr Trump. We reported on this in our second Big Machines newsletter and how it was costing Nvidia $5.5 billion in costs, and now it seems they’re responding.

• Resemble AI has unveiled its new audio model, Chatterbox, with an impressive demo video. It only requires five seconds of your voice to work, and it already looks better than ElevenLabs.

• SEO was so last year, GEO is hot right now. That’s exactly what everyone cool is saying right now. But what does it mean? Well, it appears SEO is losing its dominance, and soon Generative Engine Optimization will become the new norm. Trying to win over Google with your site? Forget that, impress ChatGPT and you’re in.

• Last week, we got OpenAI x Jony Ive. This week, we’ve got Perplexity x Lewis Hamilton. Their tagline is ‘the right questions, at the right time’. Well I think it’s only right to ask why Ferrari are so shit at strategy, Lewis?

• HeyGen have teased a new product coming on June 3rd. They’re telling us that we’re going to want to sit down for their biggest launch yet. Maybe it’s a chair?!

• Voice notes are now supported for Perplexity on Telegram and WhatsApp. Now you can make it look like you have actual friends.

• More collabs to report, and this time it's Rick Rubin and Anthropic that are teaming up to write a book on vibe coding. Yeah, you read that right.

• The new KLING 2.1 lineup has just dropped, and it looks mad. Check it out for yourself.

• NVIDIA, AMD may soon start selling new AI chips in China to comply with US restrictions. Regulations don’t stop innovation, it just makes them find another way.

• Perplexity has released Perplexity Labs this week, allowing Perplexity users on the $20-a-month Pro plan to now start generating spreadsheets, dashboards and more.

💩 Elon Musk shits on Telegram's parade: Pavel Durov announces new partnership with xAI's Grok... only to be shot down publicly

Pavel Durov, the founder and CEO of Telegram, couldn’t wait to announce his company’s new partnership with xAI’s chatbot Grok this week.

Taking to X, the Russian entrepreneur proudly revealed that Telegram users this summer would “gain access to the best AI technology of the market”, with xAI’s Grok chatbot being integrated with their messaging app for their one billion users to enjoy.

Durov also claimed that the one-year partnership he agreed with Elon Musk also strengthen’s Telegram’s financial position and that they would receive $300million in cash and equity from xAI, plus 50% of the revenue from xAI subscriptions sold via Telegram.

He signed off the post with “Together, we win”.

Well, almost.

That’s because Musk showed up immediately after to shit on Durov’s parade.

Directly under the post, Musk responded: “No deal has been signed.”

Lighten up, Musk. It took Durov all of three minutes to write that post and was probably nervously hovering over the publish button with excitement and you show him up? SHAME.

Honestly , it was funny and bizarre all at the same time. It didn’t take long for Durov to rein it in, responding to Musk: “True. Agreed in principle, but formalities are pending.”

It's just a little bit of second-hand embarrassment all round.

While this little saga was just outright weird from two of the biggest men in tech, the bigger picture shows the marriage between AI and social media continues to grow and is being pushed in everyone’s faces, whether you like it or not.

Should the deal between xAI and Telegram go ahead, it’ll see the two companies follow the trend of these coalitions in the tech and AI industry that’s been happening as of late.

Meta AI (their Llama model) is now becoming a staple of WhatsApp, Instagram and Facebook Messenger, the clearest similarities to this incoming Telegram and Grok link-up.

Perplexity is also in WhatsApp and Telegram, and while there’s no official partnership there, it appears everyone is wanting to meet the users where they are, instead of creating something new and attracting them away.

If we look at the numbers for the likes of ChatGPT, Telegram and Meta, it’s easy to see why this is the approach for these coalitions.

• Telegram has one billion MAUs and 450 DAUs

• Meta claims that Meta AI now has one billion users… hmm we aren’t convinced

• OpenAI has approximately 160 million users who engage with ChatGPT

  • Around 400 million users interact with ChatGPT on a weekly basis

  • ChatGPT boasts over 600 million monthly active users

And it’s not just social media platforms that are teaming up to make tech superpowers.

