Building Recap (Nov '22 - Jan '23)
Eight projects and a whole lot of fun
As my consulting workload has eased up since November, I’ve been working on quite a few things. Sequentially, they are:
Autodocs - generate Tailwind-style docs from Notion pages
GPT3 in Sheets - AI() custom function in Google Sheets to prompt GPT3
AI Bar - Quicksilver-like interface for GPT3
CodeGPT - VS Code extension to help you write code with the help of ChatGPT
aicommit - CLI tool that generates commit message suggestions
Flair - AI avatar generator
WhimsyAI - edit photos with text
WhimsyWorks (current project) - AI-powered personalized children’s books
I’m really proud of my output over the last 3 months. Getting in the habit of building things is hard but once you do it, it’s really great. Then, you can focus on actually buildings things that are useful for people.
Releasing things is even harder than building things. For aicommit, the code is very simple but as a Python noob, packaging it up so it you can easily `pip install aicommit` took me several days of work. Similarly, iOS apps require lots of non-coding work before they can be released: App Store screenshots, privacy policy, various other forms on App Store Connect.
Of the 8 projects, I never released 2 of them: Autodocs and AI Bar. Autodocs was not that engaging of a project. Competing with the likes of HelpKit would have taken months of work, and all for a relatively small market. Besides, AI is the transformational technology of our time, and I’d like to be a part of that revolution rather than stuck working on a boring SaaS product.
AI Bar was abandoned mostly because ChatGPT was released in early December and ChatGPT was so much better for most use cases. There might be room for AI Bar yet. Although Raycast extensions might fill that need quite well instead.
GPT3 in Sheets is the most popular thing I’ve built on that list. It’s an open source snippet of code. It seems extremely popular in SEO circles. This viral tweet uses my code snippet:
I spent some time thinking whether to productize GPT3 in Sheets but I wasn’t really sure how much user value I could add besides being a GPT3 wrapper. And SEO is not something I’m passionate about in the least bit. Shubhro, whose original tweet inspired me to build the Sheet script, did built a product and I’m curious to see how it does.
Building, releasing and watching users use these products has been a great learning experience. I’ll be writing a bit more about those learnings in future posts.
For now, I’m excited about the approach we’re taking with my current project, WhimsyWorks. We’re validating the user need first. And by validating, we’re not just talking to people, we’re getting them to spend $50 to buy a book. And we’re only building just enough to validate the hypothesis.
We are putting the user first and validating their needs before building a full product.
wow, so many projects!
you're on fire with this spree, looking forward to more, well done!