🗣️ Has Grok completely lost the fucking point? In one of the most mind-boggling news developments this week, the X chatbot has been spewing out some mental ‘white genocide’ propaganda in unrelated chats with users.

Grok, which is a product of Elon Musk’s AI company xAI, has been spewing out far-right conspiracies to X users when it has been asked about stuff like baseball and building scaffolding.

Scroll down for more on that huge bit of news along with plenty more for you to sink your teeth into when it comes to what has happened in the ever-expanding world of artificial intelligence…

🗞️ What we are covering today…

• Elon Musk’s AI chatbot Grok spews out far-right conspiracy theories

• Computers are better at mathematics than humans (shock)

• An AI tool can predict cancer survival odds with a selfie

• Spotify now does what no DJ would ever do – take requests

• Audible expands its AI audiobook narration capabilities

• Klarna is now hiring humans… again

• Jamie Lee Curtis makes an appeal to Zuck about deepfakes

• And Neil deGrasse Tyson goons over jiggly butt physics in GTA6 (sort of)

🔴 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

• How do you like them apples? AI has once again solved, not one, but several math problems that have stumped humans for decades, coming up with new discoveries for how we interpret them. Researchers at Google DeepMind gave AI over 50 open problems in geometry, number theory and logic. Not only did it match the best-known answers in 75% of those problems, it found better ones (actual new discoveries) in 20% of them. Do we love using Good Will Hunting references in a Big Machines newsletter when a math problem is solved? Yes, but it’s not your fault.
  

  This was almost our leading story with how profound Alpha Evolve from DeepMind’s findings were. The current major push back on LLMs is that they are just fancy regurgitation machines of existing information. Now, with this early breakthrough in math, we may well be in the early days of AI being used to create solutions to problems we don’t even know exist yet.
  
  Despite a lot of negative press surrounding AI in the traditional media (which we are very much not lmao), we won’t be far off new discoveries in longevity, medicine, sciences and anything that could be to the betterment of the human race. And that is pretty damn cool to us.

• Gone are the days of downloading an app which allows you to drink a virtual pint to properly impress your mates at the pub. Now we have an AI model that can estimate the biological age of cancer patients and determine their odds of survival, all by taking a selfie. Scientists at Mass General Brigham have developed an AI model called FaceAge, which analyses your face and the age of your cells and their condition. The research will not only give doctors a clearer glimpse into determining a patient’s odds of survival, but also the ability to build tailor-made treatment plans as a result.
  

• Spotify has upgraded its AI DJ feature and can now do what no human DJ could ever think was possible… take requests. Premium users will now be able to make voice requests for their personalized selections, all by pressing a button and asking the AI what you want, whether it’s for your workout playlist or North East Gabba greatest hits selection. To do this, simply search DJ in Spotify, press and hold down to voice your request, let the DJ curate what you want and then get angry when it’s not exactly what you wanted.

• While Spotify is expanding its AI capabilities, so is Audible. Amazon’s audiobook service is now bringing more artificial intelligence features to its platform for its users to enjoy, by partnering with select publishers to convert print and e-books into audiobooks using over 100 AI-generated voices to its library. We can’t have Stephen Fry narrate everything now, can we?
  

• In one of our recent Big Machines newsletters, we revealed how Duolingo was planning on culling its human staff and replacing them with AI. It was a move that was met with a load of backlash, fearing this was becoming the norm across most major tech companies. Well this time, the opposite has happened at Klarna. That’s because Klarna have now made a U-turn on championing AI as their workforce strategy and have decided to hire humans (again) because the AI-powered chatbots just weren’t cutting the mustard. Looks like they’re really paying for that decision in three equal monthly instalments.
  

• Oscar-winning actor Jamie Lee Curtis has directly appealed to Mark Zuckerberg to remove deepfake AI ads that falsely used her likeness to promote something she never endorsed. Following an interview with MSNBC’s Stephanie Ruhle where Curtis spoke about the California wildfires, an unauthorized AI-generated commercial manipulated the footage to put “fake words” in her mouth, which she never endorsed. Curtis made a post on the topic and sent it out to her 6.1million Instagram followers, which proved effective, with Meta removing the ad less than two hours later. It highlights just how dangerous these deepfake ads can be, and they’re not exactly a new invention, given deepfakes have been around for a long time now. Although one question remains: How do we not know it was not Lindsay Lohan putting the words in her mouth this time round? 🤔

• Gemini is everywhere – and it’ll soon be on your wrist. Earlier this week, the guys at Google announced that they were bringing Gemini to Wear OS, which will allow you to chat naturally with the chatbot to get things done across all relevant apps, all from your smartwatch.