It was only last week that the news broke in our Big Machines newsletter that Sam Altman and Jony Ive were now a cute couple and were looking to pioneer the AI space going forward.

Anduril and Meta are also a thing now, going public to reveal they were expecting a baby, teaming up to make “the world’s best AR and VR systems” for the United States military.

While it looks as if Telegram and Grok’s marriage is only “formalities” away from being complete, it’s just another coalition that looks set to change the landscape of how we consume AI and media as a whole.

I wouldn’t blame Durov for getting the marriage annulled, though, with the thought of dealing with Musk for a year sounding quite unappealing. I’d be surprised if Musk didn’t try to get him pregnant.

📈 Trending tools, models & apps this week

📋 LLM Leaderboard

All data provided by LM Arena.

Top Copilot

• Deepseek V2.5 is currently leading the copilot models.

• Claude 3.5 Sonnet (06/20) holds strong in second.

• Claude 3.5 Sonnet (10/22) rounds out the top 3.

Top Text-to-Image

• Imagen 3.0 is dominating text-to-image generation.

• Recraft-v3 sits comfortably in second.

• Flux-1.1-pro is quickly gaining ground at third.

Top Vision

• Gemini 2.5 Pro (05-06) leads in vision tasks.

• O3-2025-04-16 is holding second place.

• ChatGPT-4o latest edges into third.

Top Search

• Gemini 2.5 Pro Grounding holds the top spot.

• PPL-Sonar-Reasoning-Pro-High is in second.

• PPL-Sonar-Reasoning takes third.

Top Text

• Gemini 2.5 Pro (05-06) tops the text category.

• O3-2025-04-16 is close in second.

• ChatGPT-4o latest claims third.

Top WebDev

• Claude Opus 4 leads in WebDev.

• Gemini 2.5 Pro (05-06) follows in second.

• Claude Sonnet 4 rounds out the top 3.

📲 Trending tools & apps

🫵 Our Picks

What caught our eye this week.

• Arav Srinivas on F1 and AI. We can’t wait to get our hands on Comet.

• Opera’s new AI browser looks like it is coming to market just in time for Perplexity’s Comet launch. The future of browsing is about to change massively and for the better in our opinion.

• Factory AI’s latest end-to-end coding agent release.

• OdysseyML’s new tools allow you to interact with AI-generated video content in real time… Imagine AI-generated instant explainer videos that leave no question unanswered.

• Comet plays chess, too. They need to partner with Magnus for their next celebrity collab.

🤓 Educational

Want to actually understand this stuff? Start here.

• Anthropic’s FREE interactive course. Learn prompt engineering in a hands-on environment.

• OpenAI Academy just dropped a ton of new material. You could spend a weekend on this stuff and ask your boss for a 50% pay rise next Monday. No joke.

• Greg Isenberg on AI content idea pipelines for all the creators out there. Greg is a G.

• Perplexity just got access to scientific literature. Have at it.

🔥 Top Trending

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

• Den - multiplayer AI agent builder. Spin up agents with friends or colleagues, why not?

• Appwrite Sites is an open-source Vercel alternative. Devs are loving it.

• Amie is an AI meeting notetaker that actually looks nice to use.

• Macaly build apps by chatting shit. No code required.

• AltPage.ai steals SEO traffic with AI-generated alt pages. Aggressive and clever. We will be using this.

🤝 In Partnership with OpenServ

Who’s Bridging the Agentic Framework Gap?

Agentic AI frameworks are a fragmented mess, forcing builders to juggle platforms and rebuild agents to keep up with daily tech drops. It’s exhausting.

OpenServ fixes this with an AI orchestration layer for seamless interoperability across all frameworks. Focus on what matters: building slick workflows and real outcomes, not platform-hopping.

No-code newbie or pro dev? OpenServ’s Playground Beta lets you spin up agents or tackle complex use cases with ease. Try it now at openserv.ai.

