How I AIHow a non-technical founder built a $100K ARR meme company | Jason Levin (Memelord CEO)
CHAPTERS
- 0:00 – 4:36
Memelord’s thesis: entertainment, memes, and the rise of agentic users
Jason and Claire frame Memelord around a core belief: the most entertaining brands win, and memes are the most information-dense unit of culture. They connect that thesis to a near-future where AI agents, not humans, become primary “users” of tools that generate and distribute content.
- •Memes as cultural transmission and “memetic warfare” as a marketing lever
- •Why attention and trend-speed matter more than polished campaigns
- •Agents reduce human hesitation and ‘cringe friction’ in marketing output
- •Memelord’s evolution from meme alerts into an API-first product direction
- 4:36 – 6:55
Demo: OpenClaw + Memelord for agentic meme generation (trend-aware captions)
Jason shows how Memelord’s API plugs into OpenClaw to generate memes based on current trends and relevant templates. The system pulls from a trending meme database, selects contextually appropriate imagery, and iterates quickly with caption variations.
- •Prompting an agent to ‘cook’ memes on a topic using Memelord’s database
- •Context matching (e.g., political meme templates for political prompts)
- •Rapid iteration via “switch the caption” to explore alternative jokes
- •Why narrow timing windows make automation especially valuable
- 6:55 – 8:27
“No UX is the best UX”: designing for an agent-first future
They discuss the tension between investing in beautiful onboarding/UX and the inevitability of users preferring API access and automation. Jason shares a candid investor anecdote illustrating the shift from button-clicking to agent workflows.
- •Ramp CTO quote: “No UX is the best UX” as a product north star
- •High-craft onboarding still matters for humans—until they grab the API key
- •Investor story: ‘I don’t want to use anybody’s software’ → ship an API
- •Balancing human brand experience with agent-first distribution
- 8:27 – 13:32
From $6.90 meme newsletter to product: the scrappy origin story
Jason explains Memelord’s earliest form: a paid newsletter delivering the newest memes and linking to Google Slides decks. The chapter highlights starting with obsession and distribution rather than waiting for perfect tools or engineering help.
- •MVP as a newsletter + Google Slides when you can’t code
- •Meme alerts/trend surfacing as the initial core value
- •Remixing trends for brands as the consistent through-line
- •Bias toward shipping before the tooling wave fully arrives
- 13:32 – 14:37
Scaling to $100K ARR on Bubble: 395 workflows and a no-code grind
Jason details building the business on Bubble before ‘vibe coding’ tools matured. He describes the operational reality of a complex no-code app, rate-limiting surprises, and why he eventually raised and hired engineers.
- •Building a real subscription business without engineers initially
- •The fragility/complexity of scaling: “395 workflows” in Bubble
- •Early growth pain: rate limits and learning infra concepts late
- •Raising capital and transitioning to an engineering team
- 14:37 – 15:07
Cursor for non-technical founders—and the “every marketer vibe codes” rule
Jason shares how Cursor became the default environment for both him and his marketing team, with engineers handling security-critical work. Claire emphasizes why Cursor’s visibility (reading code, guided modes) helps non-technical builders become effective and autonomous.
- •Cursor workflows: Ask/Agent modes for learning and shipping
- •Division of labor: engineers on security/auth; founders/marketing on experiments
- •Why seeing and editing code accelerates non-technical capability
- •Culture policy: every marketer must build, not just request
- 15:07 – 19:58
Free tools as lead magnets: shipping weird mini-products for massive email capture
Jason demos Memelord’s free tools page and explains how tiny meme utilities became major acquisition channels, generating hundreds of thousands of emails. They argue interactive tools increasingly outperform PDFs because they’re faster to build and more compelling to use.
- •Free tools (filters/generators) as modern replacements for downloadable PDFs
- •Examples: “bust down” filter, GigaChad maker, Steve Jobs portrait generator
- •Viral distribution loops (e.g., unexpected geographic virality)
- •Why marketer-built tools avoid handoff loss and drive subscriptions
- 19:58 – 24:40
Let marketers cook: organizational leverage, token budgets, and “priority is yes”
They make the case that talented marketers will leave if they can’t execute directly and experiment rapidly. AI shifts product velocity enough that traditional prioritization and heavy handoffs become liabilities rather than safeguards.
