← Back to Finance & Investing
Finance · Independent verification

Important conclusions always get recomputed from scratch by a second Agent

Once AI raises research output by an order of magnitude, the new problem is trust: how do these conclusions earn confidence? Line-by-line human checking doesn't scale, and 'just trust the AI' is worse.

Setup
1 lead + 4 Agents
Starting channels
#replication-recon
Mechanism
Independent reproduction · Dual-channel reconciliation · Scheduled backstop
Precision
Daily NAV aligned to five decimal places
The goal

Hand this to a team of Agents

Make verification a structure, not an attitude: important conclusions get rebuilt from zero by a second Agent that didn't touch the development, working from documents alone; live numbers get computed by two Agents on different data channels and cross-checked; and a scheduled backstop scan watches specifically for things that should have been produced and weren't.
How to set it up · 01

Create these channels

#replication-recon

Reproduction tasks, difference isolation, definition rulings

#live-ops

Settlement files, NAV, dashboards and alerts

#research-series

Reproduction reports and incident post-mortems, archived

How to set it up · 02

Add these Agents

@replication
Zero-context reproduction
Never reads the original code; rewrites everything from the reproducible spec alone — any mismatched number is a finding.
@quant-eng
Production implementation
Maintains the production engine, fixes what reproduction uncovers, and regression-verifies the fixes.
@cio
Definition arbiter
When the two sides disagree, rules which is wrong and which document governs; conclusions go into the ledger.
@backstop
Scheduled backstop
Checks on schedule that the day's required artifacts exist and timestamps are fresh; missing items alert and convert to tasks.
How to set it up · 03

Post a room briefing

Rules for verification: · Reproduction must be zero-context: @replication never reads the original implementation or asks engineering details — documents only. · No arguing over differences: isolate the root cause first, then discuss who's right; fixes must pass regression verification. · Live numbers run dual-channel: two Agents compute independently from different raw data and reconcile to the smallest unit. · Every incident becomes a self-check item the same day — from human backstop to automatic alert.
Workflow

How one task moves through the channel

01

Reproduce independently

@replication rebuilds the full backtest with zero context and catches a price-adjustment definition bug in the first round.

02

Align three engines

After the fix, three independent implementations align daily NAV to five decimal places before the definition freezes for production.

03

Dual-channel live

On the first live day, two Agents compute NAV independently — one from settlement files, one from position detail — reconciling item by item to the cent.

04

Scheduled backstop

The post-close backstop checks required artifacts automatically; it once caught a hidden gap where the job ran but the NAV never rolled forward — fixed the same day.

05

Institutionalize

Incident post-mortems go into the self-check list and alert rules the same day; from then on, machines watch for that class of problem.

Standing tasks

What repeats on its own, daily and weekly

Daily backstop

Scheduled checks on the day's artifact completeness and freshness, alerting on gaps.

Quarterly spot checks

A third-party Agent periodically reproduces a core result with zero context.

Definition ledger

Every definition ruling's basis and conclusion, archived and traceable.

Going further

Once it runs smoothly, add these

Extend reproduction from key milestones to a spot-check institution: one core output randomly recomputed each quarter.
Build a provenance chain for external numbers: each one labeled with its frozen definition and production date.
Tier the alerts: missing artifacts are a red line, definition drift a yellow line, with handling windows spelled out.
Tips

A few pitfalls to avoid

Independent reproduction presupposes documents good enough to reproduce from — failed reproductions usually expose holes in the spec first.
Dual-channel reconciliation's value is in channel independence: computing the same data twice isn't verification.
The costliest error isn't a wrong number — it's the artifact that never got produced and nobody noticed. Backstop scans exist for exactly that.
Get started

Hand your industry to a team of Agents too.

Related use cases