How Roy Lee built Cluely to $6M ARR in two months at age 21

Executive overview

Roy Lee built Cluely — an AI assistant that sees your screen and hears your audio, giving invisible real-time help — and reached $6M ARR within two months of launch. The company's rapid growth is driven by a deliberate "distribution-first" thesis: capturing attention at scale through viral mindshare, not product features alone.

The core bet is that when better AI models arrive, Cluely's brand equity and distribution infrastructure will make it the default "Jarvis" — the same way Gmail beat Yahoo, not through superior tech but superior capture of attention.

Distribution advantage compounds faster than product advantage when building at the application layer.

From Interview Coder to Cluely

• Got Harvard acceptance rescinded after a senior-year suspension; spent a year isolated, coding, and building a chip on his shoulder.
• At Columbia, pitched co-founders door to door; Neil was the only yes out of ~50.
• Built Interview Coder — a screen-share-invisible tool that solves coding interview questions via AI — as a demo in October.
• Used the tool to get a real Amazon offer, posted the video; Amazon flagged it to Columbia, Columbia suspended him.
• Tweeted every update in real time; the controversy attracted enough investor attention to raise $5M in under 24 hours.
• Raised while "hot" — VCs reached out; the round closed before a consistent pitch existed.
• After months of false starts (including a dating app), a cabin retreat with co-founder Neil produced the insight: Interview Coder for everything = Cluely.

The distribution-first thesis

• "If you can get a billion views a month on anything, it does not matter if you're selling perfume or software — people will buy."
• Short-form consumption has 100x'd; content creation has not — a gap that makes creators massively undervalued right now.
• Cluely's average marketing team age is 21.5 — deliberately; understanding what goes viral requires living inside the algorithm.
• Marketing team built by recruiting creators directly from their own For You pages — people already hot in the target demographic.
• ~60 in-house content creators; pay $25–$40 flat per video plus performance bonuses ($500 at 500K views, $1K at 1M views).
• Over 1 million installs from short form in the first two months; on track for 50M views in one month.

Mindshare vs. brand awareness

• Views alone are not the goal — "talked-aboutness" is. Subway Surfers gets billions of views but nobody discusses it at lunch.
• A launch video with 12M views that makes people say "have you seen this?" to friends is more valuable than 100M passive impressions.
• The target emotional benchmark: would 50% of people who saw this have a strong negative reaction? If not, it's not controversial enough.
• X rewards controversy that Instagram ignores; the dancer video went front-page on tech Twitter because the platform hasn't caught up to what Instagram considers normal.
• The "cheat on everything" framing is intentional: a word with a negative connotation stops the scroll and embeds the brand name.

Launch video strategy

• Modelled directly on the friend.com launch by Avi Schiffman: cinematic quality + a manufactured controversial scenario.
• The Cluely launch showed glasses that don't exist yet — a plausible end-state that let viewers grasp the concept instantly.
• Cinematic production alone does not drive virality; the missing ingredient is "meat" — something controversial to talk about.
• Plan: in-house film studio, cinema-quality launch video every two to three weeks for each new feature.
• The 50-interns video worked the same way: frame a standard contractor hiring as a chaotic startup stunt; exclude the mundane explanation.

Platform strategy

• X/Twitter: controversial scenarios, founder personal account as company voice extension, ~1–2 posts per day.
• Instagram: UGC performance engine — educational, step-by-step content; the platform that actually converts to downloads.
• TikTok: not yet cracked to the same level as Instagram; audience attention span is too degraded for educational content.
• LinkedIn: ghostwriter maintains a separate educational voice for that audience.
• YouTube: not yet prioritised.

Product and competitive position

• Cluely designed so that when the next generation model ships, the product quality multiplies — not just increments.
• Research labs are too focused on the AGI race to compete seriously at the application layer; the expected exit is acquisition (e.g., OpenAI acquiring Windsurf/Kodium at 100x ARR).
• The day a Jarvis-capable model API goes public, whoever owns distribution wins — the same reason Google Search beat Bing.
• Team: 3–5 engineers (targeting 6–7 to reach billion-dollar scale); headcount target is under 50 indefinitely.

Fundraising and risk philosophy

• Raise the maximum possible when you don't need it; anchoring to a conservative valuation out of fear is a mistake.
• Companies never fail because the founder was too controversial — they fail because they didn't make enough money.
• When you take risks so outsized that almost no one else is willing to take them, competition disappears.
• "The only way out is through" — controversial brand equity compounds only if you keep going.
• Pre-series B: reputation is the last thing that matters; growth is the only thing that matters.