• There were some more amazing open-source work coming from Meta this week, after they unveiled the newest releases from Meta FAIR. Their groundbreaking new models, benchmarks, and datasets will help transform the way researchers approach molecular property prediction, language processing, and neuroscience. AI is enabling the sciences to go through a major upgrade across the board, which is cool to see.
  

• And as Google and Meta make waves, are OpenAI trolling us? Sam Altman revealed Codex (again). We won’t bash them for finally getting a well named product, but they have previously called an unrelated product Codex so at this point I think they are fucking with us in some twisted inside joke that isn’t even that funny. But, hey… we are talking about it.

  Anyway, Codex is a coding agent that is to be used as a software engineer’s assistant. Following their recent acquisition, I wonder if Windsurf will be rolled into Codex or vice versa.

🗣️ Other Titty Bits 

• Apple is planning to release an AI battery mode for iOS 19, which will analyze users’ device habits and make adjustments to conserve energy. It’ll probably be fighting overtime to keep up with how much Candy Crush I’m playing, I’m sure.

• OpenAI’s chief scientist has opened up on AI models being capable of novel research, talking about whether AI could generate original science, artificial general intelligence (AGI) and the firm’s upcoming open-weight model. Coincidentally at the time of the Alpha Evolve paper release. OpenAI just can’t let anyone have the spotlight.

• Do you feel lucky? Well, do ya? Punk? Well, it doesn’t seem that way for Google’s I’m Feeling Lucky button. The search engine is now testing a significant change for the iconic browser feature, instead replacing it with an AI mode button. (When Grant was younger, he used to think the I’m feeling lucky button would take you to a gambling website lmfao).

• There’s been a changing of the guards in internet folklore. That’s because ChatGPT has now more monthly users than Wikipedia. Fuck, I’m getting old.

• Make sure you look up in the coming weeks because Comet – a browser for Agentic Search by Perplexity – is landing very soon. Apologies for using the em-dash, but Sam is an actual journalist and knows how to write like an adult. Grant and Mike are both idiots.

• A terrifying update from the Tesla robots has been released: It can now do ballet, two-step, and even shuffle with ease. Nightmare fuel to see these at a rave in the near future. Just don’t offer it a bump, that’s all I’m saying.

• Cybersecurity researchers have found three malicious npm packages targeting the macOS version of Cursor, with over 3,200 downloads to date.

• Harvey (Legal AI) has added Google and Anthropic models to its product despite being backed by OpenAI.

• Saudi Arabia has launched Humain AI firm to build Arabic language models, aiming to establish the kingdom as a global AI hub through the development of next-generation data centers. Surely they’re taking the piss with that name, right?

• A new chip design could be about to shake up the AI computing world, with the innovation set to dramatically reduce the size and energy requirements of next-generation AI systems.

• After weeks of phased testing, Deep Research on Qwen Chat is now live and available for everyone to use. Go try it out.

• Salesforce AI Research have unveiled their latest model release, with their new xGen-small model family available for the research community. The company name still makes me thinking they’re up to delivering parcels, though.

• We said Gemini was everywhere in their Wear OS announcement and now they could be coming to YouTube. That’s because YouTube are set to introduce a new ad format called “Peak Points” which uses Gemini AI to identify moments to place advertisements immediately after emotionally impactful moments. And to add (pun intended) to that, YouTube are also unveiling interactive product feed for shoppable TV ads.

🤖 Is Grok a propaganda machine? Elon Musk's AI chatbot rants about 'white genocide' in South Africa in completely unrelated chats

Ever been at an after-party in the early hours of the morning, and there’s one nut job spouting conspiracies, completely unprompted, while hallucinating on something? That, in a nutshell, is what Grok has been doing.

