How to Accidentally Destroy a Codebase (A Step-by-Step Guide) - Mariami Tsotsolashvili

Value of the Information & Strength of the Argument

The talk’s value lies in its practical, live demonstration of how seemingly reasonable choices accumulate into a tangled, hard-to-maintain codebase. The speaker effectively illustrates each anti-pattern with real code, showing the step-by-step rationale that makes each decision appear logical at the time. Her argumentation is solid: she contrasts the initial clean state with the final bloated state, highlighting the increase in files, lines of code, and the emergence of subtle bugs. The alternatives she proposes are pragmatic and grounded in well-known engineering principles, such as the rule of three and the avoidance of premature abstraction. She also emphasizes the importance of measuring before optimizing and addressing root causes rather than symptoms. The presentation is engaging and humorous, which helps reinforce the message without sacrificing technical depth.

Scientific Rigor, Source Quality, Title Accuracy

The talk does not cite formal academic sources, relying instead on the speaker’s professional experience and widely recognized software engineering practices. The NDC conference platform adds credibility, and the two provided links in the description (ndcconferences.com and ndcoslo.com) are conference websites, not specific references. The title accurately represents the content: it is literally a step-by-step guide to destroying a codebase, delivered as a live coding session. The speaker’s claims about common pitfalls align with established anti-pattern literature, though no explicit external references are given. The title-contenu alignment is strong, and the talk is internally coherent and well-structured.

The title accurately reflects the content: the speaker systematically destroys a codebase step by step with seemingly reasonable decisions, exactly as promised.

8/10

The speaker is a senior software engineer and tech lead with practical experience. The talk is based on common anti-patterns and provides reasoned alternatives. However, it relies on anecdotal evidence and personal experience rather than formal research.

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

• 02:00Introduction of the todo app and its healthy architecture (signals, service, 120 lines).
• 04:00First mistake: quick fix for urgent tasks using a magic string, not added to the model.
• 07:00Second mistake: premature abstraction with a generic CRUD class, applied with only two services.
• 10:00Third mistake: caching without invalidation, leading to extra logic and no real benefit.
• 13:00Fourth mistake: manual change detection to 'fix' performance, creating more overhead.
• 16:00Fifth mistake: boolean flag explosion, causing impossible states and bugs.
• 19:00Sixth mistake: over-engineered layers (domain, application, infrastructure) for a simple app.
• 22:00Autopsy: comparing before/after metrics (14 files, 300+ lines) and the rewind begins.
• 25:00Alternatives: rule of three, proper types, trusting framework, discriminated unions, and layer necessity.
• 28:00Final checklist of five questions to ask before making a change, and concluding remarks.

• NDC Conferences — Organizers of the event where the talk was given, mentioned in the description.
• NDC Oslo — Specific conference website, referenced in the description as the event location.

• ndcoslo.com