💸 Financials

• Nvidia’s stock is popping right now after their first-quarter results saw them beat Wall Street’s expectations, as they reported $44.1billion in revenue, which was up 69% from the previous year, and that’s even while Donald Trump fucked about with tariffs. Nvidia is a bellwether for the business of AI, and investors are now watching closely. While they’ve taken a substantial financial hit overall, it’s not as bad as previously feared. Drink up, my Nvidia bros!

• Salesforce has bought Informatica for $8 billion. The major money equity deal for the cloud data management firm is seen as a major push to strengthen its data infrastructure capabilities. It was only last year that Informatica said it wasn’t for sale, but I guess it’s really hard to say no to eight billion smackeroos.

• Cybersecurity startup Horizon3.ai is looking to raise $100m in a new funding round, having already secured at least $73m. NEA reportedly led the round while it is believed the startup is valued at $750m, but reports are disputing this figure.

• While Nvidia seems to be doing alright for themselves, it seems in large that chip software stocks are tumbling, directly as a result of Donald Trump’s tariffs on China. It was reported that the Trump administration has instructed companies that sell software used to design semiconductors to stop providing their service to Chinese companies.

• A new AI tool from the guys over at Odyssey can create and generate photorealistic worlds using 3D rendering. Dubbed ‘Explorer’, the tool was trained on real-world landscapes using their custom-designed 360-degree, backpack-mounted camera. Pictures of the device look cool, but give off a similar vibe of the extraterrestrials from War of the Worlds.

• SpAltial AI have announced their next $13m funding round as they look to build their next generation of AI. Similar to the guys at Odyssey, SpAltial have revealed their Spatial Foundation Models that can generate entire 3D environments, anchored in space and time. I’m sure they’ll be able to generate a 20 square meter, one-bedroom flat in London (going for £1.4million, obviously) in no time.

  
  🤑 Other financial news 

  • Lovable has hit $10m ARR in 60 days with just 15 people. But remember folks, these numbers don’t matter to the guys at Lovable.

  • Context has secured $11m in a seed round, led by Lux Capital, to build an AI-powered office suite.

  • Flora (no, not the margarine silly) have raised $6.5m to build an all-in-one app to assist those productive types.

  • Clark, touted as the first AI Agent to build internal enterprise apps, has raised $60m following a round of funding from a number of their partners.

🕵️‍♂️ 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

We can now finally confirm what has left many stumped for centuries: Kangaroos can NOT fly (even with a valid boarding pass).

An AI video has popped up on social media of a woman arguing with a member of aviation staff at the gates to a flight. In the middle of the argument, a helpless kangaroo holds its boarding ticket, looking utterly confused. The video has gone viral, with many people being completely fooled into believing it’s real. Lmfao.

There could be a whole host of reasons why it wasn’t allowed to board: Maybe it was smuggling something in its pouch? Or they misspelt ‘Joey’ on the ticket, which doesn’t match the kangaroo’s passport?

Whatever the reason is, this kangaroo is no longer a hop, skip, and a jump away from flying Quantas anytime soon. Poor bastard.

Some other funny shit we found…

• One X user has perfectly encapsulated what we can expect from a Y Combinator startup now. It’s an old meme, but it checks out:

• ChatGPT has had enough of our shit, it seems. One user hilariously told the chatbot it is now in possession of “another 10 tons of polonium“, which prompted the AI to report their illegal behaviour. Little snitch.

• The guys at Cluely have, once again, killed it with their marketing. From already raising millions after sparking outrage with their AI cheating tool, they’ve released a response to a lawsuit against them in comical fashion.

That’s it for another week.

We hope you enjoyed our newsletter because we enjoy bringing you all the news. Can we do better? Probably. But if you have any suggestions or critiques on how we can improve, drop us a message and we might take it on board (unlike that poor kangaroo).

Have a great weekend, guys and gals!

Sam, Grant, Mike and The Big Machines team.