luca ferrari

Unlike PE's shallow interventions and 3-7 year exit horizons, Bending Spoons' permanent ownership enables three structural advantages: (1) fluid movement of R&D/marketing resources across portfolio companies to capture fleeting opportunities, (2) employer brand and talent density that standalone companies cannot match, (3) shared proprietary technology (AI recruiting, infrastructure) amortized across businesses. These compound over decades, creating high barriers to entry.

AI acts as a force multiplier for companies that build custom integrations, proprietary tooling, and cultural adoption — widening the gap between elite operators and the majority. Bending Spoons is already at the cutting edge of using AI across engineering, design, product, and growth functions, and expects this advantage to compound for years.

Bending Spoons treats jobs as their most important product, attracting elite talent through high talent density, unique learning opportunities (27-year-old GMs running $50-100M businesses), and a culture of intellectual honesty. They use AI to predict performance from 800K applications, hire only fixed-salary employees with optional equity investment, and maintain alignment through cultural rigor rather than KPI-driven incentives.

Bending Spoons prefers permanent capital (evergreen structures) over 10-year VC/PE funds because it eliminates forced liquidation risk. They use debt markets (3.5x EBITDA) efficiently — lenders are surprisingly visionary but rigorously downside-focused — and raise equity mainly for employee liquidity, minimizing dilution (~10% total).

The majority of software R&D investment yields negative returns. Bending Spoons achieves higher output with lower cost by eliminating tangential projects, focusing on what customers 'painfully need,' and measuring impact rigorously — as demonstrated by Evernote's lower cost base but superior performance and feature set.

Bending Spoons uses fine-tuned open-source models for specific tasks (cheap, self-hosted) with frontier models only for supervision/complex tasks. Custom orchestrator routes tasks optimally. Result: 90% of code AI-written with modest spend. Challenges 'token maxing' narrative at big tech.