AI Agents / Automation
AI Daily News
An autonomous AI-powered news pipeline that researches, ranks, summarizes, evaluates, and publishes AI news every day — from source collection to final delivery.
Specific news from September 21
Showing image 1 of 4
01
AI is evolving very quickly, and new tools, models, and updates are released every day. I found myself spending a lot of time searching different sources to keep up with the latest AI news and deciding which stories were actually worth reading. I wanted a way to automatically find the most relevant AI news, summarize it, and deliver it to me every day without having to search manually.
02
AI Daily News runs a scheduled pipeline that handles the daily news cycle with no human intervention. News is collected from multiple APIs and processed through several stages. AI is used to rank and evaluate stories, generate summaries, and support content creation. Semantic similarity is used to identify duplicate coverage so that the same story isn't repeatedly published from different sources. LangGraph coordinates the multi-step AI workflow where state and controlled transitions are useful, while the surrounding application handles scheduling, storage, publishing, and notifications. Only content that passes the required processing and evaluation stages is published to the web application and send to the Telegram channel through Telegram bot.
Challenges & decisions
Automating the daily news pipeline
The system needed to run every day without requiring manual research, summarization, or publishing. The pipeline was designed to automatically move from news collection through processing and publishing on a schedule.
Handling duplicate news
The same AI story can appear across multiple news sources. Embeddings and semantic similarity were used to identify similar stories even when their titles and wording were different.
Selecting relevant stories
Not every collected story is equally relevant. An LLM-based ranking step helps select the most useful stories before they move through the summarization and publishing stages.
Adding an evaluation step
AI-generated summaries should not be published without checking their quality. An evaluation step was added before publishing to identify low-quality or off-topic results.
Orchestrating the AI workflow
Research, ranking, summarization, and evaluation needed to work as coordinated steps rather than a single large prompt. LangGraph was used to orchestrate the multi-step workflow and maintain its state.
03
AI Daily News automates the daily news cycle end-to-end using AI agents and workflows rather than manual curation. The interesting part of this project isn't the articles themselves — it's the pipeline that researches, filters, summarizes, evaluates, and publishes them on a schedule, with a web interface for reading and daily notifications for staying up to date.
How it works
A scheduled automation layer triggers the pipeline daily. Research agents gather source material on selected topics, summarization agents condense it, and an evaluation step checks the output before publishing — with LangGraph coordinating the multi-step agent workflow where stateful orchestration is required. Published content is served through a web interface, and a notification component alerts users when new daily content is available.
Key features
- Agent-driven research workflow for gathering source material
- Automated summarization of researched content
- Evaluation step to check output quality before publishing
- LangGraph-based orchestration for multi-step agent workflows
- Scheduled, automated publishing pipeline
- Web interface for reading published content
- Daily notifications when new content is published
04
Technologies & Deployment
- AI Agents
- LangGraph
- LLMs
- Python
A fully automated daily pipeline from news collection to publishing Automated AI story ranking, deduplication, summarization, and evaluation A working web interface for reading the published AI news Automated image generation for published stories Daily Telegram notifications when new content is published