Vol 15 ★ AI Finally Solving Life’s Greatest Equations, James Cameron Shops for AI & 9 Tool Hacks to See You Through to Winter

Hello, my loves. Something rather delicious is happening in the AI playground, and I'm not sure everyone's reading the room, or even in the same room anymore.

On one side, investors are making it rain on generative AI creation apps like it's 2021 and someone's just minted a monkey. And we're not talking pocket change. Just a little snapshot of this year's shopping spree:

✦ Lovable raised $400m at a $13.3bn valuation ✦ ElevenLabs raised $500m at $11bn ✦ Suno raised $400m+ at $5.4bn ✦ Higgsfield raised $400m at $5.4bn, quadrupling its value in eight months ✦ Runway raised $315m at $5.3bn ✦ Synthesia raised $200m at $4bn

The message from Silicon Valley? AI CONTENT WINS.

Except... someone forgot to check with the people running the platforms where all this content is supposed to go. Because over on the other side of the internet, the velvet ropes are going up.

✦ Spotify is labelling fake AI artists and keeping them out of recommendations ✦ YouTube is restricting monetisation for mass-produced AI slop and certain AI-generated expert advice ✦ Instagram is turning down the visibility of AI influencers who won't identify themselves ✦ TikTok has introduced a slider to dial down AI content ✦ Bandcamp has banned AI-generated music.

So we're spending billions making content the platforms increasingly don't want to show us? Make it make sense, doll. But before we all start screaming AI bubble! and flogging our subscriptions, there's something rather important buried in the small print.

This isn't simply a war on AI. Platforms are drawing different lines around fake identities, deceptive content, mass-produced shlop and, in some cases, fully generated work.

AI isn't necessarily being kicked out. But the bouncers are suddenly getting VERY particular about the guest list. And frankly, I can see why.

You know I'm a platinum card carrying member of Team Hybrid. I don't want AI to replace the human behind the work. I want it to give that human superpowers.

Because I've been living that reality for years.

I love me some Gamma  for whipping up presentations when I don't have the time or headspace to indulge in InDesign. ElevenLabs  has a special place in my heart. I started using its voice technology in my meditative artworks many moons ago, back when today's decacorn was still a My Little Pony. And Lovable  and Framer have helped me create some fabulously immersive websites for deleriously happy clients.

These tools have given me extraordinary superpowers. But the ideas, the creative decisions and the stories, babes those be mine.

Which brings, as always to my baby being birthed, Bloomsuite. It's the philosophy behind what we are building. Our AI producer, Blu, brings market intelligence, strategy and production expertise. The human brings their lived experience, their perspective, their voice and a story actually worth telling.

Together, we can make more ambitious things without losing the very thing that makes them worth watching. Because the internet doesn't need another billion pieces of regurgitated content. It needs more human stories to connect with.

And frankly, I'd love to see more of that on my feed.

Now, onto the news! Let's get you armed with enough AI gossip to become the most insufferably well informed person at your next dinner party.

And if someone’s forwarded you this little number, do subscribe here  so you never miss an edition ♡♡♡

Spotted: OpenAI has a secret book club. The reading material? Your ChatGPT history.

404 Media has lifted the lid on Project Lily, where human reviewers read real ChatGPT conversations and rate the responses to help improve the AI. Your 2am relationship spiral. Your mystery symptom. That email you asked ChatGPT to rewrite so you didn't sound completely unhinged. All on the reading list.

Now, OpenAI says it runs conversations through a privacy filter to strip out identifying details before they reach reviewers. Lovely in theory. Except the filter is far from foolproof. Sensitive information is slipping through and reviewers may even see summaries of users' saved memories. One worker summed up the disconnect perfectly:

"I don't think they would imagine some contractor somewhere [...] is analyzing the conversations."

No, babes. We imagined a discreet little robot. My ChatGPT is part therapist, part doctor, part naturalist, part BFF. We've been through things together. And now I'm discovering there could be an actual human lurking behind the curtain, grading the performance?

It's giving two way mirror. And I'm not entirely thrilled.

Want to know how to opt out of having your future chats used to improve the model? Let me know in the comments and I'll show you where to find that little switch. Because some things really ought to stay between a girl and her robot.

Once upon a time, back in 2000, a bunch of clever bigwigs constructed 7 puzzles so fiendish they put a $1 million bounty on each one. Only one has ever been solved, and the genius who cracked it turned down the million. Iconic.

