Vol 14 ★ DeepMind is having a shake up, Claude now watermarking & LinkedIn AI snitch button lands

Busy month, darlings.

And a quick hello to new subs following my recent IRL workshops. Nice to have you on board! We kick off with a quick update, dive straight in to the latest news and round off with some of the tools you really should be getting involved with. Feel free to hop, skip to your fav bit - no hard feelings. This time.

So! Update…. Excitingly, our content creation app, Bloomsuite, is in the final stages of MVP development, which basically means doors open to beta testers very shortly. If you want a period of free access to our amazing platform, the limited Beta list might just have a few spots left… try here.

And no, before you ask, with Bloomsuite we are not talking about AI generated content. Nothing wrong with it, but that isn't what we're building. Bloomsuite uses AI to help creators make their own videos... beautiful, story led ones, with a focus on building your media channel. Your story, your face, your voice. The algorithm just holds your hand and helps you get there sanely.

I'm also back to some freelancing, which has excitingly had me testing more AI tools than ever before. One thing that’s come out of testing so many new tools is how important, for me, the hybrid role is. Human and machine, working together, each doing the part they're actually good at.

I spent a few hot Spanish summer nights revamping AI XOXO, carnival style. Seen it? No? Rude. Here ya go.

And to mark the occasion, I also opened the doors to my first AI XOXO workshop. The Ultimate AI Starter Workshop, built for the beginners in your life... and yes, possibly you.

Half an hour absolutely packed full of gems, online to do at your own pace. I take you through the five big AI models and which one you should actually be using for what, how to train Claude to work with you, setting some super important privacy boundaries to ensure your data isnt sold if the AI chat company ever is, prompting hacks & more. There's a goodie bag, obviously.

And just for you, my fabulous readers, code VIPXOXO gets the first 10 a cheeky launch discount. If you were looking for somewhere to start - this is it.

And now, on to the news.

Saira xo

★ FEATURE ★

Demis Hassabis, one of the defining architects of AI, and the man who built Google’s DeepMind lab is giving up the day-to-day CEO job to become Chair of DeepMind and Chief Scientist of Alphabet, freeing him to focus on the bigger stuff: protein folding, drug discovery and casually, the future of human intelligence.

Then we have Jeff Dean, Google’s longtime Chief Scientist, who co-founded Google Brain and helped shape the company’s entire AI research strategy, also leaving the circus to start his own AI research company, Discovery Loop. (Google is a founding investor, so he isn’t exactly storming out lighting a match behind him).

Meanwhile, reports continue of top researchers drifting toward Anthropic, the younger rival behind Claude. AI research firm SemiAnalysis added fuel to the fire, arguing that DeepMind was no longer operating at the very frontier of AI. The Verge thinks that verdict goes too far, and Google insists it still has “all the ingredients to lead from here.” Still my lovelies… I feel its a curious amount of furniture moving for the lab that kept its rivals quivering.

Another thing that caught my eye; Demis casually ending Google I/O (event) by telling the world we are standing at “the foothills of the singularity.” Not AI is getting useful. Not Gemini is getting better. The singularity . Essentially his vision of the point where AI becomes powerful enough to transform society in ways we can barely predict.

Just pause on that for a second, because as I have been harping on about in previous editions - the AI race is framed around reaching AGI: systems “capable of doing economically valuable knowledge work at, or beyond, human level”. In Silicon Valley terms, whoever gets there first doesn’t just launch the best chatbot. They rewrite the hierarchy. And yet Google appears to be pointing some of its brightest minds toward medicine, biology and humanity-sized problems rather than simply sprinting for the trophy. Then again, we cant forget Google is the trust-fund baby of AI: perhaps it can afford to save humanity and win the race.

While Demis moves upstairs to think about civilization-level science, we see Koray Kavukcuoglu taking over DeepMind’s day-to-day operations; pushing Gemini and the products people can actually use. Maybe this is the master plan: let Demis chase the future of humanity while the machine downstairs churns AI into a business.

