llm/8d7ff4a1-085d-4b84-a357-52acc95c6355/topic-9-fde81546-0afb-4a3e-949e-964b664456c4-input.json
The following is content for you to summarize. Do not respond to the comments—summarize them. <topic> On-Device ML Classification # Suggestions for lightweight on-device machine learning to classify promotional vs transactional notifications, requests for smarter filtering beyond regex patterns </topic> <comments_about_topic> 1. Love the on-device approach. The fact that it never phones home is a huge differentiator — most "utility" apps these days are just data collection with a feature attached. The regex filtering is clever. Have you thought about adding ML-based classification for notifications that are harder to catch with patterns? Something lightweight like a small on-device model could detect promotional vs. transactional notifications without needing manual rules. Also curious about battery impact — how often does it process the notification stream? 2. > Have you thought about adding ML-based classification for notifications that are harder to catch with patterns? Honestly that's a little out my league. The idea did occur to me, but I'm discouraged by the amount of compute required for most ML. > Also curious about battery impact — how often does it process the notification stream? The OS sends any new notification to the app (it is a push based approach) automatically. On my own phone, this app currently shows at the bottom of the list in battery usage (<1%). 3. yes! I looked into implementing adblock on the iPhone notification tray and it didn't look like it was possible. Glad someone is working on it for android. Apps shouldn't be allowed to send notifications for Ads! I give any app on my phone one chance to be annoying and then turn them off. This feels like something where we should be able to use an on device classifier or even LLM to bucket notifications, similar to a spam inbox. Even better if they can pull any potential coupons out for use later without flavor text from the notification itself. 4. LLM would help with this immensely, if only it was allowed (not sure how though... make the ruleset available as a single text field for export/import, maybe?) 5. It would be neat to be able to do LLM filtered notifications. Perhaps with a local LLM for users that prefer. I hope that Apple does a better job of this too! I don't want Uber's ad notifications, but I do want their notifications about my vehicle status. </comments_about_topic> Write a concise, engaging paragraph (3-5 sentences) summarizing the key points and perspectives in these comments about the topic. Focus on the most interesting viewpoints. Do not use bullet points—write flowing prose.
On-Device ML Classification # Suggestions for lightweight on-device machine learning to classify promotional vs transactional notifications, requests for smarter filtering beyond regex patterns
5