As a reminder, each week I cover:
Reflections, insights, and learnings from the week's work and research;
Tools and resources relevant to entrepreneurship in the AI age;
Snippets and extra content related to The Complete Founder, particularly around entrepreneurial skills development.
The Week in Brief
👁️ Watching others succeed is a powerful way to build self-belief. A young founder's daily habits offer insight in AI-augmented thinking
🎒 Most travellers are still using AI only for basic itinerary planning. The innovation happens behind the scenes at service providers
😯 Anthropic's powerful new Mythos model has been taken offline following US government intervention. Possibly the first time a frontier AI has been withdrawn by official order
🍕In Founderval news: customer development and onboarding continues, and I'm exploring ‘snackable’ explainers to condense course content into shareable hooks
Vicarious Learning: What We Gain Watching Others
One of the Founderval mini-courses covers self-efficacy: how we practically develop belief in our own abilities. It outlines four methods, one of which is watching others complete a task. When we observe someone working, we see that something is achievable and start to understand what it would take for us to succeed.
As I upskill in AI, I mostly rely on one of the other four methods: mastery through practice. But I'm also drawn to watching how more advanced people operate.
This week, I found an interview with young founder Farza Majeed useful. A few highlights from his daily habits:
No phone for three hours after waking. He checks in with his mum and fiancée, then stops. Everything else — email, social, notifications — waits until his undistracted thinking is done.
Fifteen minutes of free writing before touching any task. He built his own app (freewrite) that forces continuous typing with no backspace. He uses it to convert raw notebook scribbles into prose, then feeds the result to Claude — turning rough notes into structured thought.
An AI agent cleans his inbox at 6am so he never has to. A script runs overnight, tagging every email as Attention, FYI, or something considerably less polite. He only opens the Attention inbox.
A personal Wikipedia built from his own life. He ran his iMessages, Apple Notes, and diary entries through an LLM pipeline that auto-generated linked articles on every topic, person, and company in his life. It self-updates and is queryable via an agent — like "a librarian who's an expert on you."
He uses Grok to consume Twitter, not Twitter itself. Once a day he asks for the top 10 AI tweets — model drops, announcements, no drama — instead of doomscrolling a feed.
Every product starts with one selfish use case. He doesn't build for an imagined audience first. He builds what he personally wants, ships fast, and lets the internet tell him what it actually is.
That last point encouraged me as it's how I prefer to start, despite it cutting against lean startup orthodoxy. His comment on building in public also stuck with me:
"I believe that when you put something out there, it changes."
Watch the 15-minute interview from Boys Club
How Are We Using AI For Travel?
I'm heading on holiday soon in the mountains, and it got me looking at how people are using AI for travel.
It's easy to imagine that savvy users now have agents doing everything — but research suggests the early majority are still largely using AI for basic tasks like itinerary planning.
The Verge put it well: "If you want to do the six most obvious things in any city on planet Earth, AI has you covered, but that's about as far as it goes."
Looking further ahead, the same piece explores the potential of Google's Gemini Spark. For those willing to trade privacy — handing over access to emails, documents, search history, photos, and messages — this concierge service operates at a different level entirely.
The Verge writer, planning a trip to Pennsylvania, found that nap times for his children and activities for his dog Frida had been scheduled automatically, without any prompting.
I spent a while hunting for novel ways that tech-savvy people are using AI for travel, convinced I was missing something. But most of the real innovation is on the service-provider side, with travel companies enhancing what they already do, like Hopper's price-watching tool.
In the end I accepted my fairly basic use of ChatGPT to identify what I'd need to pack for mountain weather, and went to buy my first fleece 🐑
The AI Model That Was Too Powerful
The big news this weekend was about Mythos and Fable.
Anthropic, the company behind the Claude chatbot, recently unveiled two new AI systems. Mythos was the company's most capable model, designed for a small number of trusted organisations. Fable was a safer public-facing version built on the same technology, with extra guardrails intended to prevent misuse.
The reason these models attracted attention was not that they could write better emails or create prettier images. Their strength was in areas like software security. They were reportedly extremely good at finding weaknesses in computer systems, the sort of capability that could help a cybersecurity team defend a network, but could also potentially help an attacker break into one.
Just days after launch, the US government stepped in.
Using export control powers more commonly associated with advanced military technology or semiconductor chips, officials ordered Anthropic to suspend access to the models for foreign nationals, citing national security concerns. Rather than try to separate eligible and ineligible users, Anthropic took both models offline entirely.
It may be the first time a frontier AI model has effectively been withdrawn by government order.
This story is probably less about one AI model and more about where the industry is heading. For years, the assumption has been that AI would follow the pattern of the internet: new technology appears, competition drives prices down, and eventually everyone gets access. Mythos and Fable hint at a different future.
The most advanced AI systems may increasingly be treated like strategic assets. Governments may restrict them. Companies may reserve them for trusted partners. Access may depend not just on what you can afford, but on where you live and who you work for.
One of my favourite resources for learning, AI With Kyle, warns us: “If you are engaging with this technology you are ahead of the world. That gives you an opportunity over the next couple of years to set yourself up. But that window is not staying open for long - models like Mythos/Fable are a warning shot.”
The Plan This Week
I'm enjoying showing people Founderval as the customer development phase continues. I'm offering a free in-person workshop for those I’m speaking to who are interested in institutional membership, to enhance engagement with users.
One challenge I keep returning to is how to create online content that feels genuinely useful but can be consumed quickly in an age of short attention spans. I'm considering developing TikTok-style explainers that condense the material even further — hoping they act as a hook for deeper engagement. I'd love to know your thoughts?
I'm planning to switch off entirely on holiday, so there'll be no newsletter next Monday. Reporting back in your inbox on 29 June 🫡
Have a good week!
Luke
🎩Hat-Tip I receive about a dozen AI newsletters a week. The one I always open is The Neuron — a good mix of news, tips, and resources.
LinkedIn | YouTube | Instagram | TikTok