Under the Hood

Purpose-Built, by Investors
for Investors.

Vester is a long-running research agent built for analysis depth and factual grounding—not quick answers. It reasons across multiple data sources, verifies every claim, and keeps working until the research is truly complete.

Module 01

Task & Plan

A user defines the objective. The agent parses intent, builds an execution plan, deploys skills and sub-agents in parallel.

New Task

Priority

~47s

Run a multi-factor analysis on NVDA pre-earnings — pull 10-K data, check sentiment, compare with AMD and INTC, and generate a full investment report.

Detected:NVDAAMDINTC10-KSentiment

Analysis Types

Equity Researchactive

DCF Valuationactive

Comparable Analysisqueued

execution_plan.yaml

Parse user intent & identify tickers

Gather earnings transcripts (NVDA, AMD, INTC)

Pull 10-K filings & financial data

Fetch on-chain & sentiment signals

Run comparative analysis models

Generate charts & visual outputs

Compile investment report

Sub-Agents

5 active·1 queued

QuantAnalyst

40%

Running Python models

2/5 models

NewsAgent

85%

Scanning news feeds

3/4 sources

MacroAnalyst

91%

Pulling macro indicators

12 metrics

SentimentAgent

62%

Scanning social feeds

8.2K signals

DataAnalyst

73%

Aggregating datasets

6/8 fields

EquityResearcher

58%

Extracting 10-K sections

3 tickers

Module 02

Data Curation

Sub-agents fan out across financial data sources, pulling real-time and historical data simultaneously.

NVDA Q4 Earnings CallFeb 21, 2026

CEOAI demand continues to exceed our supply...

CFOData center revenue grew 409% year over year...

AnalystCan you quantify the inference vs training mix?

10-K Annual Report

SEC

NVIDIA CORP

CIK: 0001045810

Revenue

$60.9B

Net Income

$29.8B

Risk Factors

Extracting financials...

On-Chain AnalyticsLive

Whale Inflows

+$42M

DEX Volume

$2.1B

Active Addresses

1.4M

TVL

$48.7B

Avg Gas Fee

12.4 gwei

Net Flows (7d)

+$128M

Sentiment Analysis12,847 signals

Bullish

78% positive · trending up

Social Media

82%

News Sources

71%

Analyst Calls

88%

Prediction MarketsPolymarket · Kalshi

NVDA beats Q4 earnings?

Yes 82%

$4.2M

Guidance raise > 10%?

Yes 67%

$1.8M

Stock +10% in 30 days?

Yes 54%

$890K

Streaming data from 5 sources...

42.3 MB ingested

Module 03

Reasoning, Verification & Memory

Every data point auto-captured into a structured fact store with source provenance across 13 tools. A separate adversarial model cross-references the final response against source data before the user sees it — catching drift in volume and computed metrics.

Source Verification1/3

Financial DataNVDA · AMD · INTC

Metric

NVDA

AMD

INTC

Revenue

$60.9B

$22.7B

$54.2B

YoY Growth

+122%

+10%

-1.4%

Gross Margin

74.7%

49.1%

41.4%

Net Income

$29.8B

$1.6B

$1.7B

Op. Margin

62.4%

7.0%

3.1%

P/E Ratio

62.3x

45.1x

N/A

NVDA

AMD

INTC

Q4 FY2025

Data verified against SEC filings

Adversarial Validation

Scanning

ClaimAgentSourceResult

RevenueSEC 10-K

$60.9B$60.9B

—

YoY GrowthEarnings Call

+122%+122%

—

Avg Daily VolExchange API

847M892M

—

Gross MarginSEC 10-K

74.7%74.7%

—

DC RevenueQ4 Transcript

$47.5B$47.5B

—

Net IncomeSEC Filing

$29.8B$29.8B

—

Matched

Flagged

Pending

0 · 0 / 6

Fact Memory Chain

13 tools active

SEC

NEWS

CHAIN

MKT

EARN

ALT

SENT

PRED

MACRO

FLOW

TECH

FUND

RISK

SrcMetricValueProvenance

SECRevenue$60.9B

10-K FY2025

CHAINWhale Volume892M

Glassnode API

MKTP/E Ratio62.3x

Market Feed

NEWSSentiment78% Bull

NLP Pipeline

SECNet Income$29.8B

10-K FY2025

EARNDC Revenue$47.5B

Q4 Transcript

Every number → source query6 facts · 0 unverified

ConfidenceHigh

91Overall

Financial Analysis94%

Sentiment Signal87%

Market Pricing91%

All primary claims verified. Proceeding with high-confidence analysis.

Verification complete

13 tools instrumented · 6 facts captured · 1 drift caught · 91% confidence

Module 04

Quantitative Methods

Writes and runs analytical code on the fly — regressions, statistical models, custom algorithms.

analysis.py

1# Comparative revenue analysis — NVDA vs AMD vs INTC

2import pandas as pd

3from scipy import stats

5tickers = ['NVDA', 'AMD', 'INTC']

6revenue = fetch_quarterly_revenue(tickers, periods=8)

8for ticker in tickers:

9 slope, _, r, _, _ = stats.linregress(

10 range(len(revenue[ticker])), revenue[ticker])

11 growth_rates[ticker] = {"slope": slope, "r2": r**2}

regression_plot.svg

R² = 0.96

output

running

>>>NVDA: slope=2.84B/qtr, R²=0.96

>>>AMD: slope=0.31B/qtr, R²=0.82

>>>INTC: slope=-0.12B/qtr, R²=0.44

>>>Regression models fitted successfully.

Module 05

Output Generation

Produces publication-ready charts, visualizations, reports, and presentation decks.

Quarterly Revenue ($B)

FY2025 comparison

NVDA

AMD

INTC

Growth Trajectory

8-quarter trend with regression

NVDA +2.84B/qtr

AMD +0.31B/qtr

DC GPU Market Share

Data center GPU revenue share

NVDA80%

AMD12%

INTC5%

Other3%

Investment Report

24 pages, PDF

REPORT

1.Executive Summary

2.Revenue Analysis

3.Risk Factors

4.Valuation

5.Price Target

Executive Summary

Executive Summary Deck

12 slides, PPTX

DECK

Investment Thesis

Strong AI infrastructure moat

122% YoY revenue growth

Dominant DC GPU market share

1 / 12

Financial Model

DCF + Comps

MODEL

D7=NPV(WACC, CF_range) + Terminal_Value

FY2024

FY2025E

FY2026E

Revenue

$60.9B

$82.5B

$105.3B

COGS

$15.4B

$20.6B

$26.3B

Gross Profit

$45.5B

$61.9B

$79.0B

Gross Margin

74.7%

75.0%

DCF Value

=NPV(...)

Income Stmt

DCF

Sensitivity

Purpose-Built, by Investorsfor Investors.

Task & Plan

Data Curation

Reasoning, Verification & Memory

Quantitative Methods

Output Generation

Purpose-Built, by Investors
for Investors.