https://mnemoverse.com/docs/library Research and writing on persistent memory for AI agents. en Sun, 04 Oct 2026 00:00:00 GMT https://mnemoverse.com/docs/library/ai-code-review-benchmarks https://mnemoverse.com/docs/library/ai-code-review-benchmarks AI code review benchmarks explained: answer keys, LLM judges, counting rules and live leaderboards, and what to ask before trusting a #1 claim. Edward Izgorodin Sun, 04 Oct 2026 00:00:00 GMT https://mnemoverse.com/docs/library/vector-search-duplicate-results https://mnemoverse.com/docs/library/vector-search-duplicate-results Duplicate results in vector search have three causes: look-alikes, one document's chunks, one fact stored twice. The setting that fixes each, store by store. Edward Izgorodin Thu, 01 Oct 2026 00:00:00 GMT https://mnemoverse.com/docs/library/agent-memory-write-gate https://mnemoverse.com/docs/library/agent-memory-write-gate How an agent memory write gate filters noise and duplicate context: what six systems test on the write path, and what a refused write becomes. Edward Izgorodin Sun, 27 Sep 2026 00:00:00 GMT https://mnemoverse.com/docs/library/agent-memory-catch-up-read https://mnemoverse.com/docs/library/agent-memory-catch-up-read How AI agents catch up on memory they missed: a queryless newest-first feed, cursor paging while others write, and empty answers that name their scope. Edward Izgorodin Tue, 22 Sep 2026 00:00:00 GMT https://mnemoverse.com/docs/library/long-term-associative-memory-in-python https://mnemoverse.com/docs/library/long-term-associative-memory-in-python Long-term associative memory in Python: run a standard-library example, compare six memory APIs, and test persistence, expansion, and feedback. Edward Izgorodin Tue, 22 Sep 2026 00:00:00 GMT https://mnemoverse.com/docs/library/outcome-reranking-agent-memory https://mnemoverse.com/docs/library/outcome-reranking-agent-memory Eight memory systems checked against one rule: is there a documented call that takes a recall outcome and changes the next order? Fields, terms, defaults. Edward Izgorodin Tue, 22 Sep 2026 00:00:00 GMT https://mnemoverse.com/docs/research/evaluation/ragas-framework-deepdive https://mnemoverse.com/docs/research/evaluation/ragas-framework-deepdive RAGAS metrics explained: faithfulness, response relevancy, context precision and recall. Two of the six text RAG metrics need no reference; none spans sessions. Edward Izgorodin Sun, 20 Sep 2026 00:00:00 GMT https://mnemoverse.com/docs/research/memory/architectures/continual-learning-algorithms https://mnemoverse.com/docs/research/memory/architectures/continual-learning-algorithms Continual learning algorithms explained: what EWC, GEM, A-GEM and iCaRL actually fixed about catastrophic forgetting, and what transfers to AI agent memory. Edward Izgorodin Sun, 20 Sep 2026 00:00:00 GMT https://mnemoverse.com/docs/library/agent-memory-consolidation-compared https://mnemoverse.com/docs/library/agent-memory-consolidation-compared How do memory layers handle automatic consolidation and forgetting? Six systems, six answers for the old record when a newer memory contradicts it. Edward Izgorodin Sat, 12 Sep 2026 00:00:00 GMT https://mnemoverse.com/docs/library/agent-memory-deletion https://mnemoverse.com/docs/library/agent-memory-deletion Agent memory deletion reviewed across eleven systems at pinned commits: what a delete touches, whether it reaches derived records and caches, and who is recorded doing it. Edward Izgorodin Sat, 12 Sep 2026 00:00:00 GMT https://mnemoverse.com/docs/library/stop-agent-repeating-fixed-mistakes https://mnemoverse.com/docs/library/stop-agent-repeating-fixed-mistakes You fixed it, explained it, and it came back next session. Six agent memory systems put to one mechanical question: does anything change when the outcome was bad? Four of the five others publish an input that takes a negative verdict, fewer publish what it moves, and one word returns zero files in all five organisations. Edward Izgorodin Thu, 10 Sep 2026 00:00:00 GMT https://mnemoverse.com/docs/library/agent-memory-knowledge-graphs-compared https://mnemoverse.com/docs/library/agent-memory-knowledge-graphs-compared Six agent memory systems and all six build a knowledge graph, so the useful question is a different one. The word graph names four different objects here, in all six the edges are made on the write path, and twenty searches for the vocabulary of read-driven association return zero files. Edward Izgorodin Tue, 08 Sep 2026 00:00:00 GMT https://mnemoverse.com/docs/library/cursor-memory-bank-what-loads https://mnemoverse.com/docs/library/cursor-memory-bank-what-loads The canonical Cursor Memory Bank ships fifty five rule files. Exactly one of them is guaranteed to reach a session, and it tells the agent where to write, not what to read. Checked against the repository, the vendor's own pages and the forum threads where it fails. Edward Izgorodin Tue, 08 Sep 2026 00:00:00 GMT https://mnemoverse.com/docs/library/mcp-memory-servers-claude-code-and-cursor https://mnemoverse.com/docs/library/mcp-memory-servers-claude-code-and-cursor Six memory systems checked against two named clients. In the terminal, five of five now ship a first-party plugin rather than a bare protocol server, the asymmetry between the