Stenobird

Episode

How I lost my (old) job to AI

Podcast
Go Time: Golang, Software Engineering
Published
Sep 18, 2024
Duration seconds
4704
Processing state
processed
Canonical source
https://changelog.com/gotime/331
Audio
https://op3.dev/e/https://cdn.changelog.com/uploads/gotime/331/go-time-331.mp3
JSON
/v1/public/podcasts/go-time-golang-software-engineering/episodes/how-i-lost-my-old-job-to-ai
Markdown
/podcast/go-time-golang-software-engineering/how-i-lost-my-old-job-to-ai.md

Actions

  • POST https://stenobird.com/v1/public/podcasts/go-time-golang-software-engineering/episodes/how-i-lost-my-old-job-to-ai/transcription-requests
    Idempotently request low-priority transcript generation for this episode.
  • GET https://stenobird.com/podcast/go-time-golang-software-engineering/how-i-lost-my-old-job-to-ai.md
    Read the agent-friendly Markdown representation of this episode resource.

Summary

Veteran engineers debate the actual impact of AI on software engineering, arguing that while coding assistance is improving, the complexity of long-term maintenance remains a human-centric challenge. The discussion explores the limits of LLM reasoning and the potential for a future where engineers focus more on oversight than raw implementation.

Topics

  • Software Engineering
  • Artificial Intelligence
  • Cloud Development Environments
  • Code Maintenance
  • LLMs
  • Platform Engineering
  • Developer Productivity
  • Tech Industry Trends

Highlights

  • Main idea: AI excels at boilerplate and testing but struggles with complex, multi-file logic and deep architectural reasoning
  • Practical takeaway: Use AI for unit tests and tab-completion, but maintain manual oversight for critical business logic
  • Failure mode: Relying on AI for complex functions can lead to 'garbage' code that requires more time to debug than writing from scratch
  • Industry insight: The demand for engineers to maintain AI-generated code may actually increase short-term workload
  • Critical perspective: The 'death of the software engineer' narrative ignores the immense difficulty of maintaining legacy systems and complex dependencies

Chapters

  1. 1:00 Understanding Cloud Development Environments: An explanation of Coder and how cloud-based development environments (CDEs) solve infrastructure and dependency issues for platform engineers.
  2. 13:10 The Limits of AI Reasoning: A discussion on whether AI has reached its practical limits and the concerns regarding training data and copyright.
  3. 36:20 The Prompt Engineering Struggle: A firsthand account of trying to use detailed prompts to fix broken AI-generated functions and the resulting failure.
  4. 48:10 The Maintenance Burden: Why the need for senior engineers persists due to the long-term necessity of making software human-readable and maintainable.
  5. 54:00 The Future of Software Engineering: Speculating on a future where engineers transition from writing code to maintaining and servicing AI-generated software.
  6. 1:00:05 The Power of IDE Refactoring: Reflecting on how traditional IDE features like robust search and refactoring still outperform current AI-driven workflows.
  7. 1:11:50 The Hype Cycle and Job Markets: A closing discussion on the volatility of tech hype and the suspicious patterns in modern job postings.