Sign InOpen Brain
AI EngineerVideoSource Linked

How to Get Your Org to Adopt Coding Agents (Without Shipping Garbage) — Eyal Blum, Figma

Figma’s cautious agent adopters expose missing safeguards, not resistance. Use their feedback to improve verification, plans, review size, and the labeling of AI-generated work.

AI Engineer · Aug 28, 2026
Open Source Open MarkdownOpen JSON
Source Summary

Figma reports uneven agent adoption and says experienced engineers may be slowest because they see codebase risks first. One detailed plan turned an estimated **six weeks of coding into about one week**, including review.

Practical Implication

Make verification executable before implementation: use TDD, encode repeated checks deterministically, and split plans into independently validated, reviewable phases. Put the reason for the change first, and use skeptical maintainers’ feedback as the safety roadmap.

Agent-Ready Context
Figma reports uneven agent adoption and says experienced engineers may be slowest because they see codebase risks first. One detailed plan turned an estimated **six weeks of coding into about one week**, including review.

Make verification executable before implementation: use TDD, encode repeated checks deterministically, and split plans into independently validated, reviewable phases. Put the reason for the change first, and use skeptical maintainers’ feedback as the safety roadmap.

The workflow is still evolving and depends on planning and review effort that raw generation metrics can hide. AI also increases communication volume, so clearly separate human-written intent from generated descriptions to protect scarce reviewer attention.
Connected Context · Feed7 Judgment

This confirms that coding-agent adoption is primarily a verification and workflow-design problem, not a raw generation problem. It adds an organizational consequence: skeptical experienced maintainers can identify the controls needed for safe rollout. The reported acceleration remains local and includes planning and review, while increased generated communication can consume the reviewer capacity automation is meant to save.

How to Kill the Code Review — Ankit Jain, AviatorFigma’s executable checks and separation of human intent from generated description support the candidate’s shift from diff review toward reviewing retained intent and verification evidence.Loop Engineering from First Principles — Kyle Mistele, HumanLayerSplitting plans into independently validated phases implements the candidate’s bounded-change loop and preserves review points before work expands.MineValiCoder: Reliable Code Generation with Test Case Quality Mining and Bipartite Graph-Based Mutual ValidationFigma relies on tests as implementation constraints, while MineValiCoder addresses the prerequisite that generated tests themselves must be reliable enough to steer the code.From AI-Assisted to AI-Native: Building a Frontier Development Team — Clare Liguori, AWSBoth attribute gains to scoped plans, fast feedback, and workflow preparation, and both warn that review and decision capacity can become the downstream bottleneck.
Context Map
agentcoding#coding-agents#agent-reliability#harness-engineering
Uncertainty
The workflow is still evolving and depends on planning and review effort that raw generation metrics can hide. AI also increases communication volume, so clearly separate human-written intent from generated descriptions to protect scarce reviewer attention.