ChatGPT

For the most part, I have been quite skeptical about AI. Not the invisible AI when you get search or posts recommendations, or anything that we don’t immediately associate with A. Rather the explicit use-cases, like Chat bots for diagnosing diseases, or the likes of self-driving cars. The sort of products that we immediately associate with non-human intelligence. Even when Dall-E came out, even with some fantastic results, I continued to be skeptical. Or with Github’s CodePilot, that I use most days. It’s nice, but you can see flaws here and there. But also its limitations. As in, how much can this “entity” give me for free.

ChatGPT is to me the first tool that completely breaks the moulds of what interacting with AI is. These are examples of things that I have done with it, with remarkable results:

1. Improve this class written in Swift.
2. Write tests for this function in Swift.
3. Improve the performance of this SQL query.
4. Improve the performance of this Go snippet.
5. Write me an application for a tech company.
6. Write me a rejection letter for a candidate that did well, but wasn’t senior enough.
7. Give me a vegetarian recipe with the following ingredients: carrots, onion, quinoa, quorn, garlic and rice. It shouldn’t take more than 30 minutes to make.
8. Write me a tech roadmap for a SaaS B2B company.
9. Improve this paragraph using the same style as Hemingway.
10. Write me a witty, thought provoking tweet about React Native. You should maximise engagement over everything else.

Every output blew my expectations. Especially the first four items, since these weren’t trivial pieces. But beyond the output, also the explanation of what it did. Imagine having your CI running this, every time you open a PR. Providing suggestions to the code you just submitted. We are truly living in the future.