10 tips to level up your ai-assisted coding - Aleksander Stensby - NDC AI 2026

Item: 10 tips to level up your ai-assisted coding - Aleksander Stensby - NDC AI 2026
Rating: 4
Author: PostSingularity Institute

by Aleksander Stensby

Applied Sciences & Engineering AI — Models & Research UYQArtificial intelligence (AI)UYQFGenerative artificial intelligence

🎙 Aleksander Stensby 👥 227K 📅 July 1, 2026 ⏱ 62 min 👁 6K 📄 expert opinion 🧭 2026-08-13

Available in: English (current) Français

Keywords

context windowrulesskillsplan modecompound engineering

Summary

In this NDC AI 2026 talk, Aleksander Stensby shares practical tips for maximizing the effectiveness of AI coding assistants like Claude Code. He emphasizes a pair-programming mindset and the concept of compound engineering, where the AI learns from each interaction. The core advice revolves around managing context: treating the context window as a finite resource, monitoring its usage, and guiding auto-compaction. He recommends using rules files (Claude.md/agents.md) but cautions against over-scaffolding, suggesting to periodically prune rules as models improve. Skills are highlighted as a key investment, allowing users to encode preferences and capabilities, with a recommendation to build custom skills rather than downloading many. Planning is stressed as essential, with plan mode serving as a safe exploration and specification tool. He also touches on the importance of verification, using plans to check work, and the potential of MCP to connect AI to external tools. The talk concludes with a call to embrace the changing role of developers and to continuously ask the AI for what’s possible.

166 words

Critical Evaluation

Value of the Information & Strength of the Argument

The talk provides high-value, actionable advice for developers using AI coding tools. The speaker’s experience and role as an Anthropic ambassador lend credibility. Arguments are well-structured, with clear explanations and practical examples. The emphasis on context management and skills is particularly valuable, as these are often overlooked. The argumentation is solid, though some claims about model capabilities are based on benchmarks and personal observation rather than detailed evidence.

Scientific Rigor, Source Quality, Title Accuracy

The speaker demonstrates scientific rigor by referencing specific tools and practices, such as Claude Code, skills, and MCP. However, he does not provide detailed citations for benchmarks or claims about model performance. The title accurately reflects the content, which is a list of tips. The talk is well-organized and the advice is grounded in practical experience.

139 words

Title / Content Match

The title accurately reflects the content, which offers a set of practical tips for improving AI-assisted coding.

Quality & Reliability

8/10

The speaker is an experienced developer and Anthropic ambassador, providing practical, experience-based advice. Claims are generally supported by references to tools and practices, but some statements (e.g., model capabilities) are based on personal observation and benchmarks without detailed citations.

Key Moments

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

Cited Sources

NDC AI Conference — Conference where the talk was given
NDC Conferences — Organizer of the conference

Concurring Sources

Model Context Protocol (MCP) — Standard for connecting AI to external tools, mentioned in the talk

Contribution & Novelties

The talk offers a comprehensive set of practical tips for AI-assisted coding, with a strong emphasis on context management and skills. It introduces the concept of compound engineering and encourages a proactive mindset. The speaker’s experience as an Anthropic ambassador provides unique insights into Claude Code’s capabilities.

Pour aller plus loin :

Model Context Protocol (MCP) — Official documentation for MCP, a standard for connecting AI to external tools.
Anthropic’s Claude Code documentation — Official guide to Claude Code, including features like skills and context management.
The concept of ‘compound engineering’ — Article by Every that coined the term, explaining how AI can compound learning.
Skills documentation — Official documentation on creating and using skills in Claude Code.

118 words

Radar Profile

The radar profile shows high scores in quantity and quality of information, with a moderate technical level. The speaker provides a wealth of practical advice, but the technical depth is not extremely high, making it accessible to a broad audience. The overall reliability is strong, given the speaker's expertise.

Reliability 8/10