AI summary by TwinMind
It’s actually over now
Theo - t3․gg
52m 13s
Initial Arc Browser Recommendation and Current Disappointment
- Creator previously switched from Chrome to Arc browser, made dedicated video promoting it
- Brought "unbelievably large number of users" to Arc through recommendation
- Now feels burned by Browser Company's handling of Arc abandonment
- Josh (CEO) cancelled promised call twice, poor communication throughout pivot
Browser Company's Letter to Arc Members Analysis
- Published article "Letter to Arc Members" explaining pivot to DIA browser
- Three things they would do differently if starting over:
- Stop working on Arc a year earlier
- Embrace AI fully sooner and unapologetically
- Communicate very differently (acknowledging communication failures)
- Confirms they have fully stopped working on Arc development
- Arc eating MacBook battery in 2 hours instead of 10 hours
- Windows Arc development completely abandoned
Arc's Usage Statistics and Feature Adoption Problems
- Only 10% of users who tried Arc stuck with it (retention problem)
- Multiple spaces used by only 5-12% of users
- Live folders used by 4.5% of users
- Calendar preview used by only 0.4% of users
- Strong retention if users stayed 3+ days, but near-zero return rate if they quit after first day
- "Mom problem" - Josh couldn't see his mom using Arc, indicating mass market inaccessibility
Technical Architecture and Open Source Decision
- Arc built on custom ADK (Arc Development Kit) - internal SDK for browser UI
- ADK allows rapid prototyping without touching C++ parts
- ADK is foundation for DIA, making open sourcing Arc impossible
- Originally built with TCA and Swift UI, moving away due to performance issues
- Download tab freezes browser with 100+ files due to rendering limitations
- Company considers ADK their "secret sauce" and core value
DIA Browser Strategy and AI Focus
- DIA positioned as Chrome-like browser with AI integration
- 40% of DIA users use chat-with-tabs feature vs 5% Arc users using spaces
- Goal: make browser "mom-friendly" with less learning curve
- Three key principles for DIA:
- Simplicity over novelty (terrible saxophone vs piano analogy)
- Speed as foundation (DS architecture faster than Arc)
- Enhanced security (team grew from 1 to 5 security engineers)
Company Scale and Financial Burn Rate
- 136 employees based in NYC
- Estimated $26M annual salaries ($150K average)
- Total operational costs ~$30M annually
- $1.5M rent, $300K software subscriptions, $400K infrastructure
- Raised publicly known $128M since 2019, likely more undisclosed
- Zero revenue despite massive burn rate
- Comparison: Zen browser built by 3 people provides better experience
Browser Alternatives and Recommendations
- Zen: Firefox-based, best UX experience but Firefox limitations
- Vivaldi: Chromium-based, customizable but janky
- Helium: Very early Chromium fork by Cobalt creators, promising but not ready
- Chrome: Still recommended for most users despite Google data concerns
- Ungoogled Chrome: Good alternative for privacy-conscious users
- Creator contributed money to both Zen and Helium development
Venture Capital Impact and CEO Cosplaying Problem
- VCs likely didn't recommend killing Arc - would have advised evolution instead
- "Cosplaying as CEO" behavior: hiring before users, elaborate marketing videos, fancy conferences
- Accountability to investors drives vanity metrics focus over real progress
- Smart founders can rationalize any decision, even wrong ones
- Pattern of looking at percentage improvements in meaningless metrics vs actual revenue
Creator's Personal Experience and Lessons
- Went through similar struggles with Ping (video collaboration tool)
- Had to lay off team when product couldn't cover even one salary
- "Mom problem" comparison: mom trying to support Ping vs actually using T3 Chat
- Learned difference between vanity metrics and revenue-focused updates
- Now focuses on "we doubled our revenue" vs "increased retention 16.4% to 18.2%"
Technical Performance Issues
- DIA significantly slower than Chrome's built-in AI features
- Demo showed DIA taking much longer than Chrome's instant AI summaries
- Browser crashed during demonstration
- Animation performance issues throughout interface