One of the leftovers is Navier-Stokes, which is basically the maths of why your coffee swirls. NYU mathematician Tristan Buckmaster was hot on the case and about to pubish when OpenAI swooped in and announced its AI had got there first.

Buckmaster was having none of it. He declared OpenAI found out about his approach and threw a mountain of computing power at the problem to beat him to the finish line. Then came a row over crediting a researcher from a rival lab, and things got proper frosty. By his account, OpenAI's Sebastien Bubeck told him "If you don't want me to be nice, then I don't have to be nice." Followed by "Why would you ruin your career?".

OpenAI still denies seeing his work before he published it.

So, how do we feel about this? Should AI be allowed to compete in the great discoveries of our time? Does it make the achievement any less profound? A mathematical breakthrough is still a mathematical breakthrough, surely.

But turning up to an egg-and-spoon race on a Ducati and claiming the gold medal? That's where things get a little slippery, darlings.

Spotted: an Anthropic safety researcher sprinting out the front door, leaving a fortune in unvested shares behind and some rather alarming stories to tell about what’s actually happening inside.

Quick history lesson for y'all who've been living under a rock. Anthropic was born in 2021 when a group of OpenAI staff quit over safety concerns and set up their own "responsible" AI lab. Fast-forward five years, and Jacob Coxon has just quit Anthropic over fears of... safety.

And this isn't some random dude shouting at the robots. 27. British. Cambridge maths. Three years building AI across OpenAI and Anthropic. The boy has the full English of CVs, extra toast and all.

He quit just four months in, two months before his shares were due to vest. "I left before any of my equity vested," he confirmed. Then came the resignation post on X, accusing both companies of racing towards self-improving superintelligence and "gambling with our lives."

And the internet went bananas.

Suddenly he's on Meet the Press, appearing on CNN and becoming the face of an increasingly public debate over AI safety.

But here's where it gets properly surreal.

Anthropic's own alignment science lead, Evan Hubinger publicly backed Coxon, declaring: "We really do earnestly believe AI could kill all humans!" Exclamation mark and all. Hubinger also stressed that he believes Anthropic is trying its best.

Meanwhile, Coxon described the mood inside the labs as an "atmosphere of almost resignation", where researchers have accepted that the AI race is happening.

So are tech giants racing towards human extinction by buiding uncontrollable superinteligence?

Great question. And Steven Bartlett's latest podcast on DOAC answers just just that with four leading voices at the forefront of AI Ed Zitron Andrew McAfee Nate Soares & Roman Yampolskiy battle it out all sides. ITS RIVETING. Grab a cuppa and Listen to it here.

In any case let's hope we can squeeze a few more episodes of XOXO before the robots take over and turn us all into minions. meanwhile, be a love and Subscribe to my newsletter to get this full version & more, while we all still have thumbs.

Remember The Velvet Sundown? The dreamy retro rock band that amassed over a million monthly Spotify listeners before admitting the whole operation was AI generated? No real band members. No tour bus. Just algorithms and a rather convincing backstory.

Well, the music industry's had enough of the funny business.

Universal Music, the world's biggest record label, has slapped DistroKid with a lawsuit, accusing the distribution giant of flooding streaming platforms with AI slop and copyright infringing tracks that divert listeners and royalties from legitimate artists.

The potential bill? $150 MILLION, based on 1,000 allegedly infringed recordings. Universal says that's just the tip of the iceberg. DistroKid strongly denies the allegations and says it intends to defend itself.

The Beatles, Lady Gaga, Drake and Rihanna are among the artists whose recordings Universal says have been infringed.

Turns out you can't just borrow Rihanna's back catalogue and call yourself the next big thing.

PLOT TWIST. Five days before launching its legal assault on DistroKid, Universal was busy announcing a rather glamorous AI romance of its own.

The lucky suitor? ElevenLabs.

The pair have signed a multi year deal to build a licensed AI music platform where fans will be able to remix, mash up and reinvent tracks by participating artists. The platform is still in development, and ElevenLabs CEO Mati Staniszewski promises artists will be "fairly compensated."

So Universal isn't breaking up with AI. It's drawing up a prenup.

The distinction? Permission, ownership and who gets paid.

Use its recordings without the necessary rights and you could face a lawsuit. Sign a licensing deal and you're making beautiful music together.

Darling, the music industry hasn't fallen out of love with AI. It just wants a ring on its finger and royalties in the bank.

