Keynote: AI-Powered App Development - Steve Sanderson - NDC London 2026

Keynote: AI-Powered App Development - Steve Sanderson - NDC London 2026

by Steve Sanderson

Applied Sciences & Engineering AI — Models & Research UYQArtificial intelligence (AI)UYQFGenerative artificial intelligence
🎙 Steve Sanderson 👥 227K 📅 February 25, 2026 ⏱ 54 min 👁 32K 📄 expert opinion 🧭 2026-08-13
Available in: English (current) Français

Keywords

coding agentsAI-assisted developmentCopilot CLIsub-agentsplan mode

Summary

In this keynote, Steve Sanderson from the GitHub Copilot coding agent team discusses the transformative impact of AI on software development. He begins by surveying the audience’s usage of AI coding tools, noting that a significant portion already use CLI-based agents. He argues that coding agents represent a major shift in the industry, enabling developers to produce code much faster. He demonstrates a simple feature addition using Copilot CLI, highlighting the speed and ease. He then explores how this changes developers’ relationship with code: code becomes cheap, prototypes are free, and design processes become more iterative. Sanderson introduces key features of coding agents, including sub-agents, plan mode, skills, delegates, memories, and hooks. He demonstrates a parallel code review using multiple AI models (Opus, Haiku, Gemini) to compare their findings. He also discusses the rapid evolution of tools and the FOMO it creates, mentioning projects like Gastown and Ralph Wiggum. He emphasizes the importance of planning and clear communication to avoid agent laziness. The talk concludes with a live demo of adding a product customization feature to an e-commerce app, showing how plan mode helps structure the work. Sanderson reflects on the changing role of developers and the need to adapt to a new paradigm where AI is a collaborative partner.

210 words

Critical Evaluation

Value of the Information & Strength of the Argument

The talk provides valuable insights into the practical use of coding agents, backed by real-world examples and live demos. Sanderson’s argumentation is solid, drawing on his experience on the Copilot team and presenting concrete metrics (e.g., 200 PRs merged in a week). He effectively demonstrates the capabilities and limitations of current tools, and his advice on planning and using sub-agents is actionable. However, some claims are anecdotal and not scientifically validated, and the talk is inherently promotional of Microsoft’s products, though he does acknowledge alternatives.

Scientific Rigor, Source Quality, Title Accuracy

The talk is scientifically rigorous in its practical demonstrations, but it lacks formal citations. The speaker references internal team data and personal experience, which are credible but not independently verifiable. The title accurately reflects the content, focusing on AI-powered development. The talk is well-structured and the technical details are accurate based on current knowledge. No external sources are cited beyond the conference links, but the speaker’s authority adds credibility.

169 words

Title / Content Match

The title accurately reflects the content: a keynote on AI-powered app development, focusing on coding agents and their impact on developer workflows.

Quality & Reliability

8/10

The speaker is a recognized expert from the GitHub Copilot team, providing practical insights and live demos. Claims are based on personal experience and team metrics, but not peer-reviewed. The talk is opinionated and forward-looking, with some speculative elements.

Key Moments

Markers derived by PSI from the transcript: the creator did not define chapters.

Cited Sources

  • NDC Conferences — Conference organizer and source of the talk
  • NDC London — Specific conference where the talk was recorded

Concurring Sources

  • GitHub Copilot — Official product page for GitHub Copilot, aligning with the talk's focus.

Dissenting Sources

  • Critiques of AI coding agents — Some developers express concerns about code quality and job displacement, which are not fully addressed in the talk.

External References

Contribution & Novelties

The talk provides a practical, insider perspective on using coding agents, with live demos and actionable advice. It highlights the shift in developer mindset and the importance of planning. The ‘Pour aller plus loin’ section suggests further exploration:

Pour aller plus loin :

  • GitHub Copilot documentation — Official documentation for GitHub Copilot, including CLI and agent features.
  • Claude Code documentation — Documentation for Claude Code, a leading CLI coding agent.
  • Sub-agents in coding agents — Wikipedia article on software agents, providing background on autonomous agents.
  • Plan mode in AI tools — Wikipedia article on planning, relevant to the planning approach discussed.
  • Ralph Wiggum project — GitHub repository for Ralph Wiggum, a tool to force coding agents to complete tasks.

119 words

Radar Profile

The radar profile shows high scores in information quantity and quality, reflecting the talk's rich content and practical insights. The technical level is moderate, suitable for a broad developer audience. The overall reliability is good, though not fully verifiable due to lack of formal citations.

Reliability 7/10

💬 Sur les 0 commentaires analysés, aucune tendance n'est disponible.