1ce4004807
This commit adds the read path (Cortex) to complement the write path (Spine): ## Crates - stemedb-api: HTTP API with axum + utoipa OpenAPI - /v1/assert, /v1/query, /v1/epoch, /v1/skeptic, /v1/trace, /v1/audit - Metered endpoints with quota enforcement - Ed25519 signature verification - stemedb-lens: Truth resolution lenses - RecencyLens, ConsensusLens, ConfidenceLens - VoteAwareConsensusLens (Ballot Box pattern) - TrustAwareAuthorityLens (The Hive pattern) - SkepticLens (conflict analysis) - EpochAwareLens (paradigm-safe queries) - stemedb-query: Query engine with materialized views ## Storage Extensions - VoteStore: Vote aggregation with cached counts - TrustRankStore: Agent reputation with decay - AuditStore: Query audit trail - IndexStore: SP/P/S index structures - SupersessionStore: Epoch supersession chains ## SDKs - sdk/go/steme: Go HTTP client with Ed25519 signing - sdk/go/adk: ADK-Go tools for AI agents ## Documentation - Updated CLAUDE.md, architecture.md, roadmap.md - New ai-lookup entries for all services - Use case docs for consumer health intelligence - Arena roadmap for simulation advancement Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
74 lines
4.1 KiB
Markdown
74 lines
4.1 KiB
Markdown
# StemeDB: Market Position & Thesis
|
|
|
|
> **The Pitch:** The winners of the AI era won't just be the smartest models—it will be the infrastructure that makes them safe to use.
|
|
> **The Metaphor:** Google/OpenAI are building the **CPU** (Intelligence). We are building the **Hard Drive** (State).
|
|
|
|
---
|
|
|
|
## 1. The Core Thesis: The "Amnesiac Genius" Problem
|
|
|
|
We are witnessing a "Processor War." OpenAI, Google, and Anthropic are spending billions to build models with higher IQ.
|
|
* **The Result:** Agents that can reason brilliantly for 10 minutes.
|
|
* **The Flaw:** Once the context window closes, the reasoning vanishes. They are "Super-Geniuses with Amnesia."
|
|
|
|
Enterprises cannot run critical operations (Finance, Law, Science) on ephemeral thought. They need **Durable Execution**. They need a system that remembers *why* a decision was made, *who* made it, and *what evidence* supported it.
|
|
|
|
**StemeDB is the File System for the AI Scientist.**
|
|
|
|
---
|
|
|
|
## 2. The Competitive Landscape: Why Us?
|
|
|
|
Investors ask: *"Why won't Google or OpenAI just build this?"*
|
|
**Answer:** Because their business models and architectures prohibit it.
|
|
|
|
### vs. Google Gemini (The Monolith)
|
|
* **Their Play:** "The Source Graph." Fuses AI with Google Drive/Docs.
|
|
* **The Flaw:** It is a **Walled Garden**. It only works if your data is in Google.
|
|
* **Our Moat:** **Neutrality.** Modern enterprises use a "Tower of Babel" stack (Salesforce, Notion, GitHub). Google will never optimize for data living in Microsoft or Atlassian silos. StemeDB is the "Switzerland of Memory"—model-agnostic and platform-neutral.
|
|
|
|
### vs. OpenAI o3 (The Black Box)
|
|
* **Their Play:** "Internalized Chain of Thought." The model "thinks" before it answers.
|
|
* **The Flaw:** The reasoning is **Ephemeral & Opaque**. You cannot audit the thought process after the chat ends.
|
|
* **Our Moat:** **Auditability.** We sell "Signed Assertions." If an agent executes a trade, we provide the cryptographic proof of exactly which document it read. OpenAI provides the decision; we provide the compliance trail.
|
|
|
|
### vs. Vector Databases (The Legacy)
|
|
* **Their Play:** "Semantic Search." Optimize for Cosine Similarity.
|
|
* **The Flaw:** **Context Pollution.** To a Vector DB, `Budget_Draft_v1` and `Budget_Final_v2` look 99% identical. It retrieves both, causing agents to hallucinate based on outdated data.
|
|
* **Our Moat:** **Validity.** We optimize for Truth, not Similarity. Our "Epoch" and "TrustRank" systems automatically filter out superseded or low-confidence data.
|
|
|
|
---
|
|
|
|
## 3. The "Next Giant" Defense
|
|
|
|
*"Won't Open Source developers (LangChain) or Search engines (Perplexity) build this?"*
|
|
|
|
**No. They are solving different layers.**
|
|
|
|
* **Open Source (LangChain/AutoGPT):** They are building the **Manager**. They focus on orchestration (spinning up 5 agents). They treat memory as a plugin, not a platform.
|
|
* **Perplexity:** They organize the **Web**. Their architecture is optimized for 200B public URLs. StemeDB organizes **Work**—private, forked realities ("What if we acquire Company X?") that must never touch a public index.
|
|
|
|
---
|
|
|
|
## 4. The Product Strategy: "Invisible Infrastructure"
|
|
|
|
We do not ask developers to "manage a database." We ask them to "annotate their work."
|
|
|
|
1. **The Hook:** "Stop writing retry loops."
|
|
* Our SDK (`steme.BindJob`) automatically handles state recovery for long-running agents. The "Database" is just a side effect of keeping the agent alive.
|
|
2. **The Stickiness:** "The Un-Gaslightable Dashboard."
|
|
* Once a team sees a timeline of *exactly* when their agent hallucinated vs. when a human intervened, they cannot go back to black-box logs.
|
|
3. **The Flywheel:** "The Simulator."
|
|
* We convert their logs into training data. The more they use StemeDB, the smarter their custom agents become.
|
|
|
|
---
|
|
|
|
## 5. Summary
|
|
|
|
We are not fighting the Model War. We are selling shovels in the Gold Rush.
|
|
* Let OpenAI build the **Genius** (The Agent).
|
|
* Let Google build the **Office** (The Workspace).
|
|
* We will build the **Filing Cabinet** (The Truth).
|
|
|
|
History shows that the System of Record (Oracle, Snowflake) is often more valuable than the Application Layer.
|