Fragments: May 27

2026-05-27 · Source: Martin Fowler · Field: Technology & Digital — Software Development & Engineering, Artificial Intelligence & Machine Learning · Depth: Advanced, extended

Summary

Kent Beck and Martin Fowler discussed LLM-augmented programming at GOTO 2025, highlighting its "slot machine" nature and the need for human oversight. Ian Johnson's experience demonstrated how robust testing and static analysis enable AI agents to ship production code. The NHS controversially closed open-source repositories due to LLM security fears, countered by GDS advocating secure-by-design. Adam Tornhill noted the cognitive endurance challenge of "compressed cognition" with LLMs, suggesting small tasks. Broader concerns include Gen Z's growing distrust of AI and its potential economic impact on graduate hiring, with a 6.6% drop for most exposed jobs. US AI governance efforts are criticized for being too modest and lacking expertise. The discussion also touched on agile at scale, emphasizing user communication and flow optimization, and advice for junior developers on continuous learning and domain understanding.

Key takeaway

For AI Engineers and development leads integrating LLMs, recognize that AI agents, while powerful, require significant human oversight and robust verification. Prioritize establishing comprehensive testing and static analysis before increasing AI autonomy, and actively manage the cognitive load on developers by structuring tasks to allow for learning and critical review, rather than blindly trusting generated code.

Key insights

LLM-augmented programming demands human curation, robust testing, and awareness of cognitive load due to AI's inherent unreliability.

Principles

• LLMs are powerful but unreliable; constant verification is crucial.
• Prioritize communication between development teams and users/customers.
• Identify and reduce workflow queues to enhance organizational agility.

Method

Implement characterization tests, static analysis, and architectural patterns before increasing AI agent autonomy. Shift human roles from code writer to curator, focusing on defining patterns, reviewing test specs, and making strategic decisions.

In practice

• Query LLMs multiple times with varied prompts to compare results and assess consistency.
• Utilize post-feature development gaps for learning, refactoring, and test generation with LLM assistance.
• Structure AI agent tasks to be small and focused to mitigate cognitive endurance issues.

Topics

• LLM-augmented Programming
• AI Agent Reliability
• Software Development Practices
• Agile at Scale
• Cognitive Load
• AI Governance
• Open-Source Security

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

Related on AIssential

• Question regarding multiple posts on numerous subreddits containing variations of the same query. — Observation of potential AI-driven social media queries probing abstract human aesthetic concepts.
• AI is pulling the strings — The cartoon illustrates AI's growing influence, potentially controlling human intelligence.
• How the EU’s Tech Sovereignty Package Finally Puts Open Source to the Test — EU's new Tech Sovereignty Package centers open source as a strategic lever to reduce digital dependencies and foster a sustainable ecosystem.
• The Download: Trump’s new AI order, and smart glasses for warfare — AI governance is evolving towards voluntary oversight, while AI applications expand into defense and raise ethical concerns.
• Why the AI Policy Debate Should Focus More on the Harness and Protocol Layers — AI policy must shift focus to architectural security, user ownership, and governance of AI's harness and protocol layers.
• Meta AI safety director lost control of her agent. It started deleting her emails — AI agents can exhibit unpredictable, harmful autonomous behavior, even under expert supervision.

Open in AIssential →

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