What's Easy Now? What's Hard Now?

2026-05-18 · Source: Marc Brooker's Blog · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Software Development & Engineering · Depth: Advanced, medium

Summary

The article discusses how AI coding agents are transforming software development, particularly focusing on the role of feedback loops. It posits that agents excel at tasks with effective, immediate feedback, drawing an analogy to electronic circuits where feedback enables complex operations from simple components. The author argues that while open-loop models might find UI/website development "easy," the "feedback loop hypothesis" suggests that system software, like high-performance database storage engines, will become "easier" for agents in the long term due to the availability of precise, quantifiable feedback. Conversely, tasks like architecture or concurrent programming, where feedback is subjective or delayed, will remain "hard." This perspective challenges conventional intuition about AI's capabilities in software engineering.

Key takeaway

For AI Engineers developing agent-driven software, recognize that the "ease" of a task for an agent correlates directly with the quality and immediacy of feedback. You should prioritize robust specification and integrate tools like Rust, Hydro, or TLA+ to create strong feedback loops. This approach will make complex system software development more tractable for agents than traditionally "easier" UI tasks, shifting your focus towards building effective feedback mechanisms.

Key insights

AI coding agents' long-term capabilities are primarily determined by the availability and effectiveness of feedback loops, not just open-loop model behavior.

Principles

• Feedback transforms component behavior.
• Agent capabilities are limited by feedback availability.
• Quantifiable feedback simplifies complex tasks for agents.

In practice

• Prioritize robust specification for agent-driven development.
• Utilize compile-time and modeling tools like Rust or TLA+.
• Employ property-based testing for agent-generated code.

Topics

• AI Coding Agents
• Feedback Loops
• Software Engineering
• Large Language Models
• System Software
• Specification

Code references

• verus-lang/verus
• p-org/P

Best for: AI Product Manager, AI Engineer, Software Engineer, Director of AI/ML

Related on AIssential

• AI tools for hearing difficulties — helpful or harmful for language learning? — AI assistive technology presents a dilemma for language learners with hearing difficulties, balancing aid with skill development.
• How Enterprise Marketing Teams Are Actually Using AI Agents in 2026 — Enterprise marketing is shifting from human-assisted AI tools to autonomous AI agents that redesign entire workflows.
• [AINews] Microsoft Build: MAI-Thinking-1 and MAI Family models — Microsoft's MAI models showcase a transparent, full-stack AI strategy emphasizing clean data and local agent execution.
• Are we getting better at explaining things because of our interaction with LLMs? — Crafting clear LLM prompts may enhance human explanatory and communication abilities.
• OpenAI expands Codex with role-specific plugins to build a general-purpose app for non-developers — OpenAI is transforming Codex into a general-purpose, role-specific AI work app, expanding beyond coding for non-developers.
• How to Use Claude Managed Agents? — Anthropic's Claude Managed Agents simplifies AI agent deployment by providing a hosted, stateful, and secure execution environment.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Marc Brooker's Blog.