TechPrufer
Get started
AI Engineer
AI Engineer logoAI Engineer · Feature trail

AI-Assisted Code Verification and Deterministic Testing

4 videos · Oct 2026

What this trail covers

This post contributes to the field of AI-assisted software development by highlighting the bottleneck of verification for AI-generated code. It introduces a system that records user flows, replays them on every pull request, and provides visual diffs for near-exhaustive frontend verification with zero developer effort. Key technologies include fully mocked network traffic, a deterministic browser, and coverage-guided flow selection, enabling code coverage to finally mean 'code tested' by capturing user interactions.

All AI Engineer trails
  1. Build with Perception Agents — Emile Baizel & Shruti Arora, Amazon AGI LabThis talk details the development of a system that integrates coding agents with visual perception primitives for frontend development. It introduces two core components: deterministic visual annotation, which allows users to select elements, draw, or pin on a webpage to export structured feedback (JSON) for coding agents, and visual verification, which checks the rendered site against design expectations and user flows. The system connects a browser extension (Amazon Nova Act) to a local bridge and a coding agent, creating a loop where feedback is applied, code is edited, verification is run, and a report is generated. The talk showcases this loop with a podcast landing page example and discusses various audience-driven integrations and concepts, emphasizing visual human-agent collaboration and verifying rendered results over agent claims.
  2. AI Writes More PRs. Who Validates Them? — Ali-Reza Adl-Tabatabai, SonarThis talk introduces an agentic validation system designed to address the bottleneck of code review and CI for AI-generated code. The system automates PR review, analyzes CI failures, fixes issues until the PR is green, and can auto-approve and merge based on user-defined rules. It combines agents with program analysis (SonarQube) to provide more comprehensive validation than either approach alone. The talk also discusses building trust in automated validation and provides insights for platform teams and leaders.
  3. Why AI Didn't Actually Make You Ship Faster — Gabriel Spencer-Harper, MeticulousThis talk introduces a novel approach to verifying AI-generated code by shifting the bottleneck from AI code generation to human review. The proposed solution, Meticulous, employs a record-and-replay mechanism for frontend verification. It achieves this by: 1) Mocking network traffic to ensure consistent responses. 2) Utilizing a deterministic browser to eliminate flaky tests. 3) Employing coverage-guided flow selection to ensure all critical user paths are tested. This system provides visual before/after diffs on every pull request, effectively making code coverage synonymous with tested code by capturing screenshots of every interaction step.
  4. Why 99% Accurate Browser Agents Still Fail — Derek Meegan, BrowserbaseThis talk addresses the critical challenge of reliability for browser agents in production, particularly for long trajectories. It highlights that a 99% per-step accuracy compounds to a low success rate over many steps. The core contribution is a production architecture that mitigates this by: 1. Using deterministic tools for predictable browser interactions. 2. Employing OCR verification for robust content validation. 3. Separating authentication to prevent agent failure. 4. Utilizing 'skills' as standard operating procedures to keep agents on the critical path. The talk emphasizes measuring success per transaction rather than per run and prioritizing performance, cost, and maintainability as engineering problems.