10 Tips To Level Up Your AI-Assisted Coding - Aleksander Stensby - NDC London 2026

Value of the Information & Strength of the Argument

The talk provides high practical value, offering concrete, actionable tips that developers can immediately apply. Stensby’s argumentation is solid, grounded in his personal experience and observations of industry trends. He effectively counters common objections by emphasizing the rapid evolution of AI models and the importance of adapting workflows. The advice is well-structured, moving from mindset to specific techniques, and is supported by real-world examples and references to tools like Claude Code and Cursor. The speaker’s enthusiasm is balanced with practical warnings about costs and the need for human oversight. The argumentation is persuasive but relies heavily on anecdotal evidence rather than empirical data, which is a limitation.

Scientific Rigor, Source Quality, Title Accuracy

The talk demonstrates good scientific rigor in its practical approach, though it lacks formal citations. Stensby references specific tools and practices (e.g., Claude Code, Cursor, MCP) and mentions industry figures like Boris and Andrej Karpathy, but does not provide verifiable sources. The title accurately reflects the content, and the talk stays on-topic throughout. The speaker’s credibility is enhanced by his 20 years of development experience and his active engagement with the AI coding community. However, the absence of external references and the reliance on personal testimony limit the scientific rigor. The talk is more of an expert opinion than a research-based presentation.

The title accurately reflects the content: a list of ten practical tips for improving AI-assisted coding, delivered in a conference talk format.

8/10

The talk is based on the speaker's extensive practical experience and references to well-known tools and practices. It lacks formal citations but provides actionable, reproducible advice. The speaker acknowledges limitations and encourages critical thinking.

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

• 02:00Introduction: Speaker shares his background and sets the stage for the talk.
• 05:00Discussion on why developers are skeptical, referencing the rapid improvement of AI models.
• 10:00Emphasis on mindset shift: treating AI as a co-worker rather than a tool.
• 15:00Introduction to compound engineering and the importance of learning from mistakes.
• 20:00Tip 1: Context is king - providing detailed context to reduce hallucinations.
• 25:00Managing context windows and using subagents for parallel tasks.
• 30:00Tip 2: Point AI to relevant documentation and use markdown files.
• 35:00Introduction to skills and how they can trigger context-specific behaviors.
• 40:00Tip 3: Always start with a plan using plan mode and review it.
• 45:00Using AI to generate ideas and leveraging human taste for guidance.

• NDC Conferences — Mentioned as the conference organizer and source for future events.
• NDC London — Mentioned as the specific conference where the talk was recorded.

• Every - Compound Engineering — The concept of compound engineering is referenced in the talk and aligns with the speaker's advice on iterative improvement.
• Anthropic - Claude Code — The talk heavily references Claude Code features, and this documentation supports the claims about skills and plan mode.

• Critiques of AI coding assistants — Some developers argue that AI coding assistants can introduce security vulnerabilities and reduce code quality if not properly supervised, a concern the speaker acknowledges but does not fully address.