Product build

Everything-Bot
A fact-checker that lived in the group chat

Solo build AWS Lambda Node.js S3 OpenAI 2025

You said "robot" in the group chat and it answered. Ask it to check a claim and it ran a web search and linked the sources. Drop in a screenshot and it read the text off the image. Lose a message somewhere up the thread and it found it. There was no server to keep alive, so an idle chat cost nothing. I built it solo for the AWS Lambda Hackathon.

The Everything Bot landing page: a bold headline reading Your AI Agent That Keeps Team Conversations Flowing, listing web search, audio transcription, media analysis and memory, beside a phone mockup of the bot in a group chat.
The Everything-Bot landing page, from the hackathon submission.

The problem

Group chats run on claims nobody checks. The answer is one tab away and nobody opens it, so the wrong thing just sits there and becomes true by repetition. The usual fix is a bot, and a bot usually means a box you rent around the clock that bills you whether anyone is talking to it or not.

The build

There was no box. Every message hit an API Gateway webhook, woke a Lambda function, did its job, and went back to sleep. A fact-check ran a web search and came back with sources. Images went through OCR so it could read the text in a screenshot. Voice notes got transcribed. You could search the chat history in plain English, and it had a few personas to switch between. FFmpeg rode along as a Lambda layer to handle the media.

The whole thing was about five thousand lines, packed into seventy-four megabytes, and it woke up in under three seconds. When a chat got loud, the serverless setup just ran more copies. When it went quiet, the bill went to zero.

The outcome

It worked, it scaled when a chat got busy, and it sat at zero when nobody used it. The features were the easy half. What took real work was keeping a cheap model in character and inside its budget, run after run. The bot that replaced it, letmecheckbot, works the same way and is still running every day.