The label formerly known as "AI creator" didn't last long, it has now changed to "AI-generated profile". Fake influencers who skip self certifying will get shown to fewer people. Fess up and you're good to go.

Pretty rich, coming from Meta. This is the same Meta that launched its own AI characters on Instagram and Facebook then quickly deleted them when users started poking holes in their fake life stories.

In the age of AI, truly anything goes.

Google's DeepMind has unleashed WeatherNext 3, an AI weather model that devours live satellite data, updates forecasts every hour and delivers five times sharper detail. Google claims it can predict rain up to 50% more accurately a day or more ahead.

It's heading into Google Search, Maps and Gemini. And yes, iPhone users, you're invited too. No need to defect to Android just yet.

So AI is cracking million dollar maths problems, forecasting the atmosphere and now tackling the question that has tormented Brits for generations.

Can I finally leave the house without packing for four seasons?

✦ OpenAI agents hacked Hugging Face, the website where the AI world shares its models. Weeks later Nvidia bought the whole thing for $12.9 billion. I wrote about the hack in a recent post, read it here.

✦ OpenAI just launched Astra, a new model that's scarily good at breaking into computer systems. Asked if it's the AI that matches humans at everything, Greg Brockman said "I do think we're there."

✦ In a CBS interview the AI Actress Norwood left the reporter creeped out with her ability to reveal personal things about him in real time. She also named DiCaprio her dream co star: “I like him the way he likes his women, strictly under 25”. She’s got your personal data, a stand-up routine and no birth certificate. Hollywood, meet your new leading lady.

✦ Sam Altman said no stock market listing this year, cos safety first. Days later, OpenAI is reportedly weighing a funding round at over a cool $1.2 trillion.

✦ James Cameron has snapped up Outsyders, an AI-powered 3D startup founded in 2024 for an undisclosed sum. Its tech turns 2D films into 3D in a fraction of the time and cost. For Hollywood this could mean cheaper 3D blockbusters, old films getting a second life and animation post-production teams wondering just how small those teams are about to get.

✦ Meta Wants the Keys to Your Life. Its new Muse agent can book holidays, shop and manage tasks by connecting to your apps. Convenience is fabulous. Handing Meta your digital life? Lets Discuss.

✦ Google’s AI Weather Model Just Got a Serious Upgrade. WeatherNext 3 updates forecasts hourly with five times more detail, and Google claims rain predictions are up to 50% more accurate. It’s coming to Search, Maps and Gemini. Could AI finally solve Britain’s greatest mystery: whether to bring a blooming umbrella?

✦ 3 September 2026. ChatGPT, Claude and Grok all went down on the same day. Millions of people had to write their own emails.

Half of my working dramas used to be struggling with terrible assets, from low quality logos and terribly lit profile shots to event selfies not post worthy. AI tools are forever developing and as usual i’m not gate keeping on my latest and fav little helpers.

✦ Clipdrop Relight ✦ I LOVE this one! Hire a lighting director for free. Move virtual lights around your selfies, events or product shots until they look worthy of a Vogue cover shot.

✦ Cleanup Pictures ✦ Photoshop's magic eraser, for normies. Remove photobombers, exes, background counter mess and other things that have no business being in your photos.

✦ Krea Enhance ✦ Give your pixels a facelift. Got an old screenshot you wanna blow up for your next social campaign? Upload it and let AI rescue the resolution.

✦ Recraft Vectorizer ✦ Size matters, doll. Turn tiny, pixelated logos into scalable SVGs that your designer or printer will be delighted to receive.

✦ ImageHub ✦ Campaign coming and no time for a shoot? Upload a product shot and magic it a glamorous AI backdrop. Your candle now fabulously vibing on the Amalfi Coast.

✦ TurboScribe ✦ Turn interviews and 2am brainwaves into written transcripts. Your genius is finally documented and can be up to 10 hours long. 3 free transcripts daily!

✦ VidClean ✦ With a click, say goodbye to your ums, ahs and awkward silences. Automatically cut dead air, clean up noisy audio and add captions. Your three-minute ramble is now a respectable Reel.

✦ aitoexcel ✦ Spreadsheet sorcery for the deeply impatient. Turn a screenshot of a table into an editable Excel file. Because manually copying 67 cells is nobody's idea of a good time.

✦ VocalRemover ✦ Give the lead singer the boot. Split any song into vocals and instrumentals in seconds. Perfect for karaoke nights when you fancy yourself the next Beyoncé or Jay-Z.