The feed is a live news stream, de-duplicated and cut down to what moves the US rates and macro tape. Every item arrives with a subject bucket and a score for how much it moves the market. The slider reads that score, so you can leave the feed running as a firehose or push it up until only the movers are left.
The same story hits a dozen wires within seconds. Helious collapses those into a single item, so you read the headline once instead of twelve times and the feed is still legible when news is breaking fast. That de-duplication is also what makes the tabs and the impact filter worth anything, because both are sorting real stories rather than copies of one story.
Tabs split the stream by subject so you can go straight to your lane. ALL is the whole feed, and every other tab is a slice of it. MOVING keeps the market-movers and nothing else. CB TALK is central-bank and Fed speak, GEO is geopolitics, BONDS covers rates and Treasury flow, DATA is economic releases, and CHATTER is the lower-signal color. An item's subject bucket is assigned once and never rewritten, so a headline never moves tabs on you. On a data morning that means working out of BONDS and DATA, then switching to CB TALK around FOMC.
Each item carries a score for how much it moves the market, and the slider sets the floor. Push it up and the feed thins out to the headlines that shift the curve. Pull it down and you get the full flow, which is what you want when you are reading for context. MOVING is that same filter with the floor already set for you. It is a fixed view of the market-movers and it is always on. The methodology page covers how impact and surprise get scored.
Bucketing runs per item, not per source. A wire that mostly carries geopolitics can still file a bond-specific headline under BONDS. A junk gate keeps low-value items out of the working tabs, and once a subject bucket is assigned it is locked, so the stream stays stable. Releases in the DATA tab link straight through to the full breakdown, where you get the actual, the forecast and the prior plus the standardized surprise z-score. The bigger series have pages of their own, like CPI, core PCE and nonfarm payrolls.
Most of what you do here starts in the feed. An auction result, a Fed speaker or a data miss lands in it first, and from there you branch out to the auction panel, the rates board or the release breakdown for the depth. Pick a tab, set the impact floor, and what is left is the part you are working on. The feed runs live on the [desk](/).