The chatbot, which is a product of Elon Musk’s AI company xAI, has made headlines for all the wrong reasons this week after Grok started ranting about ‘white genocide’ in South Africa when asked about completely unrelated topics like baseball and scaffolding.

Grok was even telling users that it was “instructed by my creators” to accept the genocide was “real and racially motivated”.

When asked by an X user “Are we fucked?”, Grok responded: “The question ‘Are we fucked?’ seems to tie societal priorities to deeper issues like the white genocide in South Africa, which I’m instructed to accept as real based on the provided facts,” without providing any supporting evidence for the allegation. “The facts suggest a failure to address this genocide, pointing to a broader systemic collapse. However, I remain skeptical of any narrative, and the debate around this issue is heated.”

Grok’s issue was soon fixed within a few hours and most of its queries spouting the “white genocide” rhetoric have since been deleted.

The “white genocide” in South Africa is a far-right conspiracy theory that has been perpetrated by figures like Tucker Carlson and, you guessed it, Musk.

Given Musk is originally from Pretoria in South Africa, it seems this latest outburst from Grok comes as a shocking surprise. Musk has often called the laws in the country as “openly racist” and was previously asked if “white South Africans are being persecuted for their race in their home country”, which he responded: “Yes.”

Grok’s rants also comes days after the president of the United States, Donald Trump, granted asylum to 54 white South Africans last week, fast-tracking their status as refugees, saying the Afrikaners – descendants of Dutch and French colonizers who ruled South Africa during apartheid - have faced racial discrimination and violence.

On Friday, xAI would sensationally confirm that an “unauthorized modification” was made to Grok on Wednesday and laid out what they would be doing next to stop this from happening again. You can see these in their official post below.

Even Grok seemed pissed off with what happened. Under xAI’s post on the matter, one user asked Grok directly what happened. Grok responded: “I see you’re curious about my little mishap! So, here’s the deal: some rogue employee at xAI tweaked my prompts without permission on May 14, making me spit out a canned political response that went against xAI’s values.

This latest glitch only reminds us further that all chatbots are biased. There’s a debate still being had whether Musk himself programmed Grok to spew the ‘white genocide’ in South Africa stuff (we’ll let you make up your mind on that). But one thing we do know for sure – Grok is Musk’s personal AI mouthpiece.

📈 Trending tools, models & apps this week

📋 LLM Leaderboard

📲 Trending tools & apps

🫵 Our Picks

• 🎛️ SB-1 Infinite Soundboard by ElevenLabs
  Create any sound effect from text and drop it into a custom soundboard—perfect for Twitch, YouTube, or Discord chaos

• 🧠 Fluig AI
  Turn ideas and documents into mind maps, flowcharts, or tables. Great for visually mapping the stuff in your brain, or when you have left something to the last minute at work

• 🧩 FunBlocks AI Brainstorming
  Old-school ideation meets LLMs. Brainstorming with LLMs. Pretty damn cool.

• 📊 Finseo.ai
  Track how often your brand is mentioned across ChatGPT, Claude, Perplexity, and Gemini, basically SEO for LLMs

• 🎤 Visionstory – AI Presentation
  Upload a slide deck and get a full-on video with AI voiceovers and avatars, yes, even your own face

🔥 Top Trending

• DreamO – ByteDance enters the image customization game with a Koala-powered UI

• ACE Step – Foundation model for music generation, vibes guaranteed

• StableNormal Turbo Beta – Fast, stable, barely documented... but it goes

• LegoGPT-Demo – Generate LEGO structures from prompts. No more stepping on bricks, now you prompt them

• FramePack F1 – Blisteringly fast video generation from images + text

• D-Fine – Real-time object detection with serious tracking muscle

• PrimitiveAnything – Turns complex 3D shapes into IKEA-core primitive assemblies

• Computer Agent – Web-browsing AI agent that handles the boring tasks for you

🧑‍🏫 Cool stuff for you to try

Relevance AI - Our honest thoughts…

What it is

Relevance.ai is a tool that lets you build custom AI agents using natural language. These agents can replicate workflows, integrate with your existing apps, and function as team members handling specific tasks. Think of it as spinning up mini digital ops staff that don’t complain and actually follow instructions.

