ChatGPT provides an API for Fine-tuning that allows you to train it on your own data, so I used it to learn from my past tweets and created a bot that tweets in a style similar to mine. I wrote it in Deno, which has just had its v2 release, and it runs on Deno Deploy (Deno Cron).

I decided to leave Hatena Blog, which I had been using for about 10 years since 2014, and move to a self-made (big lie) blog system.

I created a web app for a "ato de yomu" (Japanese for “read later”) reading list.

I'm not sure if "reading list" is a common term, but it refers to the reading list feature that comes with Safari. The "ato de yomu" app has the following features:

• You can save web articles that you want to read later.
• You can keep a history of articles you have read.
• You can view and subscribe to other people's lists.

I created Editjiro, a clone app of Editaro, which is a text editor for writing drafts before posting to emails or chats.

I created a ChatGPT Slack bot using Slack's next-generation platform, which is in beta as of 2023. When you mention the bot, it calls the ChatGPT API and provides a response.

While many people have created bots with similar functionality, and you can find code on github.com, the one I made has the following features. Mine is the best!

• You can add the bot to any channel later, not just the channel specified with fixed values in the file.
• By saving conversation history in Slack Datastores, you can send a complete series of conversations to the ChatGPT API, allowing it to understand the context and respond appropriately.
• You can set the system message passed to the ChatGPT API for each channel where the bot operates, and you can change it later.