Half Stack Data Science S05E01 - Will GenAI kill software?
Mar 16, 2026
A lively debate about whether generative AI will swamp software with output and reduce signal in organizations. Discussion of AI’s impact on education, assessment, and how juniors can develop judgment when tools do the work. Conversation about cognitive debt from AI-written code and why human oversight, customer contact, and consequences still matter for product quality.
AI Snips
Chapters
Books
Transcript
Episode notes
AI Adds An Asynchronous Middle Manager
- Generative AI is creating a new layer of asynchronous productivity that can both produce and consume work on our behalf.
- Shaun and David note this risks toxic productivity and an explosion of low-signal slide decks and inter-office communications unless organisations rethink decision formats.
Prompt AI With Audience And Detail Requirements
- Use AI to draft targeted technical communication by specifying audience and level of detail rather than to multiply low-value outputs.
- Shaun suggests AI is useful for first drafts of technical slides when you provide audience context to avoid slop.
Grade The Process Not The Polished Output
- Education and assessment must shift from grading final polished outputs to evaluating human process and provenance.
- David and Shaun argue grading the research process, logbooks, or supervised exams preserves learning when AI can produce similar end products.
