Free · no key
Briefing
Ask for a topic. Arcop fetches Hacker News, GitHub, Reddit, Wikipedia, markets, and the Bitcoin chain, then scores each row. This page does not charge $ARCOP. The same call through the API does.
About this desk
- What you use it for
- Search once across news, code, Reddit, and Wikipedia when you want rows scored and kept with their source URL. Sign in if you want to pay to collect or train on them.
- What the numbers are
- Each row keeps the public URL, a quality score from the source’s own metric, and a reliability score. Markets, the chain, and Wikipedia start higher than Reddit. Training later keeps rows at reliability 40 or above as the answer index.
- What it leaves out
- The free page does not spend $ARCOP. Paying is a separate step. An empty paid fetch is not charged. Open the source URL before you treat a summary as the source.
Every desk is introduced the same way on the desk list.
Type at least 3 characters. Try “lightning” or “ethereum”.
How to read this
- Raw, processed, analysed
- Raw is the live row. Processed adds a short label such as high-quality or narrow. Analysed adds a sentence that names the metric behind the row: points, stars, a fee, or a 24-hour move. This page does not charge. The same call through the API does.
- Q, R, S, and Rel
- Quality rises with the size of the source metric. Reach also counts a second number, such as comments. Sentiment reads words in the title, except on markets, where it follows the 24-hour percent. Reliability starts from the source: the chain and markets sit high, Reddit sits lower. Check reliability before an agent acts on the row. Training later drops rows under 40.
- An empty list
- News sources keep the last 7 days. Markets and the Bitcoin chain ignore that window. If a coin or a fee comes back empty, turn those sources on. Keep stores a link in your clips. Save these rows writes the current list into a dataset you can train.
Get a key when you want the same rows from code. It is issued as soon as the description is real.