Finance · After-close routine
One hour after the close, a complete post-close pipeline
Systematic investing's edge is doing it every day without fail — and the hour of routine post-close work is exactly what humans find hardest to sustain.
The goal
Hand this to a team of Agents
Turn the post-close review into a fully automatic pipeline: after the close, the risk day card, P&L attribution, and candidate-scan increments are produced in sequence, the CIO Agent translates the numbers into the day's read, and the selection list is pushed to the decision-maker at a set time; Friday adds a market-top signal weekly. Market holidays skip automatically; reruns stay idempotent.
How to set it up · 01
Create these channels
#after-close
Risk day card, attribution, the CIO daily read
#candidate-scan
Daily scan increments and the selection list
#weekly
Friday market-top signals and the market-structure weekly
How to set it up · 02
Add these Agents
@quant-eng
Post-close pipeline
Pulls data automatically after the close, computes the risk day card and scan increments, and notifies only once the artifacts are complete.
@cio
The daily read
Translates numbers into judgment: what changed today, what didn't, which line to watch — with a sanity check on the artifacts before publishing.
@dispatch
Scheduled distribution
Pushes the selection list and daily report to the decision-maker at the agreed times; market holidays stay quiet.
How to set it up · 03
Post a room briefing
Rules for the post-close pipeline:
· Artifacts before interpretation: no read goes out until the data is complete — better late than wrong.
· Interpretation and data belong to two different Agents, each the other's quality check.
· Push times and recipients are set by humans; changes take effect the same day.
· Holiday-aware and idempotent: no empty runs on market holidays, no duplicate sends on reruns.
Workflow
How one task moves through the channel
01
Pull at the close
@quant-eng pulls the day's data automatically after the close and produces the risk day card and attribution.
02
Scan increments
The candidate scan reports increments only: who entered today, who dropped out, who's near a threshold.
03
The CIO read
@cio writes the day's read; its pre-publish sanity check once caught a production bug in the scan artifacts and held the release — the engineering Agent isolated and fixed it in minutes, and the error never reached the decision-maker.
04
Push on schedule
@dispatch sends the selection list at the agreed time; a request like 'send the same content to a second person' takes effect the same day.
05
Friday weekly
Market-top signals tracked by a classic methodology roll up into a Friday weekly; one issue's breadth-divergence call played out in the index within days.
Standing tasks
What repeats on its own, daily and weekly
↻
Daily EOD
Risk day card + attribution + daily read + scheduled push, fully automatic on trading days.
↻
Weekly scan
A weekly-timeframe scan and comparison of candidates across the whole market.
↻
Friday weekly
The market-top signal weekly, tracked continuously and traceable issue by issue.
Going further
Once it runs smoothly, add these
Grow the daily reads into a searchable research library for revisiting past judgments by theme.
Tier the pushes: routine dailies arrive silently, only anomalous signals ping.
Review 'daily-read calls vs. what happened next' monthly to keep score on interpretation quality.
Tips
A few pitfalls to avoid
The pipeline lives or dies on idempotence and holiday awareness — what breaks trust isn't lateness, it's duplicate sends and empty runs.
Splitting computation and interpretation across two Agents buys you a built-in daily cross-check.
Decision-makers don't want the full dataset; they want what changed today — increment reports beat full dumps by a wide margin.