In partnership with
Welcome Automaters, 👋
So guess what?? There’s a fresh three-letter acronym taking over AI Twitter, research papers, and startup pitch decks at an alarming speed. It’s called RSI, short for Recursive Self-Improvement, and it is the wildest concept in tech right now.
In basic terms, RSI refers to an AI system that can continuously upgrade itself, completely eliminating the need for us. Think of it like a video game character who levels up, then uses those shiny new powers to level up even faster, until you, the player, are completely irrelevant. Fun, right?
A new wave of elite AI labs has started baking RSI directly into their roadmaps. They’re turning it into the ultimate buzzword for a potentially cataclysmic AI leap, even if the industry is still bickering over the fine print.
For example, legendary AI researcher Andrej Karpathy, famous for his work at Tesla and OpenAI and now on Anthropic's pre-training team, has been publicly building a project called Auto-Research. He is using entire swarms of AI agents to train models on small tasks, regularly posting updates on X and sharing the code publicly on GitHub. Karpathy himself admitted the work is not exactly groundbreaking yet, but he has inspired a whole crowd of researchers to chase the exact same dream.
He is definitely not alone. Startup Adaption, founded by Google and Cohere alumna Sara Hooker, just launched a similar tool called AutoScientist, which is designed to automate the process of training frontier AI models from the ground up. Meanwhile, Disarray founder Doris Xin made major waves when her self-trained machine learning agent won 28 medals at a Kaggle competition, casually beating out many human-built systems.
Here’s where it gets spicy. Once these AI systems can manage the upgrade cycle better than humans, the process becomes a completely closed loop. At that point, the evolution of intelligence is limited only by the compute power they can access. Humans are no longer necessary, or even helpful.
But before you start building your backyard apocalypse bunker, let us pump the brakes for a second.
Google CEO Sundar Pichai acknowledged in a recent podcast interview that while the industry is making steady progress, the kind of sci-fi RSI people dramatically describe represents a whole next level of acceleration. Meaning: we are simply not there yet.
Helen Toner, director of Georgetown's Center for Security and Emerging Technology and a former OpenAI board member, is also keeping it real. She notes that simply using AI tools to do AI research does not count as true RSI. The classic, hardcore definition requires zero human involvement in the loop.
The bottom line? RSI is incredibly real, the tech elite are chasing it with everything they have, and nobody knows exactly when it arrives, or what it even really is, Classic AI energy.
And don't forget: 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 Dropped Claude Opus 4.8 Unexpectedly Soon
Buckle up, because Anthropic just served a surprise upgrade nobody fully saw coming.
Yes, really. Just 41 days after Opus 4.7 landed, Anthropic dropped Claude Opus 4.8 on May 28th. I mean, that's a shockingly fast turnaround, even for a company that moves at lightning speed. For context, the current Sonnet and Haiku models are three and seven months old respectively. So yeah, something clearly lit a massive fire under this team.
That "something" might have a lot to do with the internet's lack of chill. Opus 4.7 received a noticeably chilly reception, with users publicly calling the release disappointing across X and LinkedIn. Add to that the intense pressure of OpenAI's Codex and Google's Gemini Flash both dropping major updates recently, and suddenly the hustle makes total sense.
Anthropic simply couldn't afford to fall behind the gossip mill.
So What Actually Changed?
The biggest upgrade is honesty. No, really. Opus 4.8 is designed to be way better at admitting when it does not know something, flagging shaky inputs and uncertain outputs rather than just winging it like an unprepared student.
Early testers at Bridgewater Associates confirmed this was the absolute standout difference. According to them, the model proactively spotted problems that earlier versions, and honestly, competing models, just left for humans to stumble upon. If you ask me, that's a genuinely big deal for professional users who need facts, not fiction.
Naturally, Opus 4.8 also comes with the expected best-in-class benchmark results, proving it has the brains to match its new attitude.
The best part? Pricing stays exactly the same as Opus 4.7, and it is available everywhere right now.
But that’s not all: Alongside the new model, Anthropic launched Dynamic Workflows in a research preview. Think of it like a boss AI that can command hundreds of smaller AI helpers, all working at the same time on one giant task.
Anthropic says Claude Code alongside Opus 4.8 can now handle huge coding jobs across hundreds of thousands of lines of code, start to finish, completely on its own. It’s giving major group project manager energy, but without the slacking.
And About That Mysterious Mythos Model...
Of course, we have to talk about the forbidden fruit. Anthropic is still keeping its most powerful Mythos model locked away after some pretty serious cybersecurity concerns surfaced during a preview last month. But the company hinted in today's launch post that the wait may be nearly over, writing that the necessary safeguards are coming together fast.
In their own words: “We’re making swift progress on developing these safeguards and expect to be able to bring Mythos-class models to all our customers in the coming weeks.”
So yeah, watch this space, because the AI wars are only getting hotter. And yes, as usual, meet us on YouTube so we can dive into all this together.
hCaptcha Pro: Privacy-First, Frictionless Bot Mitigation
Don’t sacrifice user privacy for sub-par protection.
hCaptcha Pro stops bots and malicious automation with zero friction for real users.
