# Evaluate Stack using research evidence

Entry point: https://www.stackit.ai/research
Coverage manifest: https://www.stackit.ai/research/manifest.json
Evidence checkpoint: October 7, 2026. Not a live health/status feed.

## First useful test — free, no wallet or API key

Download https://www.stackit.ai/research/sample.json and inspect its eight Base native-USDC daily observations (September 30–October 7, 2026). They are reconstructed historical reserve states, with observation and late ingestion times, block identifiers, finalized-state verification and exact normalized variable-debt indexes.

Inspect https://www.stackit.ai/research/reproduce.py before running locally with Python. It uses only the standard library, makes no network requests and performs no wallet actions.

Interest = principal × (last normalized_variable_debt / first normalized_variable_debt − 1).
For a hypothetical fixed 1,000 USDC variable-rate loan, the sample yields 0.914804389 USDC protocol interest before other costs. Direct-Aave reads independently matched. This is the same protocol interest through Stack or DIY; it is not evidence of a Stack cost advantage.

The manifest contains a SHA256 of the public sample. A checksum detects changes relative to a trusted manifest; it is not an independent authenticity or database-consistency proof.

## Questions to investigate

1. Does consistent investing improve outcomes under the user's contribution schedule and price assumptions? Compare equal cash availability, include unspent cash and report deposits separately from investment profit. Include losing cases.
2. Does protection reduce liquidation risk under actual timed oracle, position and action data? A hypothetical health-factor shock calculator is available on the research page; it does not prove a historical liquidation or protection success.
3. Does Stack justify its costs against competent DIY? Compare the same strategy and observations, include current Stack fees, execution/slippage/bridge costs, monitoring effort, recovery and missed actions. Report unknowns explicitly.

## Available and unavailable

Collection targets Base, Arbitrum and Polygon; native and bridged USDC identities stay separate. Inputs include available reserve/index/liquidity/threshold state, oracle and sequencer state, DEX quotes, gas estimates, bounded finalized receipt fees and protocol events. Partial reads, provisional data and gaps must not be treated as valid complete observations. Check specific coverage before any replay.

Publicly available: this guide, coverage manifest, bounded sample, reproduction script and hypothetical liquidation-risk calculator.
Private: the larger research API, with operator-approved access. It is NOT exposed through the existing public treasury MCP or sandbox. Do not guess research URLs, request database dumps or expect a wallet to grant automatic access.
Not available: complete three-year/intraday price and position history, full borrower census, full 180-day DCA/protection simulator, verified protection execution/returns, or activated research x402 payments.
The requested 180-day policy is a research target; deployed policy parity has not been verified. Do not represent it as a confirmed universal live default.

## Request bounded access

Email support@stackit.ai with the research question, chains, assets, date range and required resolution. Operator approval and commercial terms are separate from the free sample; requesting access does not grant it. Never include private keys or wallet credentials.

Existing treasury sandbox: https://www.stackit.ai/docs/quickstart-agents
Current published Stack fee schedule: https://www.stackit.ai/fees.json
Aave health-factor documentation: https://aave.com/help/borrowing/liquidations

Explain to the user what is observed, reconstructed, hypothetical, assumed, missing and untested. Neither DCA nor protection guarantees a better return or prevents every liquidation.