Interview, Fireside Chat
Your Agent Can't Fix What It Can't See | Milin Desai, Sentry | RAISE Summit 2026
- Code creation is projected to undergo an exponential rise to 100x historical volumes with significant increases in existing code modifications, while the industry faces a prevalent problem of production code breaking due to accelerating development velocity.
- Self-healing software is expected to become an industry standard within the "rest of the year and going into next year," replacing the "wild experimentation phase" of the "past six to nine months" with "more granular, deterministic loops of automation."
- AI is anticipated to resolve the current 30% to 40% of relatively trivial issues without human intervention, with deployments and root cause analysis occurring within "five to ten minutes" rather than the current "hours later" timeframe, though success relies on a combination of technology improvements and model advancements.
- The developer role is projected to evolve from pure coding to encompassing "design experience," "system-level thinking," "operating system fundamentals," and "contextual communication," with an increased emphasis on "telling stories" to effectively communicate with users and customers.
- While the core practice of building software will persist and immediate hiring needs remain, a reduction in headcount is forecasted for the longer term as automation increases, potentially leading to smaller teams where engineers must handle customer deployment and feedback.
- Organizations are expected to adopt a multi-model strategy utilizing "a little open source," "a Little Anthropic," and other models, mirroring the evolution of cloud monitoring into a horizontal capability rather than a single-vendor solution.
- Current AI costs are identified as a significant concern, with "3% to 5% of the engineering team" now carrying a dedicated cost line, necessitating a shift from "token maxing" to establishing a direct correlation between spending and revenue.
- Sentry aims to serve the entire install base of 200,000 teams, from three-person startups to billion-user companies, by providing unmatched breadth and depth of coverage across languages and interfaces to deliver ubiquitous self-healing software.
- The market is expected to see increased pressure on companies to launch proprietary monitoring, yet the industry will favor specialized horizontal capabilities over building internal solutions, driven by the cycle where more software creation leads to more users and subsequent demand.
- Future efficiency is predicted to improve in terms of token usage and outcome speed, supported by the combination of model advancements like Cloud 6.0 and layered context or memory, despite current limitations in determinism and the risk of AI errors requiring verification agents.