We sat down with Leor Gayr from Exploring ChatGPT to discuss things like building a 60K+ subscriber newsletter through daily publishing with zero outside marketing, why AI is a deeply individual experience rather than a universal tool, the human-in-the-loop content model that keeps philosophical writing from becoming AI slop, why the next big AI topics will be agents, security, and ambient AI environments, and the counterintuitive lesson that flooding your distribution channels might get the same results as doing it by hand.
Plus: the case against perfectionism, what the “how to make money with AI” course bubble signals, and why Substack feels fundamentally different from every other platform.
About the guest
Leor Gayr is the creator of Exploring ChatGPT a Substack newsletter and YouTube channel where AI, philosophy, and speculative science meet. In a year of daily publishing, he has grown Exploring ChatGPT to 60,000+ subscribers entirely through organic Substack growth, no outside advertising, no paid promotion. He runs the newsletter alongside a day job and a real estate business he founded years ago, with a new baby arriving next month. He also wrote a book using AI-assisted writing — the same human-outline-plus-AI-expansion model he uses in his newsletter.
Connect with Leor on Substack and YouTube.
About the host
Peter Benei co-founded AI-Ready CMO, the daily intelligence platform for senior marketing leaders. Peter has been serving as a CMO, marketing leader, and consultant to high-growth B2B scaleups for the past 10+ years. He has a background in advertising, working with Fortune 500 brands.
Connect with Peter on LinkedIn or read his newsletter.
About the show
AI-Ready CMO Live is an unscripted, unedited conversation about AI and the practical applications of AI systems for marketing. We share case studies, examples, and use cases from a wide spectrum of marketers, creators, influencers, service providers, and domain experts.
Want to join us as a guest?
If you have a great case study to share or a valuable use case to discuss, contact us at info@aireadycmo.com. We would love to hear from you and host you as a guest on our livestream.
Top 10 Takeaways
Daily publishing is a willpower game, not a skill game. — Anyone can use the human-outline-plus-AI-expansion workflow. Not everyone has the discipline to show up every single day for a year.
AI is an individual experience. — ChatGPT trained on a year of your conversations will give you fundamentally different answers than it gives someone who opened an account last week. Time in the seat is a competitive moat.
The reflections stay human; the meat can be AI. — Lior’s intros and closing reflections are 100% his voice. The body content is AI-expanded from his outlines. The point of view is always his. That’s the line.
Separate your chats, separate your contexts. — One ongoing chat for Substack. One for YouTube thumbnails. One for future projections. Mixing contexts degrades quality. Most people overcomplicate what is actually a very simple setup.
AI agents and AI security are the next two big topics. — Agents are the tree. Security is the underreported branch that will matter most. Human judgment and critical thinking are the counter-narrative already gaining traction.
Flooding distribution works at scale — if you’ve built the library first. — Automated AI publishing at 2-3x daily gets the same engagement numbers as handwritten posts, once you have 6+ months of content for the agent to learn from. The library is the prerequisite.
59K subscribers. One year. Zero outside marketing. — Entirely within Substack. No LinkedIn, no ads, no newsletter swaps. The algorithm rewards consistency and personality. Leor ran the experiment, and it worked.
The “how to make money with AI” course bubble is a tell. — When platforms algorithmically reward content about the platform itself, you get a feedback loop of repetitive, low-value courses. Recognize it for what it is.
Perfectionism is the enemy of daily publishing. — The advice isn’t “do it better.” It’s “give up on perfect.” The worst outcome is embarrassing yourself online. Most people survive that. The content you’re most proud of often performs the worst.
Substack engagement happens in email, not on the platform. — 700,000–800,000 monthly views come from daily email opens, not platform comments. The engagement is there. It’s just invisible to people looking for likes and comments.
5 Topics Worth a CMO’s Attention
1. The human-in-the-loop content model at scale
The standard debate about AI content is binary: either you write it yourself, or AI does. Leor runs something more nuanced — and more replicable.
His model: every article starts with a human idea, captured on a phone the moment it appears. He writes bullet points. Builds an outline. Takes that outline into ChatGPT. The intro and reflection sections — the parts that carry point of view — stay 100% human. The body content is AI-expanded from his structure, then reviewed, edited, and fact-checked before publication.
This isn’t “AI writes my newsletter.” It’s a division of cognitive labor. Leor owns the thinking. AI handles the expansion. The result reads as philosophical and dense — his Substack covers topics most AI newsletters wouldn’t touch — because the ideas are genuinely his, even when the sentences aren’t.
For marketing teams, this reframes the AI writing conversation. The question isn’t whether to use AI. It’s where to draw the line. Leor’s line is clear: point of view is human-only territory. Everything else is a production problem AI can help solve.
The corollary: on days without ideas, AI becomes an idea generator. You curate from what it surfaces, combine suggestions, and use your own judgment about what’s worth developing. The creative act shifts from generation to selection. That’s a skill most senior marketers already have.
