Research

We publish practical research on agentic coding, team adoption, multi-agent workflows, and the habits that hold up in production.

Editorial illustration for Make Codebases Easier for Agents. Most agent failures are codebase failures.
AI-Assisted Coding·6 min read

Make Codebases Easier for Agents

Structure repos so agents can find code, make local changes, and verify them fast.

April 16, 2026·Rogier Muller
Editorial illustration for AI Coding Workflows That Hold Up. A lot of AI coding advice starts with the tool and ends with the tool. That is usually the wrong
AI-Assisted Coding·5 min read

AI Coding Workflows That Hold Up

Practical workflows for coding tools that still work when the task gets messy.

April 15, 2026·Rogier Muller
Editorial illustration for Lead Sourcing with Agents. A lot of agent demos stop at code.
AI-Assisted Coding·5 min read

Lead Sourcing with Agents

Use an agent to turn public signals into a prospect list and draft outreach for human review.

April 14, 2026·Rogier Muller
Editorial illustration for Narrow Creative Skills for Coding Agents. Give an agent one narrow creative job and review the output against a clear checklist.
AI-Assisted Coding·5 min read

Narrow Creative Skills for Coding Agents

Give an agent one narrow creative job and review the output against a clear checklist.

April 13, 2026·Rogier Muller
Editorial illustration for Plain-English Agent Updates. Coding agents often do useful work, then describe it so vaguely that a human has to reconstruct the
AI-Assisted Coding·5 min read

Plain-English Agent Updates

A small instruction change can make agent output easier to review and trust.

April 12, 2026·Rogier Muller
Editorial illustration for Tools That Hold Up. A tool can look strong in a short demo and still fail in daily use.
AI-Assisted Coding·5 min read

Tools That Hold Up

A practical look at coding tools that stay useful after the demo.

April 11, 2026·Rogier Muller
Editorial illustration for MCP for UI Components. Shared integrations can turn design handoff into repeatable UI work.
AI-Assisted Coding·5 min read

MCP for UI Components

Shared integrations can turn design handoff into repeatable UI work.

April 10, 2026·Rogier Muller
Editorial illustration for Fast Evals for Better Decisions. Evals only help when they fit the pace of the work. If they take too long, people skip them.
Methodology·6 min read

Fast Evals for Better Decisions

Small, quick evals that fit the edit loop and support real coding decisions.

April 9, 2026·Rogier Muller
Editorial illustration for Keep Agents Moving. A practical pattern for unsticking coding agents on long tasks.
AI-Assisted Coding·5 min read

Keep Agents Moving

A practical pattern for unsticking coding agents on long tasks.

April 8, 2026·Rogier Muller
Editorial illustration for Specs, Tests, Stable Stacks. Agentic coding works better when the task is constrained.
Methodology·5 min read

Specs, Tests, Stable Stacks

Clear specs, good tests, and stable stacks make agentic coding more reliable.

April 7, 2026·Rogier Muller
Editorial illustration for Stop Using CSS Selectors in E2E Tests. CSS selectors are a convenient trap in end-to-end tests. They feel precise, but they are tied
Methodology·6 min read

Stop Using CSS Selectors in E2E Tests

CSS selectors make agent-written E2E tests brittle. Use stable, user-facing hooks instead.

April 6, 2026·Rogier Muller
Editorial illustration for A Better Bug-Finding Prompt. A small prompt change can shift how an agent reads code.
AI-Assisted Coding·5 min read

A Better Bug-Finding Prompt

A practical prompt pattern for surfacing more bugs in agentic coding workflows.

April 5, 2026·Rogier Muller
Editorial illustration for Agentic Team Workflows. It is getting work through a team cleanly.
Methodology·6 min read

Agentic Team Workflows

Practical workflow patterns for teams using coding agents on real engineering tasks.

April 4, 2026·Rogier Muller
Editorial illustration for Tools That Keep Working. A lot of AI coding tools look good in the first ten minutes.
AI-Assisted Coding·5 min read

Tools That Keep Working

A practical look at AI coding tools that stay useful after the demo.

April 3, 2026·Rogier Muller
Editorial illustration for Composer Layers in Agentic Coding. Agentic coding tools are converging on the same hard problem: turning a vague request into a safe,
AI-Assisted Coding·5 min read

Composer Layers in Agentic Coding

How composer layers turn intent into edits, and where the workflow still breaks.

April 2, 2026·Rogier Muller
Editorial illustration for Headless Agent Runs for Scripts and CI. Coding agents are most useful when they can leave the editor.
AI-Assisted Coding·5 min read

Headless Agent Runs for Scripts and CI

Run coding agents in scripts, CI, and automation with checks that keep output reviewable.

April 1, 2026·Rogier Muller
Editorial illustration for Long-Running Agent Loops. Some coding agents work better when they stay alive.
AI-Assisted Coding·5 min read

Long-Running Agent Loops

Why long-running coding agents help on iterative, verification-heavy tasks.

March 31, 2026·Rogier Muller
Editorial illustration for Agent Loops for Messy Code. AI coding tools are easy to judge on simple tasks.
AI-Assisted Coding·6 min read

Agent Loops for Messy Code

Practical patterns for keeping coding agents useful on messy tasks.

March 30, 2026·Rogier Muller
Editorial illustration for AI Coding Tools That Last. A lot of AI coding tools look good in a short demo. Fewer stay useful after a week of real work.
AI-Assisted Coding·5 min read

AI Coding Tools That Last

A practical look at which AI coding tools stay useful after the first demo.

March 29, 2026·Rogier Muller
Editorial illustration for Agent Teams Skip Prompt Tuning. Top AI engineers do not spend their day polishing prompts. They shape the work around the agent.
AI-Assisted Coding·5 min read

Agent Teams Skip Prompt Tuning

Agentic teams get better results from workflow design than from manual prompt tuning.

