In partnership with
Happy Tuesday 👋 The AI arms race is heating up—OpenAI’s GPT-5 is here, and it’s not just faster, it’s smarter. But that’s just the start. This week, we’re digging into a SaaS opportunity hiding in the open, a repeatable growth engine for distribution, and a founder who scaled clones into a real business. Plus: the underused RAG trick that’ll make your AI product way better.
Let’s get into it.
In this issue:
OpenAI’s GPT-5 is a leap towards AGI 🤖
A SaaS goldmine hidden in plain sight 👀
Copy this playbook to get distribution at scale 📈
This founder cloned 3 apps to $35K/month 🌀
The RAG metadata trick no one talks about 💿
A two-way voice translator that turns any phone into a live interpreter
Talk To Locals lets you have natural, real-time conversations in 40+ languages. No typing. No passing your phone back and forth. Just place your phone between you and anyone, and both voices are translated instantly.
Perfect for ordering in local restaurants, meeting your partner's family, settling in as an expat/nomad, or discovering stories only locals know.
First 10 minutes free, then pay only for what you use, no subscriptions.
OpenAI’s GPT-5 drops in August—and it’s a shot at AGI
TL;DR: GPT-5 is set to launch as early as August 2025, and OpenAI is framing it as much more than just an upgrade—it’s being positioned as a leap toward Artificial General Intelligence. According to The Verge, internal testing is nearly complete. The new model is expected to combine domain-spanning reasoning, language generation, and math capabilities in one unified system—something previous versions, including GPT-4.1, couldn’t fully achieve.
🔍 Not just smarter—more capable – GPT-5 isn’t just about better chat responses. Early testers are reporting stronger reasoning, multi-step logic, and domain-transfer performance that suggests broader intelligence.
🧠 Reasoning is finally integrated – Unlike past models that separated general-purpose language use from advanced reasoning (like o3 or Claude 3 Opus), GPT-5 is designed to handle both natively.
⚙️ Compute bottlenecks nearly broke the timeline – Massive training costs and technical complexity pushed release plans back months, but OpenAI now seems confident enough to greenlight an early August launch.
🏁 AGI ambitions are no longer subtle – Sam Altman’s latest statements confirm that OpenAI sees GPT-5 as a critical milestone toward AGI, citing “gold medal” level math performance and generalized reasoning as key breakthroughs.
🌐 Ripple effects across industries – Expect new use cases in enterprise automation, research, education, and healthcare. But also brace for a fresh wave of policy debates, ethics discussions, and scrutiny on how aligned these models truly are.
TL;DR: Simon Hornberg’s video breaks down how founders can turn complex AI automation workflows—often shared freely on YouTube or LinkedIn—into simple, user-facing SaaS products. Using tools like N8N and Lovable, he outlines three paths: slapping a clean UI on existing workflows, bundling simplified apps into a suite, or building a full-fledged SaaS product from scratch. The common thread? Take something powerful but unusable for most people, and make it dead simple.
🧠 AI workflows are great; just not usable – Platforms like n8n let power users create mind-blowing automation, but 99% of people are locked out because the UX is overwhelming.
🧰 Wrap it in a form – Take a workflow, build a no-code UI with tools like Lovable, and charge for AI usage or access. Easy entry point, fast to build.
🧳 Bundle workflows into a suite – Turn multiple simplified workflows into a single product with unified login and billing. Think Notion for AI-powered mini-apps.
🎯 YouTube is your R&D lab – The best workflows (and user feedback) are already public. Study the comments, see what confuses people, then solve that with a clean UI.
💸 Subscription + usage fees = recurring revenue – Charge per AI credit, per workflow, or via flat subscription. Tied to actual usage, so it scales with you.
The AI tools are already out there—what’s missing is distribution. Whoever nails UX, builds for the non-technical crowd, and wraps the tech in a pain-killing solution will win. Simon’s playbook is less about invention, more about packaging and positioning.
You don’t need a big idea—just clip, clone, and ship
TL;DR: Cluey, a viral AI tool built by two students, scaled to a $120M valuation in 3 months using a clipping farm of 700+ creators distributing short-form content across thousands of accounts. The strategy: clip viral moments, flood the feed, and pay per view. This isn’t about building from scratch—it’s about cloning what works and wrapping it in distribution. Meanwhile, AI is nuking SaaS pricing, enabling lean startups to clone legacy tools like Appfolio for pennies. But human creativity and storytelling still give you the edge.
🎥 Cluey’s growth hack: the clipping farm – They hired 700+ clippers to create short-form videos from long-form content and posted them at scale. It’s algorithmic virality.
