Don't start a company — solve a problem worth solving

Executive overview

Starting a company for its own sake is a trap. Russ d'Sa, CEO of LiveKit, made this mistake repeatedly before finding a problem genuinely worth solving.

The shift: stop optimising for "founder" status and start optimising for problem-market fit. LiveKit — infrastructure powering ChatGPT's voice mode and 25% of 911 calls — emerged from recognising that real-time audio/video had no general-purpose open-source stack.

The real goal is to solve a hard problem; the company is just the vehicle.

Lessons from early failures

• MeetU (YC 2007): got distracted by advice from other founders, lost focus on the core product vision
• Stock brokerage idea: Paul Graham called it "something you work on when you don't have a good idea"
• Mobile ad streaming: built cool technology with no actual problem behind it
• Pattern across failures: starting from a solution or from ambition, not from a problem

What working at a company teaches you

• Twitter was understaffed and fast-growing — created space to own large surface areas quickly
• 12–14 hours a day felt like play, not work, because impact was immediately visible
• Fires are a good sign: crashes and scale failures mean you've built something people want
• Engineers fall in love with code; reframe it — code is a means to an end, not the end
• Embedded media on Twitter: built on personal time, became one of the platform's defining features
• 23andMe lesson: picking the right market isn't enough — you also need a path to traction while waiting for adoption

The LiveKit origin

• Co-founded with the same David from YC 2007 — relationship built over a decade of parallel careers
• Problem: the internet was built for text (HTTP), not real-time audio/video
• Timing trigger: pandemic forced everything onto the internet; cameras and microphones suddenly essential
• Framing: "What if we build the Postgres for real-time networking?"
• OpenAI partnership revealed the bigger vision: LiveKit is the nervous system connecting humans to AI brains
• Every keyboard and mouse will be replaced by a camera and microphone — that's the addressable market

Advice for founders entering AI or robotics

• Both spaces are guaranteed to be multi-trillion dollar markets — but they're early and moving fast
• Counter-intuitive edge: joining a frontier AI company first can be more valuable than starting immediately
• Learn the foundational elements, see how fast things change, meet future co-founders
• Solo founding is tunnel vision — you miss the adjacent problems other builders are hitting
• Don't force inspiration; join something exciting, dig in, and let the idea surface
• When it comes, you'll know — then go all in