2022: Week 49

Automated deployment, a minor feature addition, and ChatGPT .

Deployment

Since day one, pretty much, pushing to the repos triggers build pipelines for Minmail’s applications, but I’d avoided automating the actual deployment. Up until just recently I would run each deployment command manually (there aren’t many). I eventually put them into a bash script, which still had to be executed manually.

This week I finally got around/decided to connect deployment to the build process. I typically try to automate anything that can be automated, but I was feeling extra cautious about Minmail deployments. In the early days there was a greater probability of something breaking that I’d need to fix immediately to avoid interruption to my inbound emails. That’s not been the case for months now, though.🙄

Pinned threads

Search is usually sufficient to find emails you need, but sometimes you need a message to remain “in your face”, so to speak. Starring emails is okay, but when more come in, starred threads get pushed out of sight. This is the case in most email services.

Now we can

pin messages to the top of the Minbox for easy access, and as a constant reminder there’s something that needs taking care of.

It works well, but I found that I kept having to scroll past pinned messages to get to new, unpinned messages. I added a button to fix that.

It’s all about iteration. No doubt I’ll make more changes to its appearance and functionality.

ChatGPT

I spent too much time with OpenAI and ChatGPT this past week, like many others, I’m sure. Some things I discussed with it:

My experiments with it taught me a few things about interacting with OpenAI’s models.

Specificity

Perhaps it should go without saying, but if you need a specific output, you need to provide specific input. Be descriptive about the result you want. If you can, provide an example or two.

Collaboration

It’s impressive how receptive ChatGPT is to critique and suggestions — use that to your advantage! It maintains the context of the conversation, so it’s possible to have a bit of a back-and-forth with ChatGPT while you help it reach the desired output. Work with it.

Drilling

When you don’t know the answer (e.g. you’re looking for ideas and inspiration), start broad then drill down. ChatGPT will easily stream out plenty of suggestions around a topic. You can then feed these back into it to expand on each point of interest. This is a great way to quickly extract a lot of information.

Caution

ChatGPT’s responses can be very convincing, but I found on a few code-related occasions that it would respond with answers that looked plausible but that I knew were flat out wrong. I’d point out its error and it would apologise and correct itself. My exchanges with it made it clear that it could also be gaslit quite easily. As always with secondhand information, verify!

Quick view on ChatGPT

ChatGPT, and OpenAI’s models in general, are a game-changer. If you haven’t, I strongly suggest you try their playground . ChatGPT won’t take any jobs yet, but in the right hands it can greatly reduce the need to hire for some more entry level work. Need some basic copy or to create a tiny bit of functionality? The right series of prompts can get you the results you need in a fraction of the time and at potentially zero cost. It’s exciting and terrifying.

I’m fascinated to see where it goes and watching the space closely. In the meantime, I’m already incorporating it into my workflows. You should be doing the same, no matter what line of work you’re in.

The week ahead

There are some minor things in the UI that I’ve been putting off for too long. I’m hoping addressing them will improve usability .

Stay tuned.