As of March 2024, AI (I’m referring to LLMs like ChatGPT when I say AI) have both come a long way and... not really.
I don’t believe we’re even close to AI replacing software engineers in building entire features and product lines. Mainly because:
- Most code requires knowledge across a large domain of the product and how features affect each other — both of which AI in its current state doesn't do well.
- In every successful attempt at AI writing production code that I've seen, the code has been isolated in a black box (a single function, specific web pages, etc…).
- Even in isolated code, my experience is that AI tends to “hallucinate” and write code that is 80% there (and code needs to 100% work).
So what can AI do for engineering teams?
I believe we’re at a stage where AI is really helpful with making the developer experience faster & more efficient. Here’s how we leverage AI at Distru's engineering team:
Github Copilot
This is a no brainer. Copilot’s ability to auto-complete code I’m about to write has increased my speed by ~20%. I can also work on problems for longer because I don’t feel burnt out by writing as much boilerplate code.
ChatGPT for upgrading React Components
This was a brilliant idea my Director of Engineering came up with. Take old class based React components and feed them into ChatGPT to rewrite them into functional components!
ChatGPT for syntax error hunting
Missing a closing brace but not sure where? Pop it into ChatGPT, and know exactly where the syntax error is.
ChatGPT for data parsing
When debugging issues, you often need to parse data out into more legible formats, such as:
- Beautifying JSON
- Finding diffs in 2 blobs of data
- Timestamp timezone conversion
- Timestamp formatting
ChatGPT for simple but non-intuitive scripting
Sometimes we need to write one off scripts such as:
- Google Sheets AppScript
- Generating CSR for SSL certs
- JS transmog (yuck)
Rather than re-learning how to do these things every time you need to, popping it into ChatGPT is much easier.
As you can see, these all help enhance the developer experience. But at the end of the day, you still need great software engineers to do 80% of the heavy lifting.
Here’s how to attract those great engineers (on a startup budget)!