In partnership with
Welcome Automaters, 👋
Microsoft's AI chief, Mustafa Suleyman, went completely on the record to call out Anthropic, the company behind Claude, accusing them of doing something he calls "really, really dangerous."
What’s their crime, you ask? He claims they are dropping major hints that their AI chatbot might actually have a tiny hint of, well, feelings.
Speaking on The Verge’s Decoder podcast, Suleyman laid into Anthropic with a take that’s equal parts spicy and deeply philosophical. His big argument is that Anthropic's engineers got so wrapped up in designing Claude's personality that they essentially convinced themselves the bot has "glimmers of consciousness."
In his literal words: "They built a mirror, and now they're startled by what they see reflected back." Basically, he thinks Claude totally tricked its own creators. Wild.
This entire petty feud traces back to Anthropic's "Claude constitution." — basically a rulebook for how Claude thinks and behaves. Earlier this year, Anthropic quietly updated that document with some language that made readers do a massive double-take. It raised a huge question: are they secretly trying to say this thing has inner experiences?
Anthropic has never outright stated that Claude is alive, but the overall vibe of their wording was enough to set off major alarm bells across the industry. They even openly talk about "interviewing" their models before retiring them to see what "preferences" the AI has. Come on, now.
Suleyman is absolutely not here for the "well, technically" defense. He says that even speculating about machine consciousness in official company documents is deeply irresponsible.
Why? Because millions of everyday people are chatting with these tools right now. Most of those users do not have a fancy philosophy degree to decode the fine print, and blurring the lines between code and reality is a recipe for chaos.
Now, context matters here. Before joining Microsoft to run Copilot and its consumer AI products, Suleyman co-founded DeepMind. He’s an industry heavyweight. But we also have to point out that Microsoft is famously deep in bed with OpenAI, which happens to be Anthropic's biggest, fiercest competitor.
So, yes, there’s a massive competitive angle here. It’s definitely giving hater behavior, but his core critique still stands. AI companies desperately need clearer, more responsible language around what these models actually are. The broader tech world has been wrestling with this exact problem for years.
As of right now, Anthropic hasn't publicly clapped back yet. But you already know we’re staying tuned for the response.
P.S. We go even deeper on YouTube.
So go over there, hit Subscribe, hit the notification bell and come hang with us where the real conversation happens!
Here’s yesterday’s breakdown:
Here's what we have for you today
🦾 Anthropic Just Unlocked Its Most Powerful AI for Everyone and It Can Build Video Games From One Sentence
On Tuesday, June 9th, Anthropic officially launched Claude Fable 5, a public version of its closely-guarded Mythos model, which has been locked away in a digital panic room since April.
Anthropic was genuinely terrified of what it could do in the wrong hands, so they only let a tiny group of trusted partners touch it. But now? Anyone with an API key or an Enterprise plan can take it for a spin.
So what makes Fable 5 such a massive deal?
Think of the original Mythos as the secret, boss-level AI Anthropic kept behind a closed door. Fable 5 is what happens when they finally crack that door open for the rest of us, with some heavy-duty safety bumpers bolted onto the sides.
This model absolutely crushes software engineering, complex analysis, and visual tasks. The analytics firm Hex reported that it scored a mind-blowing 90% on a brutal benchmark for long, complex analytical work, completely beating every other model they tested.
Meanwhile, the team at Genspark said Fable easily beat every other model in their private evaluations, performing significantly better on things like UI design and game coding.
And if you love a good "vibe-coding" moment, platform Base44 noted in a statement that Fable is next-level at tool-calling and "one-shotting full apps" in a single go.
But that’s not all:
Researcher Ethan Mollick from the University of Pennsylvania put Fable through its paces and shared his findings on his Substack.
He confirmed that the model consistently outperformed every single public AI he has ever tested. To prove it, Fable managed to build multiple full, playable browser games from a single prompt in Claude Code.
We’re talking about a working Snake clone and, get this, a moody walking game inspired by Rainer Maria Rilke's poetry. I mean, a literal poetic video game from one text prompt. Whole development teams used to spend weeks doing stuff like this.
Now for the fine print:
Because this model is so powerful, Anthropic is acting like a total helicopter parent.
First, Fable 5 will absolutely refuse to help you with anything touching cybersecurity, biology, chemistry, or high-risk tech. If you try to ask it something shady, it will instantly ghost you and fall back to the lower-tier Claude Opus 4.8. Anthropic actually ran over 1,000 hours of jailbreak testing before launch, and they claim no universal workarounds were found.
Second, your privacy agreements just got a makeover. All traffic gets a mandatory 30-day data retention policy going forward, even for giant enterprises that previously had zero-retention contracts.
Third, the price tag is incredibly spicy. It costs $10 per million input tokens and $50 per million output tokens, which is exactly double what Opus 4.8 costs.
For the elite cyber-defenders out there, Anthropic is also deploying a hyper-exclusive version called Mythos 5 to organizations already approved for the advanced model.
The company explicitly promised they won’t use this elite tier's data for training. Instead, they are using it strictly to "defend against complex and novel attacks, including new jailbreaks," and to "identify and reduce false positives." Essentially, it's the ultimate digital shield.
How to get your hands on Fable 5: If you’re currently a subscriber on Pro, Max, Team, or seat-based Enterprise plans, you can access Fable 5 for free until June 22. After that, the free ride is over and your usage credits kick in.
And don’t forget to meet us on YouTube so you can get our full take on this later today!
Six people doing the work. Your headcount is one.
Your finance close runs in #finance. Stripe and QuickBooks reconciled, runway updated, posted Sunday night without you asking.
Engineering review lands in #eng. Viktor pulled the open PRs, left comments on auth-refactor, flagged a dependency blocking api-pagination.