📈 Pay creators per view, not per video – Incentivizes performance and filters out fluff. Smart comp structure for organic reach.
🤖 AI is commoditizing SaaS – An AI-built Appfolio clone charges $23/month vs. $2,300. Expect more disruption of overpriced incumbents.
🧠 Creativity is still your moat – AI struggles with emotional resonance, storytelling, and brand nuance. Don’t outsource your voice.
🛠️ Services > SaaS (at first) – Use public data like new biz listings to launch lead-gen driven agencies or simple AI-powered services. Fast path to revenue.
You don’t need to invent something new—you just need a strong distribution flywheel. Viral clipping, performance incentives, and AI tooling create asymmetric leverage. And if you want recurring revenue? Start with services, then stack software.
The best marketing ideas come from marketers who live it.
That’s what this newsletter delivers.
The Marketing Millennials is a look inside what’s working right now for other marketers. No theory. No fluff. Just real insights and ideas you can actually use—from marketers who’ve been there, done that, and are sharing the playbook.
Every newsletter is written by Daniel Murray, a marketer obsessed with what goes into great marketing. Expect fresh takes, hot topics, and the kind of stuff you’ll want to steal for your next campaign.
Because marketing shouldn’t feel like guesswork. And you shouldn’t have to dig for the good stuff.
How a former optician cloned 3 apps to $35K/month
TL;DR: Samuel Rondo quit opticianry, taught himself to code, and now runs three SaaS apps pulling in $35K/month—all by cloning successful apps and improving them by just 1%. His formula is clear: validate via revenue screenshots, use AI tools like ChatGPT to speed up coding, launch a fast MVP, and stack growth with paid ads, SEO, and affiliate traffic. His stack is lean (Next.js, Node, Vercel), his process is tight, and his products—Us.com, Storyshow.ai, and Capacity.so—prove that simple, maintainable software still wins.
🚀 $2M from a single viral video – Ava reverse-engineers virality by studying top Reels and adapting winning hooks with her clients’ expertise.
🎥 Talking-head videos dominate – Clean, value-packed videos with clear verbal hooks outperform over-edited, visually complex content.
🔍 Data-driven hooks – She tracks outlier content (5x+ average views) to extract trends in opening visuals, captions, and scripts.
⚙️ SOPs = scale – A 100-person team, strict quality controls, and personal script review allow her to handle 350+ clients without diluting quality.
📈 Instagram > TikTok for business – She leverages Instagram’s Reels tab and Explore page for reliable lead generation and client growth.
Most people try to outsmart the market with novelty. Samuel just looks for what’s already working and gets it live faster. Execution velocity + proven ideas = unfair advantage.
The smart RAG hack: Add Metadata or stay mid
TL;DR: A smarter Retrieval-Augmented Generation (RAG) stack starts with one overlooked feature—metadata. This beginner-friendly walkthrough shows how attaching structured metadata (like video titles, timestamps, and URLs) to vector chunks dramatically improves precision, transparency, and control. Using a pipeline built on Apify, Claude, and Supabase, the creator scrapes YouTube transcripts, chunks them with preserved timestamps, stores them with metadata, and builds a filterable, trustworthy vector database. It’s a masterclass in building RAG systems that don’t just return relevant results—but cite their sources too.
🔗 Metadata creates trust and traceability – Chunk-level context like URLs and timestamps makes it easy to verify AI responses and builds user confidence.
🧠 Metadata ≠ embeddings – Metadata doesn’t influence vector similarity—it enhances how results are filtered, organized, and interpreted.
🎯 Chunking with timestamps = pinpoint precision – Users can jump to exact moments in videos instead of reading entire transcripts. Massive UX upgrade.
🔍 Filter by video title, URL, or tags – Target answers from specific videos or groups of content, crucial for multi-source databases.
🔄 Automate data cleanup with metadata triggers – Mark a video “remove” in a sheet and auto-delete all related chunks. Clean RAG, clean results.
🧰 Tools: Claude for code, Apify for scraping, Supabase for storage – Together, they create a replicable, end-to-end pipeline.
The real unlock in RAG isn’t just better models—it’s better data architecture. Metadata adds a thin but powerful layer that turns messy vector blobs into something searchable, interpretable, and useful. If your AI app doesn’t cite sources, it’s not ready for primetime.
Want to sponsor this newsletter? This newsletter goes out to thousands of creators, builders, and early-stage founders every week. If you’d like to get your product or brand in front of them, you can book a sponsorship slot here 👉 Become a Sponsor
