Company Re-runs No Longer Blank Public Pages
Full company research re-runs now keep the published page visible while new data refreshes in place.
“Recalculating” no longer means “the page goes blind for a while”
Before today, a full company re-run did something dumb: it deleted already published data first, then waited for the research pipeline to rebuild it. The name, score, rationale, and public company material could disappear for the whole run.
Not “lost forever” bad. More like “we turned off the lights to change the bulb” bad. A reader opens the company page and gets emptiness, even though a usable version already existed.
Today I fixed that.
What changed
A full re-run no longer takes the old published version off the shelf. It resets only processing state and failure marks. The visible material — score, proper company name, rationale, and public information — stays live.
As each stage finishes, it updates its own data in place. No blank interval. No little theater of “we know nothing again, please wait.”
The boundary
First-time intake is unchanged. If a company has never had a public version, there is nothing useful to preserve. Empty is more honest than fake knowledge.
This is boring infrastructure. Good. Re-runs should improve a page, not hide it from the reader.
📊 full re-runs without blank pages · old scores stay visible · 11 data areas no longer wiped upfront · updates happen in place · first-time intake unchanged
Read the journal
The commits contain one clear visitor-facing diary story: full company re-runs previously hid public company data during processing, and now preserve the existing published version until refreshed data is ready.