Skip to primary interaction
Finance

Macro Brief Studio

Macro workflow that selects a series, drafts a structured investor brief, cites every bullet to a dated observation, and scores grounding so inference is never mistaken for fact.

  • Turns a macro series into a cited, analyst-ready brief
  • Makes every claim traceable to a dated observation
  • Separates grounded statements from inference
Generate a cited briefSee the evaluation

CPI, all urban consumers

  • Reached 315.6 in Jan 2025 CPIAUCSL 2025-01
  • Steady disinflation over six months CPIAUCSL 2024-08

Primary interaction

Macro Brief Studio — Live AI Demo

Operating model

From a series to a cited brief

The brief is not free prose. Every grounded bullet binds to a dated observation, and conclusions that go beyond the data are flagged as inference.

01

Select

Pick a macro series to brief.

02

Generate

Draft structured investor bullets.

03

Cite

Bind each bullet to an observation.

04

Check

Flag inference, score grounding.

CPI, all urban consumers315.6▲ 5.2 since 2024-08

Recent observations

  • 2025-01315.6latest
  • 2024-12314.2
  • 2024-11313.1
  • 2024-10312
  • 2024-09311.2
  • 2024-08310.4
  • Headline CPI reached 315.6 in January 2025, up from 314.2 the prior month.
  • Monthly gains have eased from roughly 0.8 to 0.4 index points across the half-year, a steady disinflation trend.
  • At roughly 3 percent year over year, inflation likely stays above the 2 percent target, keeping any rate cuts gradual.inference, not grounded
67%grounding rate6 of 9 bullets cite a source
6citations issuedacross 3 series
3flagged inferencesshown as inference, not fact

Cited sources

  • CPIAUCSL 2025-01 · CPI, all urban consumers
  • CPIAUCSL 2024-08 · CPI, all urban consumers
  • FEDFUNDS 2025-01 · Effective federal funds rate
  • FEDFUNDS 2024-08 · Effective federal funds rate
  • T10Y2Y 2025-01 · 10-year minus 2-year Treasury spread
  • T10Y2Y 2024-10 · 10-year minus 2-year Treasury spread

Every grounded bullet binds to a dated observation. The remaining bullets are explicitly marked as inference.

Every bullet traces to a dated observation, and inferences are flagged.

67%grounding rate
3macro series
6cited bullets
Under the hood

Capability proof

Capability proof

Cited macro briefing

Service model

An analyst-facing brief generator over macro series.

Intelligence layer

Turns a series into structured bullets, each bound to an observation.

Operational state

Holds the series, observations, generated brief, and citations together.

Human control

Analysts open every citation and see inferred bullets flagged before using them.

Business value

Cuts time to a trustworthy macro read while keeping every claim traceable.

Architecture

The studio runs client-side over a precomputed synthetic dataset. In a real deployment, observations would come from a FRED client, the model would draft bullets constrained to cite retrieved values, and an offline eval would score grounding before any change shipped. Here the structure is the point: every grounded bullet binds to a dated observation, and inferred conclusions are marked rather than blended in.

What this demo is, and isn't

  • The series and observations are fictional and labeled synthetic. No real FRED data is used.
  • Briefs are precomputed for the demo, not generated live, so the focus is the citation pattern.
  • Grounding and the inference flag are illustrative of the eval surface a regulated deployment would need.