Saronic's CTO advocates a counter-intuitive approach: slow down the software lifecycle to be more methodical about infrastructure choices upfront. Simultaneously, speed up the hardware lifecycle by buying parts early and aggressively simplifying requirements to enable faster iteration.

The exponential improvement of AI models means product development must target future capabilities. Leaders should anticipate that problems taking months to solve will soon be trivial, and build roadmaps that assume a 6-12 month leap in technology.

Due to the rapid pace of AI-driven development, Ramp has abandoned annual or multi-year planning. They now operate on a three-month horizon, which is considered a long time because it allows them to accomplish what previously took three years, making long-term roadmaps obsolete.

Unlike pure software, building software for a physical product imposes immovable deadlines dictated by hardware manufacturing and shipping lead times. This forces software teams to abandon flexible, continuous iteration in favor of a highly-focused, delivery-oriented mindset to ensure the software is ready when the hardware is.

A holistic roadmap allocates capacity across four key areas: 1) optimizing the current product, 2) strategic innovation, 3) internal capability work (e.g., platform improvements), and 4) "run the business" support. This prevents a myopic focus on just customer features.

In a rapidly evolving field like AI, long-term planning is futile as "what you knew three months ago isn't true right now." Maintain agility by focusing on short-term, customer-driven milestones and avoid roadmaps that extend beyond a single quarter.

To avoid obsolescence costs from long component lead times, hardware companies can use a dual-track approach. One track focuses on small, quick iterations rolled into the current product 'block' based on immediate user feedback. The other track works on a long-term, ground-up redesign for major technological leaps.

A product leader should actively manage development by allocating effort into three buckets: future big bets, core foundation (stability/tech debt), and growth/optimization. The resource allocation isn't fixed; it must dynamically shift based on the product's maturity and immediate business goals.

The pace of AI-driven development has dramatically accelerated. High-performing companies are no longer planning quarterly; they have completed their 2024-2026 roadmaps and are now deep into 2027 planning. This sets a new, aggressive benchmark for what constitutes a top-tier engineering organization.