Or maybe the old DeepMind… the pre Google version, with its slightly idealistic, slightly mad scientist ambition to “solve intelligence,” is finally resurfacing.

While Silicon Valley races toward AGI, Demis is turning his attention to what that intelligence could actually do: protein folding, drug discovery, disease. Perhaps he isn’t stepping away from the race at all. Perhaps he’s decided winning looks different now.

I love that for him, for us.

Side shizzle: while Google appears to be turning one of AI’s biggest minds toward medicine and humanity-sized problems, over at OpenAI its only dedicated ethicist has left after less than a year alongside a string of senior safety figures. Curious timing all round…..will explore that next time.

★ SIZZLERS ★

🎠 Claude Watermarking Everything it Touches

Anthropic is about to make AI transparency very literal. To comply with new EU rules, Claude generated content will carry a digital watermark. But sadly for those of us < yours truly included > writing something yourself and using Claude to grammar check made up words and spellchecking it could still leave the whole thing wearing a “made by Claude” badge it frankly, doesn’t deserve.

OpenAI and Google are heading in the same regulatory direction too, so this pretty much less of a Claude quirk and more of a glimpse at where all AI assisted content is going. I’ve already seen people admitting to adding spelling mistakes back in and removing every godforsaken dash just to prove something was written by a human.

So humanity is now deliberately making its work worse to prove it wasn’t made by AI. Cute. At this rate, maybe we’ll reach AGI faster simply because the humans are voluntarily lowering the bar.

🎠 Spotify Checking for AI User ID at the Door

From mid September, Spotify starts slapping an "AI Persona" badge on artist profiles it flags aren't human. Get badged and you're out of editorial playlists and every algorithmic recommendation. The only way anyone hears you is if they already follow you or type your name in. You’re effectively buried.

However this doesnt mean we kiss goodbye to AI generated music. The badge is about who you are, not how you made it.

In Spotify’s own words: it's "about the artist's public identity, not about how the music was made." So you can artificial intelligently generate every note, every lyric, every drum fill and stay perfectly clean so long as a human exists at the end of it.

Making fake music is fine. Being a fake person is the crime.

🎠 The LinkedIn Saga: From ‘Use AI’ to ‘Report AI’

LinkedIn has officially given us a button to report AI slop which means those chronically annoying “let that sink in” and “I wasn’t going to share this…” posts cluttering your feed can now be slammed behind bars with a satisfying tap.

And why should we all be using this snitch button? Well the platform will be using those reports to teach its systems what low quality AI content looks like and reduce how often it gets recommended.

Chief Product Officer Hari Srinivasan said “AI slop is a top priority for all of us. We really care about this. People come to LinkedIn to connect with real people and share their real perspectives, ideas, and expertise.”

Which is all very noble until you remember LinkedIn itself had an “enhance your post” feature designed to rewrite your posts, in AI. Needless to say that feature is now being pulled and replaced with simple proofreading. So, ya, LinkedIn encouraged us to let AI rewrite our thoughts, watched the feed fill up with AI shhhlop and then gave us a button to blitz it

A fabulous little circle of life.

🎠 Snap Says Bring a Human or Lose the Algo

AI video is having a rather awkward coming of age moment.

Synthesia raised $200 million at a $4 billion valuation this year; Higgsfield says its annualised revenue run rate has exploded to $500 million, largely from companies making commercial video and advertising; and ElevenLabs primarily voice/audio rather than a full AI-video generator is now valued at $11 billion.

It’s pretty fair to say money is absolutely flooding into tools that let brands and creators produce AI generated video content.

Just as the technology gets frighteningly good and investors are piling in, the likes of Snapchat are now drawing a distinction between making AI content and rewarding fully automated AI content with distribution.

Snap has decided fully AI generated videos will no longer qualify for Spotlight recommendations or creator rewards. You can post it, darling, but we’re not helping anyone find it.

AI assisted videos however; Absolutely fine.

Which exposes a rather interesting contradiction at the centre of our AI boom: we’re spending billions making it easier to create infinite synthetic content at exactly the moment platforms are deciding humans are the premium product.

