More The AI Native Dev episodes

Maggie Appleton: Agentic Coding Has Outgrown the Text Box thumbnail

Maggie Appleton: Agentic Coding Has Outgrown the Text Box

Published 3 Nov 2026

Duration: 00:52:11

Agentic coding requires shifting from text-based CLI tools to visual, collaborative interfaces to align human planning with agent execution, while humans retain strategic decision-making based on social context.

Episode Description

Agentic coding has outgrown the text box. Maggie Appleton, staff research engineer at GitHub Next, argues that walls of CLI output and AGENTS.md files...

Overview

Maggie Appleton, a staff research engineer at GitHub Next, discusses the limitations of current text-based, CLI-focused agent coding interfaces and argues they prefer the agent's preferences over human cognitive strengths. She proposes that agents should work proactively - researching ahead of time, presenting decisions with bespoke visual interfaces, and relieving developers' cognitive load rather than waiting to be prompted with multiple-choice questions.

She introduces the concept of "boundary objects" from sociology, applied to human-agent collaboration, arguing that while agent MD files and text-based context serve agents well, humans need social, real-time, interactive, and visual artifacts to plan effectively. Her project Chopin explores collaborative planning with agents using visual diagrams, multiplayer features, and structured classification of free-form chat into actionable plans.

The conversation also covers why individual acceleration with agents does not produce good collaborative products without team alignment, the growing importance of PMs in deciding whether to build something as building becomes cheap, how junior developers can learn by observing senior engineers' agent interactions, and why embodied social presence remains a human advantage agents cannot replicate.

What If

  • What if you replaced your text-based planning with a visual, interactive artifact?

    • Move: Set up a planning environment where you and an agent work on a shared visual document with diagrams, mockups, and interactive elements, rather than a static text file, mirroring Appleton's Chopin approach of turning "free form human conversation into a more structured artifact".
    • Why Now? Current tools are "so CLI focused" and "preferencing the agent's preferences and not the humans", but web and browser technologies already support the interactive, visual interfaces she describes, including "poke around an interactive diagram" and "move sliders".
    • Expected Upside: You could reduce the cognitive load of "constantly juggling seven different sessions" by maintaining a single visual artifact that captures decisions, constraints, and open questions in structured form, making handoffs to implementation agents clearer.
  • What if you configured your agents to research proactively before you need answers?

    • Move: Instruct your agents to perform web searches and compile comparison reports ahead of time on topics you anticipate needing during planning, rather than answering multiple-choice prompts in a text editor, following Appleton's proposal: "Don't wait for me to prompt you".
    • Why Now? She notes that "a lot of us I think are struggling with in the age of agents" due to cognitive load, and agents already have the capability to do web search and compile reports on demand.
    • Expected Upside: You could arrive at planning decisions with "reports ready for you with bespoke interfaces ready to go", reducing the time spent on individual research threads like comparing NPM libraries or evaluating design system changes.
  • What if you made your agent sessions visible to your team as a learning tool?

    • Move: Adopt an "open by default" approach where team members can see each other's agent chats and planning sessions, as Appleton describes: "everyone can see everyone else's agent chats".
    • Why Now? She argues this is "the way that you'd get junior developers to actually learn" by observing how senior engineers structure plans and ask questions in real time.
    • Expected Upside: Junior team members could learn "the quality of the level of plans that a senior engineer would hand off to an agent", potentially bridging the experience gap without formal pair programming, which she notes "must be totally dead by now".

Takeaway

  • When an agent asks you to choose between options in a text prompt, request a comprehensive comparison report instead: "Give me like a comprehensive report of how these compare" rather than picking from a multiple-choice list.
  • Build your planning artifacts as visual, interactive documents using web technologies - diagrams, mockups, sliders - rather than static text files, since humans are "much faster visual, spatial, social reasoners".
  • Configure your agents to perform proactive research and compile reports before you reach a decision point, so you arrive at planning topics with "reports ready for you with bespoke interfaces ready to go".
  • Make your agent sessions visible to your team by default so junior developers can learn by observing "the quality of the level of plans that a senior engineer would hand off to an agent".
  • After an agent completes an implementation run, request "a really nice visual report where you summarize everything you've done and why you did it and a list of all the decisions you made along the way" rather than reviewing a line-by-line diff.

Recent Episodes of The AI Native Dev

10 Nov 2026 Edith Harbaugh: Code Is Cheap. The Last Mile Isn't.

Feature flags and AI agents transform software delivery by enabling autonomous rollbacks and rapid iteration, though human judgment remains vital for defining success and maintaining stability.

27 Oct 2026 Caleb Sima: AI Attackers Move at Machine Speed

"AI-driven cybersecurity threats are evolving rapidly, with attackers leveraging speed, complexity, and swarm-like behavior to exploit vulnerabilities, necessitating advanced defensive strategies like automation and high-fidelity monitoring to counter escalating risks."

13 Oct 2026 Maria Gorinova: Feedback Loops That Fix Your Codebase

"Software factories and agentic coding require robust feedback loops to balance efficiency, reliability, and human oversight, with structured workflows and clear intent improving automation outcomes."

6 Oct 2026 Legora CTO: Why AI Agents Need More Than a Chat Window

"AI enhances legal and engineering workflows by automating tasks like drafting and coding, but human judgment remains critical for subjective or high-stakes decisions, with tools like Legora streamlining processes while addressing challenges in tooling, orchestration, and resource optimization."

More The AI Native Dev episodes