Neural Portal for the NeuroRoast Archive: Automation and Analytics
An interactive neural portal has been created for the NeuroRoast project archive, combining analytics, videos, and a web page into a single ecosystem.
The NeuroRoast project has been running for a year and a half. To systematize content, a special portal was developed that allows viewing dashboards and analytics. The creators used time series analysis algorithms to filter out negative reactions.
Automation of the process involved collecting data from the Telegram channel with the hashtag 'NeuroRoast'. An AI agent independently downloaded posts, extracted authors, and saved view metrics.
- Export of source text and metrics from the channel
- Automatic downloading of high-quality videos via ytdl
- Filtering old records without author attribution
To publish video content, Chrome's headless browser mode was used instead of standard YouTube APIs. This allowed bypassing file upload limits and processing 347 videos.
Videos were uploaded to the @SergeyTsyptsynPlus YouTube channel in chronological order. An automated task checks the channel for new materials, downloads them, and updates the database.
The project website is hosted and weighs only 237 kilobytes. It uses PHP 8.3 and plays videos directly from YouTube. All processes operate within the Google AI Pro subscription.
The portal is available at neuroroast.cgevent.ru. The project demonstrates the capabilities of autonomous AI agents in creating complex digital products without manual intervention.
"NeuroRoast" has turned into a cultural phenomenon where technology helps structure large volumes of user-generated content and make it accessible to a wide audience.








