AI unlocks consumer startup opportunities that were previously impossible

Executive overview

Consumer tech has always been gated by the tools available to create — cameras democratised photos, microphones democratised podcasts. AI is now doing the same for music, education, and dozens of "solved" product categories that were written off.

Timing and distribution remain the hardest problems in consumer. AI raises retention ceilings and enables new paid models, but founders still have to build distribution themselves.

The best consumer bets now come from revisiting overlooked categories and untouched data sets — not chasing net-new ideas.

The Anchor origin: democratising audio creation

• Started as a social audio platform in 2014–2015, inspired by Serial and the podcasting wave
• Realised creation was the bottleneck — applied mobile photo-editing expertise to audio
• Users wanted tools, not a new listening platform; kept asking for distribution to Spotify and Apple Podcasts
• Neither had APIs, so the team manually created RSS feeds and submitted them on users' behalf
• That unscalable hack unlocked growth and shaped the eventual Spotify acquisition
• Survival came from a single rule: hit 15% week-over-week growth every week for three months
• The constraint forced a pivot away from what the founders wanted toward what users actually needed

Why consumer went cold — and why it's back

• The 2008–2014 consumer boom rode platform openness; as APIs closed, distribution dried up
• B2B/SaaS had a legible playbook; consumer has always been "lightning in a bottle"
• The hardest thing in consumer: identifying trend, team, and timing simultaneously
• AI raises retention, making paid consumer models viable where subscriptions previously struggled
• Distribution is still unsolved — AI hasn't changed that yet, though new channels will emerge
• Consumer growth expertise atrophied during the B2B decade; talent is now concentrated in Eastern Europe

Suno and the music creation thesis

• No technology ever democratised music creation the way cameras democratised photos or microphones democratised podcasts
• AI is the first tool that makes music creation accessible to anyone
• Early behaviour felt like novelty, but creators are now making music they intend to keep and listen to themselves — a behaviour with no prior analogue
• The consumer opportunity mirrors Anchor: make the hardest creative act trivially easy, then solve distribution

Three phases of social media

• Phase 1 — social graphs: content distributed based on who you follow
• Phase 2 — recommendation media: TikTok-style algorithmic programming against interests
• Phase 3 — generative media: content created dynamically and in real time for each viewer; Sora is the early signal
• Phase 3 removes the need for human creators; content is generated on demand per user
• Cameo-style likeness licensing may become the new creator economy — brands and individuals get compensated when their identity is invoked in generated content
• The end state may be a fully AI-programmed feed with no human-authored content

Distribution playbook for consumer founders today

• Creator leverage (TikTok, Reels, X) is now table stakes — all top consumer pitches show charts built on it
• Micro-influencers (1k–10k followers) are still a mispriced asset; larger creators are not
• "Organic" is a misnomer — most viral growth is non-paid, not truly organic
• Retention and funnel quality still determine whether influencer-driven top-of-funnel becomes a business
• Start practicing distribution before launch: test anonymously, develop taste, lower the stakes
• The first 10–15 seconds of product description — in person or on video — determine whether anyone tries it

What to build: overlooked categories and untouched data

• Revisit categories previously written off: email clients, browsers, group software — AI changes the calculus
• Look for large data sets that are either publicly accessible or personal (health records, camera rolls, location history) and apply LLMs or media models to them
• Health data layered onto AI is a large near-term opportunity; companies like Nori (Apple Health + LLMs) and Doctronic (medical triage) are early examples
• Geo-aware, memory-rich personal agents — like Dennis Crowley's AirPod geo-notification project — illustrate what's possible with contextual personal data
• The question to ask: what surface was heavily invested in during the early Internet era and then abandoned as "done" that AI can now rebuild from scratch?

Oboe: personalised AI education

• Premise: apply AI to human intelligence, not just productivity
• Anything you want to learn, Oboe generates a course — podcast, lecture, study materials — in whatever format you prefer
• Over time it learns how you learn: format preferences, prior knowledge, gaps
• Each lesson becomes more personalised and efficient; avoids the one-size-fits-all problem of all current education
• Follows the same Anchor pattern: large ambitious surface area, clear North Star, iterate hard on the path

Advice for consumer founders

• Bet on great product builders and trust that they'll find the opportunity AI creates
• Overcapitalisation kills urgency — small teams that can fit in a room move faster and pivot more honestly
• Point at an ambitious space, start with a hypothesis, get punched in the face, and alter the path — not the North Star
• Taste matters more as building gets easier; craft and speed are now the primary moats
• Labs (OpenAI, Anthropic, Meta) have taste, execution, and resources — assume they will enter any category that looks obvious