radurevutchi

justin - helicone's great. I login in ~10 times a day. Great for debugging - I'm mostly using for analytics / observability.

WhatsApp is opened everyday by hundreds of millions of people. It's already installed on every-phone. The fatigue to open up a frequently used app is smaller than a rarely used app.

1. Cannot protect against meta doing this as well. Can only hope to make my service better than theirs. 2. Using the Meta API with a business account.

For those who don't want to install yet another app, I built a WhatsApp AI assistant to handle your daily research and manage your Google Calendar, lists, reminders.

Some challenging tasks Coco AI can complete instantly:

"Remind me to take vitamin D3 every afternoon until March" "Get child-friendly events in Dublin new years week, add to family calendar" "Find my grocery list and send my husband a reminder about it in 2 hours" "Find the next sunny day in SF and add beach day to calendar" "Add client lunch to the next available free slot on my calendar" "I found a house, remove ALL upcoming house tour events"

The Tech Meta API forwards incoming message to our webhook hosted on render.com w/ Docker. Incoming request converts to a message in our DB and we call our Agent loop w/ multiple tool calls (calendar read/write, reminders read/write, lists read/write, internet search, etc). Using GPT hosted on Azure / OpenAI. Google Calendar API for calendar event access. Practically everything is async. Postgres for DB, Next.js for frontend, FastAPI backend.

I quit my job for this in October. Would love all of your critical feedback, suggestions, and any questions!

I started to work on Coco in August after seeing how I can improve "text yourself" productivity hacks. Ended up expanding/pivoting to an AI executive assistant.

I text myself a lot many others do too for quick thoughts, reminders, to-do lists, and other 'brain-dump' items. While there's a few text yourself apps out there, Coco is the only one that doesn't require any new apps or logging into a website. It's a freemium plan (unlimited free reminders, other features behind stripe subscription) - this is needed to cover per user GPT & WhatsApp template message costs.

Features: * WhatsApp-native reminders * Sending reminders to others * Creating and managing lists (read/write/modify) * Google Calendar manager (read/write/modify) * Image / voice message analysis * chatGPT-style ask questions about anything * Built-in memory (search over your notes and messages from any period)

Tech stack: * Hosting on Render w/ Docker + Postgres + FastAPI * Next.JS frontend * Meta API * Helicone for LLM requests tracking

I'm pushing for as much feedback, let me know what you think!

Let me know if you have feedback!

We're keeping prices low for now @ $17 cheaper than monthly ChatGPT subscription. Unfortunately ChatGPT does not allow setting reminders + integrating with calendar. Any other apps on your mind?

Account settings are accessed via text (and soon dashboard w/ auto login). Receipts are done over email!

I started to work on Coco in August after seeing how I can improve "text yourself" productivity hacks with my NLP experience.

I text myself a lot and according to reddit, some articles, and my friends, many do the same for quick thoughts, reminders, to-do lists, and other quick 'brain-dump' items. I decided, let's have an assistant to text you back and actually remind you.

While there's a few text yourself apps out there, Coco is the only one that doesn't require any new apps or logging into a website. It's a freemium plan (unlimited free reminders, other features behind stripe subscription) - this is needed to cover per user GPT & WhatsApp template message costs.

Features:

* WhatsApp-native reminders / repeat reminders * Creating and managing lists * Google Calendar access (read, write, modify, delete) * Image / Voice note analysis * (of course) chatGPT-style ask questions about anything * (coming soon) send reminders to others * (coming soon) note-taking and searching over your chat history

Tech stack:

* Hosting on Render w/ Docker + Postgres + FastAPI * Meta API * Helicone for LLM requests tracking

I'm pushing for as much feedback, let me know what you think :)