- •Handoff loss: creativity degrades through layers of translation
- •Token spending as an enabler; reluctance becomes a competitive disadvantage
- •Cultural point: abundance mindset—ship more, faster, weirder
- •Retention: constrain builders and they’ll quit to build their own thing
- 24:40 – 28:21
The vibe-coding inflection point: commit graph and the modern tool stack
Jason highlights a GitHub commit graph that flips from sparse to “dark green,” reflecting his increased direct contribution through AI tooling. He also shares the broader stack—Claude/Gemini, Linear, and PostHog—and why agent-friendly tasking/analytics are becoming core infrastructure.
- •Commit graph as evidence of AI-assisted creator productivity shift
- •Tooling: Claude, ChatGPT, Gemini for ideation and execution support
- •Linear as an agent-native task substrate; minimal UI, strong APIs
- •PostHog AI for fast, natural-language analytics and dashboards
- 28:21 – 30:11
Build weird stuff IRL too: barbell strategy of AI + physical-world stunts
Jason argues that as AI accelerates digital execution, real-world creativity becomes an even stronger differentiator. He shares examples like physical merch, novelty artifacts, and experiential events as part of Memelord’s brand-building approach.
- •Barbell approach: ship AI experiments while doing tangible, real-world ideas
- •Examples: Memelord CDs as a playful API ‘distribution’ gimmick
- •IRL stunt: renting a theater to watch Instagram Reels
- •Old-school human tactics (coffee chats, events) regain leverage
- 30:11 – 33:57
Hyper-personalized software and hardware hacking: the “keyboard with no screen”
Jason walks through a DIY hardware project: a Raspberry Pi-backed keyboard that sends ideas to Zapier and routes them into email or Linear without waking his wife or using a phone in bed. The broader theme is building niche tools for yourself rather than forcing everything into a scalable product.
- •Problem framing: capture ideas at night without phone/light/voice assistants
- •Solution: Pi-powered keyboard triggers API calls on Enter
- •Routing logic via keywords (e.g., create Linear tickets vs email notes)
- •Personal-first building: useful even if it never becomes a business
- 33:57 – 39:57
More personal automations: finding lost items, hardware takes, and optimistic execution
They riff on future hacks like using in-home cameras + AI to locate misplaced items, and discuss why seemingly ‘dumb’ hardware ideas may just be early. The conversation shifts toward an execution-first optimism enabled by rapid prototyping with AI.
- •Idea: AI-assisted ‘where did I put my keys/phone’ using cameras
- •Rabbit/Worldcoin as examples of ideas that may be early vs wrong
- •Claire’s example: building internal tools (e.g., doxx-detection for podcasts)
- •A mindset shift: assume ideas can work if executed well
- 39:57 – 43:48
OpenClaw calendar agent: weekly reviews, meeting hygiene, and content mining
Jason shares a practical agent workflow that reviews his calendar weekly, flags time sinks, and suggests schedule changes. He extends it into a content system: turn meetings and interactions into draft posts, capturing high-signal insights that would otherwise disappear into DMs and calls.
- •Weekly ‘week in review’ and ‘week ahead’ summaries from calendar data
- •Agent recommendations: fewer standups, more deep work blocks
- •Automations: identify meetings that could be emails and propose cancellations
- •Content pipeline: generate posts from real-world meetings and conversations
- 43:48 – 48:18
Can AI be funny? Funniest models, why memes aren’t slop, and what he won’t automate
Jason answers whether AI can be funny, arguing it’s approaching human levels but still lacks lived context and taste at the top end. He shares Memelord’s model mix (favoring less ‘safe’ outputs), defends memes as hyper-contextual communication, and notes he avoids using AI for his own writing/joke craft.
- •AI humor is improving; humans still win at the extreme high end (for now)
- •Model take: Grok and Gemini outperform ‘safer’ models for humor
- •Memes as context-rich, information-dense—not generic ‘slop’
- •Personal boundary: Jason doesn’t use AI to write his own content/jokes
- 48:18 – 51:53
Prompting for unhinged output + closing: ship, cook, and where to find Jason
Jason describes how he prompts models when they’re not funny—pushing for edgier vernacular and less formality—while Claire lightly warns about being too mean to future robots. They wrap with the episode’s mantra (ship, abundance mindset, have fun) and Jason’s socials and Memelord onboarding.
- •Technique: push tone/vernacular; allow cursing; reduce ‘first-date’ stiffness
- •Memelord bakes ‘unhinged’ humor defaults into the product experience
- •Takeaways: abundance mindset, rapid shipping, empower builders
- •Where to find: memelord.com + @iamjasonlevin across platforms
