So in Paid Community we have been diving into Multi Agent Systems with n8n and I remembered @Deleted once asking about email summaries for her client. Overview: So I have Multi Agent System that I interact via Telegram, but I don't want it to shoot me one email at a time, nor do I want to have a to-do-list growing with actionable emails before I determine that I want to take action on it. Solution: Connect two different automations into one chat interface: 1. Every morning I receive email digest from my two different Gmail inboxes. Emails are categorised, and prioritized based on what matters to me. I get to see also sender and subject line etc. 2. I can quickly get a glimpse what is going on in my inbox from last 24h 3. And if I need to take action, like draft replay: I send a message to the same chat that triggers a Multi Agent System, that can fetch exact email for me, and I can further instruct what to do: create a task, block time on a calendar, draft and email to a person, present to me, and if I approve - it sends. The image below is from the first scenario and it is a workflow. But I found it interesting to combine rigid workflows (there you need accuracy etc) with agents that take action - all in one interface. How many of you are playing with n8n, or would be interested to learn more? 👀

Andrej Karpathy is one of the original co-founders of OpenAI. And he has been one of my trusted sources for years now. Andrej just published a 3h video explaining how ChatGPT is build and how it works. Video is made for general public, it starts with explaining how data was collected, how raw data gets turned into tokens and what tokens are.. Won’t spoil it to you, but if you will watch this do share what has been a new information to you. P.S. in the beginning Andrej shows turning tokens into emojis which has been a big prompt hacking technique for a while to use emojis as secret tokens to get LLMs say what they should not.

Hi, internet friends! May i ask you a question? Please post in the comments where you come from. This way, we can see where all the members are coming from. Afterward, take a look through the comments and feel free to like or say "hi" to members who are from your area. I'm super curious to see where you're all from and will read every single comment. Let's see which continent is dominating in this community! GG 💃

Hey automation folks! Just got back from Xmas with the inlaws. Perfect moment to chill down and work on a project 😁 I need some help with understanding how I could best achieve what I want. JarvisJr got me thinking about all the options.. **Current Setup:** - Google Sheet with booking data for 3 rental properties - Data includes bookings, guest info, dates, pricing - Currently using Claude for analysis (works great with the Excel/CSV data when uploaded) **Goal:** I want to be able to ask simple questions like: - "Who's staying at property X on date Y?" - "What's the next available 6-day period?" - "Output the top 3 bookings for the last quarter in property X, in a table format" And have the AI analyze real-time data from my Google Sheet **Requirements:** - Must read live data from Google Sheets - Ideally accessible via chat (e.g., Telegram, like JarvisJr) - Should work with an LLM (Claude/GPT) for natural language processing **Current Thoughts:** - Make/Zapier for the Google Sheets connection - Some kind of API middleware - Maybe Airtable as an intermediate database? I already simply tested downloading it as xlsx, feeding that document to Claude, and asking the example questions. This worked perfectly. Would love to hear your experiences and suggestions! Has anyone built something similar? Any pitfalls to avoid?

Did Santa visit your home too? 🤫