Scalar retro and post-mortem | Rahul Tarak

• Spent a year building WASM client-side rendering for unit economics when users didn't care—classic nerd-sniping that cost them the company
• "Deep software" products need proper technical abstractions from day one, not MVP'd infrastructure—they MVP'd their multiplayer engine and lost 18 months rebuilding it
• Being "wishy washy" about company type killed them: knew the product vision but took advice meant for different types of companies, leading to catastrophic technical debt
• Remote work was fatal for this type of product—needed in-person collaboration in SF but were too late to realize it
• Trying to "hack GTM" before product readiness (Anno.so side project, film festivals) just split focus when they needed 2+ more years of building

The author spent 3 years building Scalar, a professional collaborative video editor ("Figma for video editing"), and failed to reach PMF. The core thesis was sound—validated with hundreds of editors who confirmed the pain of needing to edit in-person—but execution mistakes compounded into failure.

The biggest technical mistake was getting nerd-sniped by WASM. They spent a year building client-side video rendering for better unit economics and free-tier experience, but it was neither performant nor something users cared about. More fundamentally, they treated their product like a typical SaaS when it was actually "deep software" requiring proper technical abstractions from the start. They MVP'd their multiplayer engine following standard startup advice, which cost them 18 months when they had to rebuild it properly. The author argues that for complex technical products, you can't MVP core infrastructure—you need to nail abstractions that compound like a deep tech company would.

The strategic failure was being "wishy washy" about what type of company they were building. They had conviction about the problem but not their approach to company building, so they took advice meant for different types of companies. This led to poor decisions around capital (too cheap, not aggressive), team structure (remote when they needed in-person), and GTM (trying to hack it with side projects like Anno.so and film festivals before the core product was ready). They also underestimated how much capital they needed and didn't use what they had aggressively enough. When GPT-4 hit and the video editing space became hyper-competitive, they couldn't raise a larger seed and entered a fundraising death spiral.

The lesson: if you're building deep software, call your shot early. Commit to the longer timeline, proper technical foundations, heavier capital needs, and in-person collaboration. Standard "move fast and break things" advice will cost you years.