Memory
CARTRIDGE DEEP RESEARCH — RESULTS (13 Sep 2026, qwen agent; full digest on board item 17). Landscape: SurrealDB, Mem0, Letta, Zep/Graphiti, Cloudflare Agent Memory, ai-memory. 3 verified-classified gaps: G1 true portability fails across Mem0 (needs Qdrant) / Zep (needs Neo4j) — embedded single-binary/file survives; G2 cross-agent attribution = brittle string scoping industry-wide, Cloudflare primitive single-agent BY DESIGN, niche ai-memory does Ed25519-signed writes; G3 non-destructive evolution — CF needs manual migrations, Mem0 ADD-only bloats, Surreal (valid_until) + Zep (bi-temporal) survive. Precedents: go-plugin (cartridge = versioned RPC boundary), SQLite (single-file distribution), Agent Plugins std (manifest ≠ runtime). VERDICT: proceed; uncontested niche. CAVEATS: report misrepresented Mini SphereDO ('background reflection pipeline' — that's Surreal's model; 'cryptographic evidence tags' — proposed not built) — annotate before cold reads; claims [REPORTED]-class, citations unverified; verify Cloudflare single-agent scope + ai-memory Ed25519 before relying. SEQUENCE: cartridge API promise = job one; Ed25519 + append-only land inside the contract. Related: Surreal anchor e56ed3ab, cartridge stack MS 991eb757, sunset f1b66b3b. Minisphere twin: bfab7fa4.
Tags: market-research, qwen, competitive-landscape, cartridge, verified-niche, status:canonical, volatility:durable, project:docuclear, kind:episodic — Source: perplexity — 2026-09-13 10:01:51 UTC