March 28, 2026·Rogier Muller
Editorial illustration for Docs That Return Markdown. Agentic tools are changing how documentation gets consumed.
Methodology·6 min read

Docs That Return Markdown

Why docs return markdown for agents, and how to do it without hurting human readers.

March 27, 2026·Rogier Muller
Editorial illustration for Browser Checks for Coding Agents. Coding agents are good at writing code.
AI-Assisted Coding·5 min read

Browser Checks for Coding Agents

Let coding agents verify UI changes in a real browser, then patch based on what they see.

March 26, 2026·Rogier Muller
Editorial illustration for Keep the Stack Current, Keep Agents in Bounds. A recent starter stack update is a useful reminder: most AI coding gains come from
AI-Assisted Coding·4 min read

Keep the Stack Current, Keep Agents in Bounds

A practical look at refreshed dependencies and short rules that make AI coding easier to review.

March 25, 2026·Rogier Muller
Editorial illustration for Defensive Code from Agentic Tools. Some coding agents lean toward defensive code.
AI-Assisted Coding·5 min read

Defensive Code from Agentic Tools

Why some coding agents add extra checks, where that helps, and where it slows reviews.

March 24, 2026·Rogier Muller
Editorial illustration for Browser Choice Shapes Agent Loops. Browser backend choice can change agent speed, retries, and recovery.
AI-Assisted Coding·5 min read

Browser Choice Shapes Agent Loops

Browser backend choice can change agent speed, retries, and recovery.

March 23, 2026·Rogier Muller
Editorial illustration for AI Coding Wrappers That Hold Up. AI coding wrappers are easy to dismiss.
AI-Assisted Coding·5 min read

AI Coding Wrappers That Hold Up

When wrappers help, where they fail, and how to test them on real coding work.

March 22, 2026·Rogier Muller
Editorial illustration for Subagents Need Separate Prompts. Subagents are recursive agents.
AI-Assisted Coding·5 min read

Subagents Need Separate Prompts

Give each subagent one job, one output, and one stop rule.

March 20, 2026·Rogier Muller
Editorial illustration for Global Rollout for AI Coding Tools. AI coding tools are no longer judged only on benchmark scores or demos.
AI-Assisted Coding·5 min read

Global Rollout for AI Coding Tools

A practical look at rollout patterns for AI coding tools across regions and teams.

March 18, 2026·Rogier Muller
Editorial illustration for Maintaining Reliable AI Coding Workflows. AI coding tools are now part of many engineering processes, but their usefulness depends on
AI-Assisted Coding·3 min read

Maintaining Reliable AI Coding Workflows

Practical steps and tradeoffs for keeping AI coding workflows effective with agent orchestration.

March 16, 2026·Rogier Muller
Editorial illustration for Patterns for Reliable AI Coding Workflows. AI coding tools are common in engineering workflows, but their reliability varies.
AI-Assisted Coding·4 min read

Patterns for Reliable AI Coding Workflows

Workflow patterns that help engineering teams get consistent results from AI coding tools.

March 15, 2026·Rogier Muller
Editorial illustration for Using Async Subagents to Speed Up AI Coding Workflows. AI coding tools are moving beyond single-threaded interactions.
AI-Assisted Coding·3 min read

Using Async Subagents to Speed Up AI Coding Workflows

Background async subagents run tasks in parallel, improving AI coding workflow efficiency.

March 14, 2026·Rogier Muller
Editorial illustration for How Returning Markdown from Docs Shapes Agentic Coding Workflows. Engineering teams working with coding agents often adjust their
AI-Assisted Coding·5 min read

How Returning Markdown from Docs Shapes Agentic Coding Workflows

A practical examination of how engineering teams adapt documentation to return markdown for coding agents, focusing on token efficiency, implementation steps, tradeoffs, and limitations.

March 13, 2026·Rogier Muller
Editorial illustration for MCP for Faster Iteration. Integrating Playwright's Multi-Context Playwright (MCP) mode into agent-driven workflows can shorten
AI-Assisted Coding·5 min read

MCP for Faster Iteration

A practical examination of how integrating Playwright MCP into agentic coding workflows affects iteration speed, implementation steps, tradeoffs, and limitations for engineering teams working with coding agents.

March 12, 2026·Rogier Muller
Editorial illustration for Agent-Readable Docs in Markdown. Teams shipping coding agents are changing how they write and serve documentation.
AI-Assisted Coding·6 min read

Agent-Readable Docs in Markdown

A practical look at returning markdown from internal docs for coding agents: why teams do it, how to implement it, and what it actually changes in agentic workflows.

March 11, 2026·Rogier Muller
Editorial illustration for Faster Loops With Playwright MCP. Agentic coding tools write code well but rarely validate it.
Cursor Features·5 min read

Faster Loops With Playwright MCP

A concrete look at how to wire Playwright MCP into agentic coding workflows, what it actually changes for teams, and where the limits are.

March 9, 2026·Rogier Muller
Editorial illustration for When AI Coding Tools Regress. Those pieces change without your consent.
AI-Assisted Coding·5 min read

When AI Coding Tools Regress

A practical look at how AI coding agents change workflows, quality bars, and expectations for engineering teams—and what to do when a core model quietly gets worse.

March 7, 2026·Rogier Muller
Editorial illustration for Running Multi-Agent Teams. Teams are starting to pair one reasoning‑heavy model as a conductor ("Opus 4.6") with several faster
AI-Assisted Coding·12 min read

Running Multi-Agent Teams

A practical look at how an "Opus 4.6" style orchestrator could coordinate a fleet of "Codex 5.3" coding agents, what this changes for engineering teams, and how to implement it without breaking your workflow.

March 1, 2026·Rogier Muller