Braintrust CEO: Evals are the new PRD for AI products

2026-06-15 · Source: How I AI · Field: Technology & Digital — Software Development & Engineering, Artificial Intelligence & Machine Learning, Cloud Computing & IT Infrastructure · Depth: Advanced, extended

Summary

Anker Goyel, CEO of Brain Trust, advocates for coding agents and rigorous evaluations (evals) as the new standard for tackling complex engineering challenges, effectively replacing traditional Product Requirements Documents (PRDs). He argues that AI models, particularly tools like Codex and GPT models, excel at writing code, shifting the engineering focus from "how" to "what." This enables engineers to define problems and success criteria, then deploy agents to exhaustively test solutions, such as optimizing slow database queries across billions of traces or performing complex schema-to-schema data migrations. This approach significantly enhances practical quality and rigor, allowing companies to address technical debt and performance issues without the prohibitive cost of manual human effort, ultimately leading to higher quality bars and faster progress.

Key takeaway

For engineering leaders aiming to accelerate technical problem-solving and ensure product quality, embrace AI coding agents and rigorous evals as the new standard for tackling complex infrastructure work, like database optimization or data migration. Re-evaluate tasks that fall "below the agent line" to free up your human engineers for higher-level work. Prioritize investment in CI and building robust feedback loops for AI products to achieve continuous improvement and backlog reduction, making "no excuse" for performance or quality issues.

Key insights

Evals serve as the modern PRD for AI products, enabling coding agents to rigorously test and optimize complex engineering challenges.

Principles

• AI shifts programming focus from "how" to "what."
• No excuse for lack of rigor or performance with AI agents.
• Practical quality of engineering with AI increases.

Method

Define problems and success criteria via evals, then deploy coding agents to exhaustively test solutions, such as database optimizations or data migrations, in a safe, iterative loop.

In practice

• Use coding agents for database query optimization.
• Automate complex schema-to-schema data migrations.
• Build robust AI product feedback loops with evals.

Topics

• Coding Agents
• AI Evals
• Database Optimization
• Software Engineering
• CI/CD
• Product Requirements Document

Best for: AI Engineer, Machine Learning Engineer, CTO, Software Engineer, Director of AI/ML, VP of Engineering/Data

Related on AIssential

• How Braintrust uses AI agents, evals, and CI to ship better software | Ankur Goyal — AI agents and rigorous evals encode "taste" and success criteria, enabling automated, high-quality software development.
• The Sequence Opinion #860: Every Company’s Last eXam: Some Reflection About Practical AI Evals — Company-specific, dynamic evaluations are now essential for production AI, forming a fourth pillar alongside compute, data, and models.
• AI Dev 26 x SF | Ara Khan: Evals Are Broken Use Them Anyway — Evals are imperfect but essential for AI agent development, requiring nuanced interpretation and iterative improvement.
• It's 2026, and We're Still Talking Evals — Effective AI agent evaluation requires moving beyond simple accuracy to real-world user behavior and production-aligned metrics.
• Evals Skills for Coding Agents — Improving infrastructure around AI agents, especially evaluation capabilities, is more critical than solely improving the underlying model.
• Ship Real Agents: Hands-On Evals for Agentic Applications — Laurie Voss, Arize — Formal evaluations are crucial for AI agents, moving beyond subjective "vibe checks" to ensure reliability and performance.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by How I AI.