★ WHATS IN MY STACK ★

I’ve had a v heavy month of AI tools. On that note, I’m loving being back to freelancing, working with new brands on everything from rebrands to websites and social integrations. I’ve remembered how much I love that slightly unhinged pace of having to deliver something amazing… yesterday.

Over the past few months however, honestly my stack is changing as i’m not interested in tools that insist on doing my entire job for me. I think I am slightly over that hype.

As a creative, and a Producer, I need to be able to get in there, move things around, change my mind, add some bespoke code, branch & play with 5 different versions any one time. AI tools/bots that automate the boring bits and getting me to the fun creative stuff faster - hell yea, take my subscription fee.

🎪 Claude + Chrome

OMG, it has finally arrived and I am here for it.

Claude Desktop now works with Claude in Chrome, which basically means Claude is no longer trapped inside a chat window who’s entire existence and conversation disappears every time it has a moment. It can actually work with you inside your browser, looking at pages, clicking around, filling things in helping me complete tasks across websites I am logged into.

For someone like me who constantly has 97 tabs open while researching, building, designing, animating, exporting different formats for assets whilst QCing client deliverables, this feels like the beginning of a very different way of using AI.

There are caveats. It’s still beta, you’re giving AI visibility into the browser tabs you let it access, and I’m certainly not sending it unsupervised into anything highly sensitive. Useful assistant - oh, absolutely. New holder of my banking password - not today princess.

🎪 Framer

When I was asked to brand Tavyaan within 16 hours, there is simply no way I could have done it at that speed without highly selective AI tools.

Framer, was one. What I love is that it sits in this sweet spot between automation and proper design control. It can accelerate the building, layouts and repetitive bits, but I can still manually control the details and add in bespoke bits, which is essential because I cannot work with a tool that decides it knows better and then hides all the buttons - or a code input - from me.

It meant I could go from new brand direction to a functioning, beautiful website ridiculously quickly without the usual design > developer > revisions > existential crisis relay race.

i wrote a Linked in Post here about the project, check it out - let me know in the comments when you do!

🎪 Formspark

This teeny tool that solved the decidedly unsexy problem of setting up Framer forms, as Tavyaan had a few, one which collected data from various bespoke fields.

Formspark connects to forms on your site and handles the submissions in the background, without needing me to build a backend or start pretending I understand server infrastructure. I hooked it into Framer, tested it, and after a bit of tinkering, it just worked.

Which is increasingly my favourite category of software: boring problem, solved.

🎪 beehiiv

Who knew a newsletter platform could make a website this gorgeous?

In case you hadnt noticed (rude) Ive given XOXO a rather major upgrade. What started as a simple Beehiv AI newsletter to keep friends amusingly informed has basically become a mothership. The new website lives there, the newsletter lives there, my first workshop is live there, and it’s capturing the people coming through the door and helping me segment them based on what they’re actually interested in.

beehiiv can build audience segments using subscriber data and engagement, while its website builder handles proper branded sites, landing pages and funnels alongside the newsletter itself. Yes, it does have an AI to help with all of tis but tbh it made too many mistakes which I ended up having to boringly unpick. For me, manually building was still the better way forward. I did however ask Claude Chrome to step in with some of the automative tasks, that bored me - like replacing titles across the site to test different ones or resizing images to make them more responsive.

Visually I’m obsessed. Who said a newsletter platform had to look like a newsletter platform? My website would like a word.

🎪 Photoshop Generative Upscale

I don’t normally shill the expensive tools here, but when a client sent me a folder full of tiny, tragic, low res images < faces included > on a deadline that did not leave room for miracles. I had basically accepted defeat when I remembered i had been meaning to test out Adobe’s latest Generative Upscale. So I threw one in and… smack, bang, suddenly 2007 had entered 2030. The images came back actually usable. Faces intact, detail restored, crisis downgraded.

Yes I am fortunate enough to have Photoshop licence but Yes, that was highly worth my sanity.