In this guide
What this covers
Cortex is the research layer: it reads onchain flow, holder structure, liquidity, and public sentiment, and it keeps reading after you have stopped asking. This guide covers turning a one-off question into a standing brief that respects your attention.
1. Define the question narrowly
Broad questions produce broad noise. Constrain the network, the behaviour, and the cohort:
Watch Base for tokens where smart-money wallets are accumulating
and brief me daily.Maren resolves "smart money" to a labelled wallet cohort and shows you the cohort size. You can inspect and edit that list — it is not a black box.
2. Choose the signals deliberately
| Signal | Reads | Fails when |
|---|---|---|
| Flow | Net accumulation by cohort | Wash trading inflates it |
| Holder structure | Concentration and distribution | Custodial wallets look like whales |
| Liquidity | Depth and lock status | Depth can be pulled without warning |
| Sentiment | Public discussion volume and tone | Coordinated posting is indistinguishable from interest |
Each has a failure mode. Confluence across several is the signal; any one alone is a coincidence.
3. Set the threshold, not just the cadence
"Report only on change" is the setting that keeps a brief useful. A daily brief that says the same thing every day trains you to stop reading it.
4. Read the brief as evidence, not instruction
Every claim is sourced. Inference is labelled as inference. Confidence is stated. What Cortex will never do is place an order — research and execution are separate on purpose, so a research error cannot become an execution error.
Onchain accumulation is observable behaviour, not a thesis. It tells you what someone did, never why. The why is still your job.
Tips
- Start with one narrow brief. Five broad ones is the fastest route to ignoring all of them.
- Ask "what would change your mind on this?" — the falsifier is usually more informative than the finding.
- Feed a promising finding into a contract audit before it goes anywhere near your capital.
Troubleshooting
Brief is too noisy. Tighten the threshold and reduce the signal count. Depth beats breadth here.
Signals conflicting. That is real information: it means the setup is genuinely unclear. Do not resolve it by picking your favourite signal.
The controls behind this workflow
- Every claim carries its source; unsourced inference is labelled as inference.
- Briefs never place orders. Research and execution stay separate by design.
- Low-confidence findings are marked rather than rounded up to certainty.
Guides describe how to operate Maren. They are not financial advice, and figures shown are illustrative. You approve every transaction before it is signed.