Check the acquisition promise
Read the store headline, screenshots and description. What problem is the app promising to solve?
Study the storefront ↗Tianqi Chen · iOS
Explore the ad evidence, study the storefront, and build your next acquisition hypothesis.
01 / Creative research
Read the store headline, screenshots and description. What problem is the app promising to solve?
Study the storefront ↗Explore apps with direct store-link matches. Compare their hooks and formats without assuming the same approach works here.
Browse observed advertisers ↗Store rankings and revenue estimates describe market context. They do not establish ad spend, conversions or ROAS.
Review the market context ↗02 / Storefront & product
Study the product story and store experience behind the acquisition funnel.
03 / Market context
Keep revenue, downloads and store rankings separate from advertising performance.
I love that it’s completely offline, with local inference and that there’s no subscription. But, it’d be a lot cooler with two additional features: (a) history of conversations and (b) ability to use the LLMs in iOS shortcuts
I was disappointed to see that I could not save a conversation other than copying one response at time. When I put app to background then returned the conversation closed and I got back to main screen. When going to conversation screen I find the app was reset and my earlier data is not present. Still, I’m impressed to have the chance to play with models on my phone.
Definitely not as good as GPT4 models, but it runs offline locally, and it’s free. Pretty amazing. *feature request* it would be really nice if I could delete models that I’ve added. The Huggingface site for MLC AI doesn’t explain the vram requirements per model, so my current workaround is to delete the app and all data then reinstall it every time I try a new model. So far it looks like only the preinstalled model is compatible with iPad mini 6 and iPhone 13 mini based on requested VRAM requirements, but reinstalling a 1.6GB app every time seems a little unnecessary. Would also be nice if we could change the settings per model for things like repetition penalties, temperature, etc. Finally, a how to for how to import any other LLM (or recompile them for use if necessary) would be really awesome.
MLC Chat lets users chat with open language models locally on ipads and iphones. After a model is downloaded to the app, everything runs locally witho...