What Claude Code Can Do That You Haven't Tried - Gui Ferreira - NDC Oslo 2026

Value of the Information & Strength of the Argument

The talk provides substantial practical value for developers looking to extract more from Claude Code. The speaker’s examples are concrete and based on real daily usage, making the advice actionable. The argumentation is coherent and structured, with a clear progression from simple setup to advanced workflows. He supports some claims with a reference to an ETH Zurich study on CLAUDE.md performance, though the citation is brief and lacks specifics. The inclusion of security measures (permissions deny) and cost awareness adds credibility. The main strength lies in the breadth of tips and the emphasis on workflow efficiency, though the lack of formal validation for many of the ‘best practices’ limits the scientific rigor. Overall, the value is high for its target audience of experienced Claude Code users.

Scientific Rigor, Source Quality, Title Accuracy

The speaker is transparent about his non-affiliation with Anthropic, positioning his advice as personal experience. He references an ETH Zurich study but does not provide a direct link or detailed methodology, which weakens the source quality. The conference links in the description are generic and do not point to specific referenced materials. The title is well-aligned with the content, as the talk indeed covers features that go beyond the basics. However, the title’s promise of ‘What Claude Code Can Do That You Haven’t Tried’ is partially fulfilled, as many tips are advanced but some may be known to seasoned users. The lack of external citations and reliance on anecdotal evidence places this talk in the realm of opinion expertise rather than rigorously researched technical writing.

The title accurately reflects the content: the talk explores advanced and less-known features of Claude Code beyond basic usage.

7/10

Speaker is a practitioner with daily hands-on experience, not affiliated with Anthropic. Claims are based on personal usage and observations, with only one reference to a study (ETH Zurich) without detailed citation. The talk is opinionated and practical, but lacks rigorous scientific backing.

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

• 03:00Introduction to the talk's structure and the four pillars: one-time setup, daily workflow, customization, and non-coding scenarios.
• 07:00Discussion on optimizing CLAUDE.md, including the /import command, avoiding duplication, and the ETH Zurich study on the negative impact of lengthy CLAUDE.md files.
• 15:00Explanation of daily workflow features: plan mode with Ctrl+G, using /ask to interview Claude, and the rewind feature for stepping back in the conversation.
• 30:00Coverage of model selection and effort levels, including /model, ultra think, and the stash feature (Ctrl+S) for managing prompts during long tasks.
• 45:00Discussion on cost tracking, context window management, compaction, and memory auditing, followed by a brief look at using Claude Code beyond software development.

• NDC Conferences — Official conference website where the talk was recorded and future NDC events are announced.
• NDC Oslo — Specific page for the NDC Oslo conference where this talk took place.