AutoHedge
Build an AutoHedge evidence lab: handoff traces and simulator counterexamples
Design an educational, no-signing lab that makes evidence provenance and fill assumptions visible without claiming a shipped trading dashboard or official roadmap.
What you will learn
- Build around inspectable evidence
- Make simulator assumptions interactive
- Define a safe completion contract
Before you start
- Basic Python, Git and dependency-management knowledge
- A fictional evidence task with no wallet or signing authority
Explain the inspected implementation and its counterexamples without mistaking a simulation or generated text for a verified financial outcome.
Key takeaways
- Make provenance and observed events visible independently of model narration.
- Use simulator counterexamples rather than decorative 3D.
- A proposed lab is not a shipped application or official roadmap.
Build around inspectable evidence
A useful follow-on project is an evidence lab, not an autonomous profit claim. Feed invented source records with timestamps and contradictions into fake research stages, then show which claims reached the final note. Keep missing dates visible and make the acceptance test independent from the generator’s self-confidence.
Record run ID, source snapshot, dependency environment, worker/tool events and the final accepted artifact. A timeline can distinguish suggested handoffs from observed calls. Treat event capture as a proposed integration: this article ships static explanations and bounded examples, not a working trace collector or live agent dashboard.
Make simulator assumptions interactive
A two-dimensional SVG view can show a flat price line, proposed bid/ask levels and separate assumed versus observed fill events. Add toggles for no fill, partial fill and fees, and show which assumptions change the accounting result. The aim is to teach model limitations rather than present simulated gains as investment evidence.
Three.js would be justified only if another spatial dimension materially helps the lesson. Here, a clear time/price chart and accessible controls are likely more informative than a rotating 3D object. The current series supplies static chapter-specific SVGs; no interactive widget, real market feed or trading control is claimed as delivered.
Define a safe completion contract
The lab should run with invented records, fake providers and no signing capability. Tests should reject invented provenance, unsupported confidence calibration and promotion of a simulated fill into a real execution record. Include negative cases and preserve null for unmeasured cost, latency and real returns.
This is an independent educational proposal, not AutoHedge’s official roadmap. The isolated source probe has zero real model calls, wallets, signatures and trades. Completing the proposed lab would require implementation and separate UI/integration tests; a published article about it is not completion of that software project.
Implementation steps
- 1
Create timestamped fictional evidence and contradictions.
- 2
Add fake handoff traces and independent artifact checks.
- 3
Expose fill assumptions and adverse cases in a 2D view.
- 4
Keep signing capability absent and unknown metrics null.
Copy-ready example
{"project":"evidence-and-fill-assumption-lab","interactiveDemoShipped":false,"officialRoadmap":false,"actualModelCalls":0,"actualWallets":0,"actualTrades":0,"measuredReturns":null}Frequently asked questions
Is the interactive lab already available on these pages?
No. These pages contain static diagrams and a proposed design; a live trace or simulator application has not been delivered.
Is this an official AutoHedge development commitment?
No. It is an independent learning project derived from the inspected code and its unverified boundaries.
Sources
- autohedge/main.pySource checked 2026-09-08
- autohedge/workers.pySource checked 2026-09-08
- experimental/market_making.pySource checked 2026-09-08