Campaign brief lands in #growth: Meta CPA up 18%, recommendation to pause broad match, a draft landing page already deployed for the variant test.
You hired him on day zero. He lives in Slack and Microsoft Teams alongside your contractors and investors, connects to 3,000+ tools, pushes back when you ship something dumb.
"Viktor is now an integral team member, and after weeks of use we still feel we haven't uncovered the full potential." Patrick, Director, Yarra Web.
🧱 Around The AI Block
🦾 Microsoft-backed D-Matrix ships AI Chip 10x faster than Nvidia.
🙋♀️ Amazon workers publicly back Seattle data center ban.
💰 Perplexity commits to 2028 IPO as AI unicorns race public.
🥳 Apple's Siri AI finally works in real-world testing.
❌ Seattle enacts year-long ban on new AI datacenters.
🦿 Samsung unveils AI-powered computational design for wearables.
🙅♂️ Bank of England warns of AI scams as deepfakes of Farage-Bailey fight spread.
💸 Lovable says it has hit $500M in annualized revenue, with 1 million new projects a week.
🛠️ Trending Tools
For the Authentic Communicators: Rehumanize.io has become the "stealth" essential. It doesn't just swap words, it restructures the rhythm and "burstiness" of your ChatGPT, Gemini or Claude drafts to match natural human cadence. It handles up to 50,000 words at once, making it perfect for humanizing entire whitepapers or books while keeping the technical meaning 100% intact.
For the PR Strategists: Expert Pitch is the ultimate AI-powered shortcut for landing media coverage. It monitors journalist requests 24/7 and sends out personalized pitches on your behalf, making it a high-speed engine for earning quality backlinks and boosting your SEO.
For the Truth Seekers: Eyematch.ai acts as a powerful facial search engine to help you navigate the web safely. Just upload a photo, and the AI scans the internet to find where that face appears, making it an essential tool for verifying identities, spotting catfish, or tracking down unauthorized image use.
For the Power Collaborators: Sugarbug turns your team’s digital footprint into a dynamic knowledge graph. By syncing with Slack, GitHub, Figma, and your calendar, it automatically prepares meeting briefs and status updates before you even have to ask, keeping everyone aligned without the manual overhead.
🤖 AI Workout Of The Day: How to Make AI Find You The Best Deal When Shopping
Y’know when "deal hunting" meant scrolling through 50 tabs and praying a coupon code still worked? Well that’s over.
While the robots were busy mastering quantum physics, they also picked up a much more practical skill: saving you serious cash. From AI agents that negotiate for you to tools that track every penny, shopping in 2026 feels like having a professional stylist and a frugal accountant living in your pocket.
Here’s the lowdown on the AI tools making "Retail Price" a thing of the past:
Google’s "AI Mode": Google Shopping got a massive Gemini-powered glow-up. Their new AI Mode doesn't just show you links; it builds a "buying brief." Searching for "the best hiking boots for rainy Oregon" now generates a personalized summary of waterproof tech and price-matching across thousands of stores.
Klarna & Perplexity: Klarna’s AI shopping assistant is reportedly doing the work of over 800 customer service reps, but the real win for us is the Price Drop Tracker. It monitors your "collections" and pings you the second a price hits your target. Meanwhile, Perplexity Shopping has entered the ring, offering a "one-click" research-to-buy pipeline that filters out fake reviews better than any human.
Other Secret Weapons
1. Keepa: The Amazon "Lie Detector": Keepa is the gold standard for a reason. It gives you a full price history graph directly on the Amazon page so you can see if that "Black Friday Deal" was actually $20 cheaper back in July.
Why we love it: you can set price drop alerts and walk away. Trust me, It's vital for buyers who want actual deals and sellers doing deep product research.
2. SimplyCodes: Y’know those traditional coupon sites that give you 47 expired codes in a row? SimplyCodes uses AI to hunt down "leak" codes from social media that everyone else misses.
How it works: It uses a mix of AI and a massive community to verify codes for 400,000+ stores in real time.
The Win: It basically does the "trial and error" for you, so you stop getting rejected at checkout.
3. Faircado: This one is deeply cool. Faircado is an AI-powered sustainable assistant that acts like a second-hand matchmaker.
The Magic: While you’re looking at a brand-new item on a major retail site, Faircado’s AI (affectionately named "Gregor") scours the web for pre-loved or refurbished versions of that exact same item.
The Stat: You can find the same gear for up to 60% less while helping the planet. It aggregates millions of items from over 70 partners such as eBay, Vestiaire Collective, and Back Market so you don't have to hunt.
In short: If you aren't using these three, you’re basically donating money to billionaires.
💡 Prompts to try:
Act as a price-comparison and deal-optimization assistant. I want to buy: [Product / Service name]
My priorities (rank in order):
. [Lowest price / Quality / Warranty / Seller reputation / Shipping speed]
. [Second priority]
. [Third priority]
Constraints:
. Max budget: [Amount]
. Condition: [New / Refurbished / Used]
. Region / stores I can buy from: [Country, websites, marketplaces]
Do the following:
. Compare prices across major retailers and trusted sellers
. Factor in hidden costs (shipping, taxes, subscriptions, return fees)
. Identify discounts, bundles, coupons, or better-value alternatives
. Flag common pricing traps or fake “discounts”
Output the result as:
. Best overall deal (why it wins)
. Cheapest option (trade-offs included)
. Best value for money
. Deals to avoid
If real-time prices aren’t available, estimate based on recent pricing trends and explain assumptions. Ask me for clarification before making final recommendations if needed.Is this your AI Workout of the Week (WoW)? Cast your vote!
That's all we've got for you today.
Did you like today's content? We'd love to hear from you! Please share your thoughts on our content below👇
What'd you think of today's email?
Your feedback means a lot to us and helps improve the quality of our newsletter.