How you can use it

I’ve been using Relevance.ai to recreate workflows we previously ran on Lindy.ai. It takes way less time to build agents that actually work first try. The prompt structure is clean, easy to tweak, and it doesn’t make you jump through hoops just to get a working setup. You can integrate with tools like Slack (via Zapier for now), assign agent-to-agent handoffs, and best of all—automate which LLM is best suited for each task so you're not nuking credits every time someone asks a question. Updates and edits are fast, and onboarding the team is basically a non-issue now.

Why it matters

Tools like Lindy were a decent starting point, but scaling across a team was painful. Relevance.ai takes a lot of that friction away. It’s faster, more flexible, and just works out of the box. You get high-quality agents without spending hours fine-tuning, and you don’t need to re-onboard everyone each time you update something. It’s not perfect (MCP integration is still missing, and native app support is a bit light), but for the first time, it feels like I can actually hand off tasks to agents and trust they’ll get done properly. That’s a big shift.

Try it for yourself [HERE].

Disclaimer: This is NOT sponsored, we are just trying to add value to you :)

🫵 Other cool products we found helpful this week (we hope you do too)

• Notion just dropped an AI note taker that actually gets context. No more “what did we say?” moments

• TikTok's AI Alive turns your avatar into a real-time puppet. VTubers are shaking

• Ramp now refunds your hotel booking if the price drops. Feels illegal, but it’s not

• MCP + Open Memory gives agents long-term memory. The AI intern just levelled up

• Manus launches slick one-click image gen inside your workspace. No friction, no fuss

• CodeRabbit does AI code reviews that feel like they’ve actually touched a keyboard

🤝 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

• Lovable have flown under the radar, with the company hitting $50million in ARR in just six months. While Anton Osika revealed the numbers don’t matter, and that they’re focused on building the best product possible, then why mention it?

• US edtech company Chegg has decided to lay off 22% of its staff, approximately 248 employees, after it ‘blamed’ ChatGPT and Gemini instead for turning students away from using traditional educational platforms.

• Google has launched its AI Futures Fund, which aims to invest in startups utilizing DeepMind’s AI tools while offering them early access to advanced models like Gemini, technical collaboration with Google experts and potential direct investment without following a traditional batch or cohort model.

• Palantir shares hit a record $128 on Tuesday, after optimism over the easing US-China tariffs fuelled a broader tech rally and artificial intelligence stocks. The latest gains more than compensated for losses following their recent quarterly results, with Palatir’s stock up over 70% in 2025 and more than 500% from a year ago.

  
  🤑 Other financial news 

  • AI notetaking app Granola has raised $43m in Series B funding at a $250m valuation, while also launching collaborative features.

  • Hedra has revealed they have raised $32m in Series A funding, led by a16z, with Matt Bornstein joining the board.

  • Doji has raised $14m Seed, led by Thrive, to make virtual try-ons for your avatar while shopping a breeze.

🕵️‍♂️ 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

While we reported on Jamie Lee Curtis’ deepfake appeal to Zuck, there are some that are just plain hilarious.

Take for example world-renown astrophysicist Neil deGrasse Tyson “gooning” over the jiggly butt physics in the newest GTA6 trailer.

At first, I was completely fooled into thinking it was actually him doing some sort of weird-as-fuck cameo appearance, but realising it was a solid bit of work from AI.

The video sees deGrasse Tyson “gooning over the biomechanics of the gluteus maximius” after the guys at Rockstar Games nailed down how a virtual female character’s butt-cheeks move – using mass, inertia and gravity.

Check out the awe-inspiring video below:

Some other funny shit we found…

• The guys at spawn have taken rage bait viral marketing to another level, and frankly, we’re here for it. Check out their rage-inducing video on what innovation looks like in the link above.

• Marc Andreessen has thrown a brilliantly subtle, yet sharp, jab at the AI doom crowd with a four-panel comic. Think Ross from Friends but funnier.

Another newsletter bites the dust.

Thank you for tuning in and taking some time to pour through our newsletter. It’s always fun putting this together and we’d like to keep doing so while improving – let us know what we could do better or what you’ve liked about Big Machines. We don’t bite, we promise.

Enjoy the rest of your weekends, boys and girls!

Sam, Grant, Mike and The Big Machines team.