When you spot a problem, build the fix — that’s the spirit behind AI @ Shippo. This week, we hit pause on business as usual — no meetings, no distractions — and gave our teams dedicated time to experiment, test, and actually build with AI. We call it AI Week: a chance for anyone at Shippo to play, prototype, and create tools that make our work smarter and our customers’ lives easier. Matthias is a perfect example: he jumped in to build a working Reddit tracker app in just a couple of days — solving a real need for better insights, faster. It’s scrappy, creative, and exactly the kind of experimentation that keeps us moving forward. #AIatShippo #Innovation #BuildingWhatsNext
Finished day 2 of our AI week. Burned 7,799,374 OpenAI API tokens. 100+ Lovable credits are goodbye (how do I get more?). Invested 10+ hours. But I've managed to build an actual useful app with value for Shippo. The idea of my app is quite straightforward: List of Reddit post that mention us or our competitors. I got this working within 10 mins. But there was a problem: Reddit is full of spam and random posts. Search for "Shippo" and most posts are not relevant at all. So I implemented AI to score all posts on relevancy with a bunch of custom criteria. I also added sentiment, categories and some other filters. I also had to implement databases (Supabase - Lovable makes it easy) and implement a user / auth logic. This magic continued and everything was working plus looking good. Just the scoring was way off. Too much spam got through. I was eager to fix it.... but I ran out of Lovable credits. So I learned how to move the project over to Cursor and continued to vibe code there. Initially I made a ton of progress, but after a while I felt Cursor was again breaking things faster than it would fix it. While it added a bunch of features, most of them didn't work properly. The AI scoring stopped working. Posts weren't being pushed to the database tables. Obviously, I have no idea about coding so I had no way to fix it beyond yelling at Cursor "it's still broken, please fix it!" and watch another 500k go down the pipe with all tables staying empty. I waited until midnight UTC to get my daily new 5 Lovable credits. With those I could fix the bugs implemented by Cursor plus get all the new features to work. The magic continued. By now I really love Lovable. And the app I built is actually useful already: It highlighted for example how upset users are about the recent price changes of our competitor. It showed posts where our CS team could jump in and help. It's overall just an easy way to see what real users are saying. Next steps: - find a way to get more Lovable credits (Anton Osika Elena Verna) - fix bugs, fine-tune scoring, double check all filters - have it create a daily summary of the most relevant posts - push summary to Slack - change LinkedIn job title to "Developer" I can't share the URL this time because I'm scared someone might find my OpenAI API key and burn through my $10 I put on there. Might turn this into an actual app one day though. But this is great. Being able to focus on AI without meetings or other distractions is amazing. Having a ton of fun and it's amazing to see what other people in the company are building. Go tell your CEO to also implement an AI week in your company. It's worth it. PS: I'm on an old Lovable Pro plan giving me 100 credits / month. Same for Cursor. Got both for a year for basically free. So I was able to play around with those tools quite a bit without paying.