AI-coding agents are now the default. What comes next?

2026-06-25 · Source: LeadDev · Field: Technology & Digital — Software Development & Engineering, Artificial Intelligence & Machine Learning · Depth: Intermediate, medium

Summary

Linear, a product development system, has fully embraced AI-coding agents, making them integral to workflows like triage, bug investigation, and pull request generation. Following an internal mandate on January 7, 2026, to use tools like Claude Code or Opencode with Opus, the company observed significant productivity gains. In February, per-engineer output increased by 30% for PRs (52.7 per author) and 33% for issues closed (44.4 per author) compared to January, with numbers climbing further in March. Over 50% of agent-generated PRs are mergeable. This shift has made context, not code generation, the new bottleneck. Linear reduced token usage by approximately 40% through context engineering, utilizing tools like rtk and emphasizing upfront planning with RFCs. While managing costs from commercial models is a growing concern, Linear explores open-source alternatives for less demanding tasks. Human oversight remains critical, with engineers approving all AI-generated changes and implementing guardrails.

Key takeaway

For Engineering Directors evaluating AI-coding agent adoption, recognize that these tools are rapidly becoming standard, offering substantial productivity boosts. You should mandate specific agent tools and prioritize context engineering to manage token usage and costs. Crucially, maintain human approval gates for all AI-generated code to ensure quality and mitigate risks, even as you scale agentic workflows.

Key insights

AI-coding agents are now default, demanding context optimization and human-in-the-loop processes for scalable productivity.

Principles

• AI agents dramatically increase developer output.
• Context engineering is critical for token efficiency.
• Human review remains vital for AI-generated code.

Method

Linear mandated specific AI-coding tools, then optimized context with rtk and RFCs, while managing costs and maintaining human approval gates for all changes.

In practice

• Standardize on specific AI-coding tools.
• Optimize agent context with tools like rtk.
• Implement human approval for all AI code merges.

Topics

• AI-coding Agents
• Large Language Models
• Context Engineering
• Software Development
• Cost Optimization
• Human-in-the-Loop

Code references

• rtk-ai/rtk

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

Related on AIssential

• Quoting James Shore — AI coding agents must inversely reduce maintenance costs to justify increased code output.
• Context Is the New Code — Patrick Debois, Tessl — Effective AI agent performance hinges on meticulously engineered context, necessitating a dedicated Context Development Life Cycle.
• Coders are refusing to work without AI — and that could come back to bite them — Widespread AI coding tool adoption masks dubious productivity gains and potential increases in maintenance costs.
• Judgment and creativity are all you need. — Coding agents move software development bottlenecks from implementation to judgment and creativity.
• The Pulse: What will the Staff Engineer role look like in 2027 and beyond? — AI-authored code is rapidly increasing, shifting developer roles and raising concerns about escalating AI infrastructure costs.
• 我的「破產版」賈維斯：調教 AI Agent 的血淚史 — Effective AI agent training requires teaching judgment and cost awareness, not just providing powerful tools.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by LeadDev.