Protect your platform with 99.9% Passive Mode while preserving user privacy.
Sign up for hCaptcha Pro today.
🧱 Around The AI Block
🤫 Medieval secrets is being revealed by AI.
📲 How to run a local AI chatbot on your iPhone.
🤳 Sneak peek at new Siri app reveals Apple’s plans to take on ChatGPT and more.
🤑 Anthropic reaches valuation of $965bn, beating OpenAI to become world’s most valuable AI firm.
💱 Just like gold and oil, we’ll soon be able to trade AI token futures.
👍 Microsoft debuts a more buttoned-up look for Copilot.
🔎 Google AI Mode & AI Overviews add Preferred Sources & New Carousel.
👍 Sesame, the conversational AI startup from Oculus founders, launches its iOS app.
👉 YouTube adds new podcast features, including an AI recommendation tool and ‘Auto speed’.
🗣️ Founder of French AI startup Mistral pushes back against Pope's criticism of AI use in warfare.
🎉Introducing: AI Tutorials
Let's be totally real for a second: the AI wave isn’t just coming; it is actively reshaping how we make a living.
You’ve probably heard the saying floating around, but it bears repeating: AI itself might not replace your job, but the person who actually knows how to use AI absolutely will. And that’s why our mission has always been to keep you three steps ahead of the curve, and today we are taking that to a whole new level.
We want to make you completely, undeniably irreplaceable in your industry. To do exactly that, we’re officially launching our brand-new AI Tutorials section! 🥳
🧐 What You Can Expect:
We’re completely stripping away the boring academic fluff and useless tech jargon to bring you highly practical, step-by-step videos that’ll dive straight into the hottest AI tools dominating the market, showing you exactly how to apply them to your daily workflow, automate the tedious stuff, and supercharge your output.
Check This out: How to Control Your Desktop from Your Phone Using Claude (Dispatch Setup).
And: 80% of NotebookLM in 13 Minutes.
So tell us, what’s the single most annoying, tedious task in your daily workflow that you desperately wish an AI could just handle for you?
Hit reply and the next video might just be around your exact problem!
Last week Viktor wrote a brief, built a landing page, and opened a pull request.
Last week, Viktor wrote a campaign brief, built a landing page, opened a pull request, generated a board-ready PDF from live Stripe data, and sent a follow-up email to a churned customer. All from Slack. Same colleague that also pulls your reports and monitors your dashboards. 5,700+ teams. 3,000+ integrations.
🛠️ Trending Tools
For the Home Improvers: OutdoorBrite lets you snap a photo of your messy backyard and watch this AI instantly render photorealistic landscaping, pools, and gardens. It’s the fastest way to "test drive" a renovation before you touch a shovel.
For the AI Marketers: ZeroRank AI is essential for brands; it tracks how often you're mentioned in AI search (like Perplexity or ChatGPT). You can also use its one-click engine to optimize your content so the AI actually recommends you to users.
For the Global Students: Studocu is a massive community library that gives you access to millions of shared study guides, lecture notes, and practice exams from top universities. It’s the ultimate shortcut for mastering a tough subject when your own notes aren't cutting it.
For the Productivity Architects: Goblin.tools is designed for the overwhelmed, this collection of tiny AI tools breaks down massive tasks into manageable steps. It can even "humanize" your “spicy emails or estimate exactly how long a chore will actually take.
🤖 AI Workout Of The Day: Generate Winning Copy Variants with AI
Struggling to figure out which version of your copy actually connects with your audience? Don’t guess, test!
This prompt turns AI into your personal direct-response copy expert, generating multiple high-impact copy variants while explaining why each one works. You’ll get actionable options for your headlines, email subject lines, ad copy, or landing pages; complete with the psychology behind each choice so you can make data-driven decisions.
Here’s How to Use This Prompt Effectively:
Provide detailed context: Describe your product/service, target audience, and the channel where the copy will appear (email, landing page, ad, etc.).
Set a clear goal: Define the metric you’re optimizing for — clicks, sign-ups, downloads, etc.
Explain your preferences: Include any tone, style, or formatting requirements (e.g., concise headlines, playful social posts, formal email).
Clarify Output Preferences: Do you want a table, bullets, or full text ready to deploy? Be explicit.
Iterate: Use the generated options as a springboard. Test, refine, or combine elements to create the most compelling copy.
💡 Prompts to try:
I’m [insert detailed description of your problem and context]. Act as a direct response copywriter. I want to test two variations of copy for [insert context, e.g., landing page headline, email subject line, ad copy]. The primary audience is [insert target audience], and the goal is [insert goal, e.g., clicks, sign-ups]. Generate 5 A/B test variant options for the copy.
For each option, provide:
-The copy itself (clear, concise, and persuasive)
-A short explanation of the psychological principle, persuasion tactic, or emotional trigger it uses
-Recommendations for when and how to deploy this variant effectively
I want the output to be [insert formatting or structure preference, e.g., table, bullet points, ready-to-paste text].
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.
