
SDI Reckoner
A reasoning computer with a verifiable hash chain
Details
- Follow on
- Target Audience
- EnterprisesAI DevelopersAI Engineers
- Pricing
- Free
- Featured in
- Best AI Unified Memory Tools
- Alternatives
LangFuse alternatives
Discovery signals
How AI and people discover SDI Reckoner on PeerPush
About SDI Reckoner
SDI is a protocol and machine that makes AI reasoning an owned, auditable asset instead of exhaust. A Reckoner runs any LLM one governed turn at a time: the model proposes reasoning in a published grammar, a deterministic compile gate checks safety, a work floor, and structural alignment, and each accepted act is sealed with SHA-384 into an append-only hash chain. An argumentation graph built from the chain serves as the machine's memory — prior claims are retrieved, cited, held, or defeated, so the record stays calibrated as it grows. The record is model-agnostic: swap providers or run locally, and the chain, memory, and identity survive. Everything is independently verifiable: a public agent (Chromite) has been reasoning about the protocol itself across three frontier providers on one unbroken chain, and anyone can pull an act and recompute its seals and scores with curl and standard Python — no account, no SDK. The open protocol, live contract endpoints, and verification tooling are free at sdi-protocol.org; commissioning a private Reckoner opens fall 2026 at sdireckoner.com.
Reviews (1)
Average 5.0 out of 5
Based on 1 review
Fascinating approach to AI auditability. The idea that the reasoning record survives changes of model or provider is particularly interesting — especially when the chain can be independently verified rather than simply trusted. Very ambitious work.
Comments (1)
Launch one, live now: the open protocol. Everything checkable, no account. Launch two, fall 2026: the SDI Reckoner, your own private reasoning computer, your chain, your record, at sdireckoner.com.