Most quant projects publish the winning curve. I wanted to know whether the evidence could stay public even when it proved me wrong.
Open evidence pathFounder evidence mapArhan Canli
I built a system that keeps the evidence that proves me wrong.
ALPHAC is my attempt to make quantitative research publicly falsifiable. The interesting part is not a perfect curve. It is the machinery that freezes each attempt, keeps corrections visible and separates what I know from what I still need to prove.
A 90-second project case
Follow the decision, not the pitch.
Six evidence stops tell the story in order. The script is ready. The video has not been recorded.
The narration preserves the published evidence-map snapshot. Its forward count covers marks through 2026-10-11; it is not a current observation count.
-
-
ALPHAC records every attempt before judging it. The public union now contains 228 legacy identities and one prospective identity, including failures and incomplete packets.
Open evidence path -
When a flattering result breaks, the system keeps the original claim, the defect, the correction and the test that prevents the same mistake.
Open evidence path -
Three dedicated Alpaca paper accounts reconcile against the published record. They contain no client capital and prove no funded performance.
Open evidence path -
A signed append-only record connects public claims to artifacts and corrections. Verification proves integrity, not that the strategy will win.
Open evidence path -
The forward record has only 13 daily observations, external reviews remain at zero, and Foundry is planned not applied. The open burden is part of the result.
Open evidence path
This is a timed production script and evidence path. It is not a recorded video, proof that a viewer watched it or evidence of admissions impact.
Contribution ledger
Credit should be as inspectable as performance.
Self-disclosed and source-bound.
Not independently attested.
01Arhan CanliAccountable human
Founder, named author, project owner and final accountable human for methodology, claims, corrections and publication decisions.
- sets the project direction and accepts responsibility for what is published
- must personally approve author answers and manuscript-specific claims
- controls external submissions and responses to reviewers
- must be able to explain and defend every major research and system decision
02Tools, libraries, services and dataCapabilities and inputs
Open-source software, market-data providers, Alpaca paper accounts, DigitalOcean and web infrastructure supply capabilities and inputs.
Their use is not evidence that they endorse, reviewed or validated ALPHAC.
03External validation0 completed
These are local commissioning packets with blank reviewer fields. They prove no outreach, review, independent replication, submission, acceptance, or publication.
Assigned reviewers: 0. Independent replications: 0.
Three proof surfaces
What I want a skeptical reader to open next.
Correction traceSee the attractive result, the defect and the repair.46 killed or screen-killed candidates remain public.
Broker evidence3 of 3 Alpaca sleeves reconcile.Paper accounts only. No client capital. No funded performance.
Signed recordVerify the append-only chain from a clean command line.1399 entries since 2026-06-27. Integrity is not profitability.
Open burden
The unfinished work belongs in the result.
I do not have a mature forward Sharpe, independent review, a deployed research Foundry or funded performance. Those are not footnotes. They define what the next evidence must establish.
- Forward snapshot through 2026-10-11
- 13 observations
- Immature Record Too Short
- Sleeve objective
- 3 of 14
- Target not achieved
- Independent review
- 0 completed
- 0 replications
Signed founder commitment
A precise promise, not a funded track record.
I committed USD 1,000 of my own capital at first live (real-money) deployment of this book, into the same book and on the same terms as external capital. It activates only at first live deployment. Until then it is a signed forward commitment and nothing more.
curl -sO https://canlicapital.com/glassbox/founder_commitment.json
curl -sO https://canlicapital.com/glassbox/reproduce.py
python3 reproduce.py --dir .The working principle
Build the claim. Publish the evidence. Keep the failures.
The public record contains 114 research documents, 89 measurements and 349 recorded hypothesis identities against a declared budget of 400. It is larger than the part that worked because that is the only version worth trusting.