How the tech briefing on 1ar.io works

I built the briefing section on the 1ar.io homepage because I was tired of checking six websites every morning. The feed updates every 6 hours, pulls tech news from RSS sources I picked, and lets me generate a one-sentence summary of any article on demand. It is a personal tool that happens to be public.

The principle

The feed is shaped by three things: interests I care about, context from past projects and work I have done, and feedback signals over time. That last part matters - the system gets better at filtering as it learns what I actually read versus what I skip.

The RSS sources are hand-picked. They include corners of the internet that algorithmic feeds chronically underexpose - small research blogs, niche industry publications, independent writers. The kind of material that gets buried under engagement-optimized content on any platform that ranks by clicks.

Prediction markets as a second signal

Next to the news tab, there is a Polymarket tab showing live AI prediction markets. Trading volume and probability data on model launches, policy decisions, company moves.

The reason these sit together: news tells you what happened, prediction markets tell you what people with money at stake expect to happen next. When both signals align, the picture is clearer than either one alone. When they diverge, that is where the interesting stories are.

The trend signals can be expanded with other connections. Google Trends integration is coming. The point is that any data source that helps separate signal from noise belongs in the same view.

How it runs

A cron job fires every 6 hours. It fetches RSS feeds, extracts full article text as markdown, and caches everything in blob storage. The homepage reads from cache, never from live sources. When you press the Sumr button on an article, the extracted text goes to a fast model for a one-sentence summary, which is then cached so it only runs once per article.

No editorial team. No content calendar. The sources are curated once, the pipeline handles the rest. Entire thing is serverless.

What this taught us

Running this for months made one thing obvious: the hard part is not collecting information. It is filtering it. Most teams that struggle with staying informed do not have an access problem - they have a volume problem. Too many sources, too little time, and the people who need the information are rarely the ones who know where to find it.

That observation led to Sumr Trace - the same underlying pattern, but configurable per team: your sources, your interest filters, your feedback loop. It is in development now.

The same foundation is being extended into other directions at 1ar labs. If your team has a version of this problem - too much noise, not enough signal - reach out.