2. AI is personal — and usage time is a moat
Leor made a point that sounds obvious but has real strategic implications: ChatGPT trained on a year of his conversations will give him fundamentally different answers than it gives someone who opened an account yesterday.
The memory that accumulates — the context, the preferences, the patterns of what he finds interesting — functions like institutional knowledge. You can’t buy it. You can’t shortcut it. You build it through sustained use over time.
This has direct implications for marketing teams debating AI adoption. The companies that started using AI tools seriously in 2023 are now operating with a meaningful advantage over those who waited. Every month of delay is a month of compounding context that competitors are building and you aren’t.
It also explains why individual power users outperform teams using shared accounts or generic prompts. The model doesn’t know your voice, your preferences, or your context. You have to train it, repeatedly, over time. The teams that assign AI tools to individuals and let those individuals develop deep, persistent working relationships with them will get better outputs than teams running everything through a shared login.
The practical advice: stop sharing accounts. Stop starting fresh conversations for every project. Build persistent contexts. Let your people develop their own AI working styles. The differentiation compounds.
3. What the AI topic landscape tells marketers about what’s next
Leor runs a generalist AI newsletter with nearly 60K subscribers, which gives him unusual visibility into which topics actually resonate with a broad audience — not just within a specific niche.
His read on the current moment: AI agents are the dominant topic, and for good reason. But the more interesting signal is the counter-narrative emerging alongside it — content about human judgment, critical thinking, and human-in-the-loop approaches. People aren’t just asking “how do I use AI more?” They’re also asking “where should AI not be?”
The upcoming topics he flagged: AI environments (the infrastructure where agents operate), predictive AI (running scenario models with variable inputs), AI security (deeply underreported given the risks), and robotics as the prerequisite for AGI claims to mean anything in the physical world.
For marketers, the security angle is particularly underserved. Leor noted what Peter confirmed from the AI-Ready CMO audience: marketers are routinely feeding company data, campaign strategy, and proprietary information into public LLMs — likely in violation of every NDA they’ve signed. The topic barely registers in marketing-specific AI content. That’s a gap.
The ambient AI angle is also worth watching. As social feeds become saturated with AI-generated content, offline experiences become relatively more valuable. Ambient AI in physical environments — smart signage, adaptive retail, context-aware OOH — could become a significant play for brands willing to move early.
4. The distribution automation question: where humans stay in control
Leor considered automating his YouTube publishing — and decided against it. His reasoning is instructive.
YouTube is permanent in a way other platforms aren’t. Delete a video, and you lose the views, comments, and algorithmic history permanently. There’s no undo. For that reason, he keeps final publishing decisions human, even as he automates adjacent parts of his workflow with Opus Clips and Descript.
Peter shared the AI-Ready CMO approach: Jenny, their internal AI agent, handles cross-platform social distribution automatically — 2-3 posts per day across all channels — with the same results as manual posting. The key enabler was having 6+ months of daily content for the agent to learn from. Without that library, the quality wouldn’t hold.
The pattern that emerges: automate where you have enough context for the AI to replicate your voice reliably, and where errors are recoverable. Keep humans in the loop where the stakes of a mistake are permanent — account reputation, legal exposure, irreversible platform consequences.
This isn’t a framework that works for teams just starting out. It’s a reward for consistency. You have to build the library first. The automation becomes possible only after you’ve shown up long enough for the AI to know what you sound like.
5. Why Substack feels different — and what it means for B2B marketing
Both Leor and Peter arrived at the same conclusion independently: Substack readers are different from social media audiences in a way that matters for how you think about engagement.
Leor’s engagement numbers look invisible on the platform — few comments, minimal public likes. But 700,000–800,000 monthly views come through email, from subscribers who open every day because they’ve opted into a daily relationship with his content. When he links his YouTube channel in a relevant article, he gets 1% conversion in 24 hours. On Instagram, that kind of audience would generate more visible engagement but less controllable reach.
The practical difference: Substack subscribers who open daily emails have made a repeated, deliberate choice. That’s a fundamentally different relationship than someone who saw a post in a feed and clicked. The conversion economics are different. The trust level is different. The tolerance for depth is different.
For B2B marketers evaluating owned media strategies, this is the case for newsletter-first thinking. Not because Substack specifically is the answer, but because building a subscriber relationship — where people opt in repeatedly to receive your content — creates a different quality of audience than social reach ever does. You can reach more people on LinkedIn. You can convert more of them from a list they chose.
Leor grew to 60K+ through Substack alone, no outside marketing, in one year. That’s not a template — it’s a proof of concept that consistent, specific, personality-driven content still compounds organically, even in a saturated market.
Found value in this conversation?
Share it with a marketing leader who needs to hear this.
Subscribe to AI-Ready CMO to catch future live episodes and get daily AI marketing intelligence that actually matters.
