AI took my job but it didn't take my passion for writing software

Source: Dev.to

Recent Package Updates

Over the past 170 hours, I updated several Go packages with the assistance of AI. Each update involved thorough code reviews, security audits, and extensive testing.

bump

• Fixed bugs that were missed by the original test coverage.
• Added new tests covering additional scenarios, including an integration test linked to existing unit, fuzz, and benchmark tests.
• Introduced a fourth test type via test.sh to catch real‑world bugs that traditional testing misses.
  GitHub Repository

verbose

• Added an Apache 2.0 license.
• Expanded test suite and improved efficiency of function calls.
• Implemented a pipeline‑style API for handling sensitive information, with selective output to STDOUT or protected files.
  GitHub Repository

goenv

• Cleaned up the Makefile by removing the summary task from each test.
• Extended the README.md documentation.
  GitHub Repository

entpassgen

• Performed a comprehensive review, adding stability improvements to the loading spinner.
• Added dozens of new passing tests.
  GitHub Repository

lemmings

• Built a brand‑new utility in less than a week with AI assistance.
• Created to fill a gap that existed since 2015 when I launched Trakify.
  GitHub Repository

genwordpass

• Updated the word list, improved core performance, and refreshed the README.md.
  GitHub Repository

room

• Developed a “waiting room” package to protect high‑traffic sites during traffic spikes, reducing the need for immediate scaling.
  GitHub Repository

sema

• Added Try* functions to the semaphore primitive.
• Expanded test coverage and functionality, making this my go‑to semaphore package.
  GitHub Repository

AI and Productivity

AI has dramatically increased my weekly productivity—by at least tenfold. My workflow is iterative:

1. Scaffold: Use AI to generate rough code.
2. Refine: Manually patch, review, and improve the scaffolded code.
3. Iterate: Hand the refined code back to the AI for further development.

The AI does not produce a finished product in a single pass, but my ability to guide and augment its output enables rapid development.

Reflections on Title and Resources

I no longer feel that “software engineer” fully describes my work; I build solutions using software as a tool. My professional identity has always been flexible regarding titles.

With a modest $20/month Claude Pro subscription (and occasional local LLM usage during cooldown periods), I can accomplish tasks that would otherwise require a far larger budget. A hypothetical $100K AI budget per month would allow me to tackle even more complex problems, but the current setup already empowers me to deliver functional software within a week.