Bryan Hunter

This feature is pretty powerful. We give our agent extra compute time and higher reasoning because it runs in the background and needs to research deeper because it needs to deliver results without a follow up. I get to dogfood Basedash with my side project RYSE (a social fitness app). I have a daily report I generate to monitor all Posts and Comments for offensive language, bugs, and things...

We don't just rely on the models themselves. There's a ton of context engineering and optimizations baked into every AI call in Basedash - which we test and validate thoroughly with our eval system (which is constantly growing as we run into crazy use cases) We use a variety of techniques like: - RAG - algorithmic schema context building - conversation summarization - learning from existing...

You know you're starting to hit PMF when you need to self-host for customers. WOOOOO

One nice benefit for me in building this new product is that I got to dogfood it extensively by replacing/consolidating my Mixpanel, AppStore, and Quicksight dashboards for my other product (RYSE). I have 4 pretty permanent dashboards now and make a lot of adhoc dashboards to track new features: - All time app-wide metrics (# users, # workouts, device type, gender distribution, app settings...

Filled out this survey last year as a developer and have since joined the Basedash team. The attention to detail on how the survey was built and how the results were statistically put together was next level. Super happy Basedash can give back to the dev community and showcase the tools/products that people actually use and love

As a developer, this looks really dope. Nice low effort way of letting someone manually help out with the product without giving them full access. Will check it out once DynamoDB is supported for sure :) One thing I was thinking during the video - is there a changelog of database updates by user & time? I imagine if someone accidentally messed something up horrifically that'd be the fastest way...