Seven months using ChatGPT 4

I started using ChatGPT Plus¹ seven months ago, and I'm surprised at how much I rely on it for day-to-day tasks. I don't need to Google as much as before, as answers tend to solve my immediate needs.

Here are a few use cases I found extremely useful.

• A custom TypeScript & React GPT for work only allowed to answer with code solutions.
• English-to-Spanish Markdown blog post translation. I only have to apply minor corrections, often because ChatGPT mixes Castilian and American Spanish.
• Write Python and bash scripts, both simple and fairly complex.
• Learning and research. ChatGPT is an avid searcher of related papers and a great explainer of machine learning concepts. (You can even ask it to draft specific neural network architectures!)
• Fix grammar and punctuation.
• Troubleshoot computer errors.

I find it mesmerizing to generate visuals and images with it — I have several parallel side projects on this, which I haven't published yet. But that's a story for another day.

Ethical and safety concerns aside, I find it quite promising that, relatively soon, we'll have models with similar performance to GPT-4 and DALL-E 3 running locally on our devices without a subscription.

I haven't measured how much more efficient as a software developer it's made me. Yet after hundreds of conversations ChatGPT has saved me some "human compute hours" and helped me develop several code solutions faster. In this case, twenty dollars a month is a no-brainer.

"We shape our tools and thereafter our tools shape us." Said John Culkin². Language models are rapidly changing how we interact with machines. The next step could be removing humans from the loop and delegating digital tasks to machines.