This is what I'm working towards: Work is a hobby and I do it because it fulfills me and brings satisfaction. It is one of many fulfilling activities that I do often. I have the freedom to work when, where, and how I want. I've got a ways to go but I'm getting closer every day.

Calorik is so accurate at tracking meals with one photo, it got my lunch accurate to 6 calories according to the recipe card

Beginners: try this 7-day micro-habit. Snap your meal in Calorik. That’s it. Day 1 starts today. calorik.ai/?utm_source=twitt…

This is hilarious and crazy and awesome, but also terrifying to think of maintaining something like this. Don't think we're all built to run projects this way but more power to him. Just goes to say that it doesn't need to be "elegant", it just needs to solve the problem.

RPA = Robotic Process Automation How did browser automation tools get such a cool name?

Is @manusai the right tool for something like this, or is there something better out there?

I tried using @manusai to automate cold outreach via a directory site for dietitians and nutritionists. I could do this manually but it would be very tedious as the site doesn't list email addresses, it only provides forms for contacting the individuals. A perfect test case for @manusai! Result: 😞 🟩 The good: even though it was pretty slow (I am on the Starter plan, not sure if this improves on the Pro plan), it actually did follow instructions fairly well to navigate the site and fill in the form fields including a lightly personalized message. 🟥 The bad: Manus sucks at captchas. It didn't reliably know which tiles to click to solve the captcha, or when it should keep clicking or submit the puzzle. I was able to tell it to stop and wait for instructions on which tiles to click, but this was even more tedious than doing it myself 😫 It has to think for like 10 seconds before each click and the directory site requires a captcha on each form submission so there was no escaping it. I pivoted my approach and am having it compile data on the listings instead, and will look at using a browser automation tool to try to automate as much of the copy/pasting as possible. Will still likely have to solve the captchas myself but hopefully that's the bulk of it! Will report back later with results!

Just started using @TryAstroApp to analyze app store keywords for @calorik_ai and @shiver_app. Much more straightforward (and much less expensive!) than other ASO tools I've played with. We've got a lot of work to do!

How important is the website for a mobile app? Asking for a friend. 🫣

We were analyzing our onboarding flow for @calorik_ai and noticed a high dropoff rate (>10%) on what we call the "Taste Test" screen. We added this so users can see the functionality in action before they hit the paywall. Apparently many users don't want this? One hypothesis is that some users aren't eating at that moment or don't have a specific meal they want to track, so they exit the app. They may or may not go back later (hard to know this). Going to try changing the placeholder text to see if that encourages users to keep going, even if they need to describe a fake meal. Also considering allowing users to skip this, or even the whole onboarding flow. Still weighing if that's a good idea; we're not sure if users will pay or trial without seeing the functionality first.

Haven't tried Claude Code. Used Claude in Cursor, and use ChatGPT regularly. Am I missing out?

Vibe coding can be super fun and satisfying! We recently released a custom camera view for @calorik_ai that streamlines capturing meal photos. No features we don't need (zoom, retake) and has an integrated photo picker. I had no idea how to do this in SwiftUI so I did the whole thing iterating in Cursor. Don't think I wrote a single line of code by hand. I tried a few different models and Gemini 2.5 Pro seemed to work the best for adding net new functionality. I actually quite enjoyed it! Have been using AI while coding for a while now but not sure I had used it to implement an entire feature before. The catch is that AI will structure the code in weird ways sometimes, especially after iterating for a while, and you need to guide it to simplify and refactor. Sometimes in very specific ways. And when it refactors it commonly breaks things that were working before, so you have to keep testing and iterating along the way. Coding with SwiftUI in Xcode, this testing can be very tedious as there is no hot-reloading--each small change triggers a full build. With web dev I'm sure the iterations would be much quicker. So it is important to at least have a foundation of coding to make something that not only works but is production deployable, meaning it is less likely to break for unknown reasons and can be reasonably augmented with new functionality. Without this foundation, I imagine it could be very frustrating, maybe not as much for an initial MVP but definitely if you're trying to maintain an app long term. But in this case it sped up my dev time at least 5-10x, and I don't feel I compromised on quality (at least compared to what I personally could produce). Vibe coding FTW!

#13 💯 Some time ago I realized that most of the entrepreneurs (and other wildly successful people) our society idolizes have made sacrifices to achieve their success that may not be worth making. Be careful who you emulate!