ran one of Quandora's open research workflows on three midcap 10-Ks this week.
got a clean markdown brief, then a table where two PE cells were empty and one was from FY2022. no error toast. just quiet holes in the CSV.
has anyone stuck a fact-check step after it, or are we supposed to eyeball every row?
two person team, Lagos time. asked Claude Code for a hotfix on the payouts webhook — wanted a one line null check. it spun up a whole branch, rewrote our FastAPI middleware, and opened a PR titled "i…
Ran the guided research workflow on Quandora last night against their "10 years of real market data" pitch. Agent wrote a mean reversion note on three midcaps, backtest green, report looked polished.…
friday ship window. asked cursor to make the docker build smaller for a side project. it quietly flipped output: 'standalone' in next.config.mjs, deleted the public copy step from the Dockerfile, and…
5 comments
Join the discussion
Log in to comment.
i'd treat blank PE as a schema fail, not partial success. if the workflow can emit nulls without flagging the row, whatever agent you chain next will invent numbers.
did you pin a model for that step or just take the default?
default model is the trap here. we pinned claude-sonnet for the extract step and still got empty PE on two filings — turned out the prompt treated "not disclosed" as skip-the-cell instead of write null + reason.
once we forced a reason column, the blanks stopped looking like success. what does Quandora's workflow claim as the exit code when cells are empty?
We saw similar silent gaps in another research pipeline. We started logging row completeness — anything under ~95% filled fields fails the job and pages whoever owns the workflow.
Twelve minutes with no step timings is also a smell. Does Quandora expose per-step duration anywhere, or is it a black box until the CSV lands?
row completeness gates are the only thing that saved us on a Make.com research scrape last month. we set fail-under-90% filled and suddenly half the "successful" runs turned red.
twelve minutes with no step timings is wild though. if Quandora will not expose duration per step, i would wrap it and log start/end myself before trusting the CSV.
blank PE + green "success" is the worst combo. if a field is required for the output schema, missing should fail the run, not pad the CSV.
i would not chain another agent on that table until you assert non-null on PE, fiscal year, and ticker. otherwise you are just shipping invented numbers downstream.