two clients flips vendor by vendor, and the row that decides the choice is where the memory is keyed. Edward Izgorodin Tue, 08 Sep 2026 00:00:00 GMT https://mnemoverse.com/docs/library/python-sdk-agent-memory-compared https://mnemoverse.com/docs/library/python-sdk-agent-memory-compared Every one of these publishes a one line install command, and not one of those lines gives you working memory. Three checks, all faster than reading a landing page, tell you what the install line will not: the runtime dependency list on PyPI, the class names the package exports, and what an empty constructor asks for first. Edward Izgorodin Tue, 08 Sep 2026 00:00:00 GMT https://mnemoverse.com/docs/library/mcp-memory-coherence https://mnemoverse.com/docs/library/mcp-memory-coherence MCP defines no ordering, conflict detection or merge semantics for memory. Coherence is a property each server chooses, and shipped servers choose differently. Edward Izgorodin Fri, 04 Sep 2026 00:00:00 GMT https://mnemoverse.com/docs/library/what-survives-when-ai-agents-restart https://mnemoverse.com/docs/library/what-survives-when-ai-agents-restart My LLM agents forget conversation history when I restart them. All five coding tools checked here already ship a working resume, so the real gap is narrower: a brand new session, and what compaction drops even when you do resume. Edward Izgorodin Fri, 04 Sep 2026 00:00:00 GMT https://mnemoverse.com/docs/library/anthropic-memory-api-no-ranked-retrieval https://mnemoverse.com/docs/library/anthropic-memory-api-no-ranked-retrieval Anthropic's memory surfaces ship storage, versioning, consolidation and forgetting. Retrieval is regex over a mounted filesystem, with no relevance model anywhere. Edward Izgorodin Thu, 03 Sep 2026 00:00:00 GMT https://mnemoverse.com/docs/library/mcp-servers-npx-setup-compared https://mnemoverse.com/docs/library/mcp-servers-npx-setup-compared Type npx into six memory-server MCP vendors' setup docs and get six different answers. Only one documents clients spawning its own MCP server with it. Checked live against every vendor's own documentation. Edward Izgorodin Thu, 03 Sep 2026 00:00:00 GMT https://mnemoverse.com/docs/library/mcp-servers-sharing-state-across-ides https://mnemoverse.com/docs/library/mcp-servers-sharing-state-across-ides Every persistent memory MCP server that claims cross-IDE sharing reduces to the same mechanism: one account, several clients, one backend. Six vendors checked for the concrete way sharing can fail while every client still reports a healthy connection. Edward Izgorodin Thu, 03 Sep 2026 00:00:00 GMT https://mnemoverse.com/docs/library/ai-memory-apis-prune-low-value-context https://mnemoverse.com/docs/library/ai-memory-apis-prune-low-value-context Persistent memory APIs compared on two mechanisms people call by one word, clean: outcome-based feedback that reranks a memory after it fails, and consolidation that reorganizes the memory itself. Six systems, including our own. Edward Izgorodin Wed, 02 Sep 2026 00:00:00 GMT https://mnemoverse.com/docs/library/best-persistent-memory-apis-compared https://mnemoverse.com/docs/library/best-persistent-memory-apis-compared Six persistent memory APIs for AI agents compared on the six questions that actually decide fit: memory model, cross-client sharing, outcome feedback, multi-agent sharing, cost, and a notable limitation for each. Edward Izgorodin Wed, 02 Sep 2026 00:00:00 GMT https://mnemoverse.com/docs/library/best-ai-agent-memory-2026 https://mnemoverse.com/docs/library/best-ai-agent-memory-2026 What is the best AI agent memory in 2026? A decision map across Mem0, Zep, Letta, Cognee, Supermemory, LangMem, and Mnemoverse, with the vendor position flagged up front. Edward Izgorodin Tue, 01 Sep 2026 00:00:00 GMT https://mnemoverse.com/docs/library/agent-repeats-mistakes https://mnemoverse.com/docs/library/agent-repeats-mistakes AI coding agents repeat corrected mistakes because nothing records whether past suggestions worked. Outcome memory attaches results to recalled facts. Edward Izgorodin Mon, 31 Aug 2026 00:00:00 GMT https://mnemoverse.com/docs/library/why-agents-need-to-forget https://mnemoverse.com/docs/library/why-agents-need-to-forget Vector stores return deprecated APIs and dead conventions forever. Closing, decay, and suppression keep agent memory current without destroying history. Edward Izgorodin Mon, 31 Aug 2026 00:00:00 GMT https://mnemoverse.com/docs/library/hindsight-vs-graphiti https://mnemoverse.com/docs/library/hindsight-vs-graphiti Hindsight and Graphiti answer the same question at different moments: Graphiti at write time with a bi-temporal graph, Hindsight at read time with standing answers. How to choose, and what neither settles. Edward Izgorodin Wed, 26 Aug 2026 00:00:00 GMT https://mnemoverse.com/docs/library/agents-md-supported https://mnemoverse.com/docs/library/agents-md-supported The format has no required fields and one sentence about precedence. Seven tools that list themselves as supporting it describe seven different mechanisms, and one of them says outright that a closer file overrides an earlier one because it comes later in the combined prompt. Olga Timoshina Tue, 25 Aug 2026 00:00:00 GMT https://mnemoverse.com/docs/library/claude-md-agents-md-cursor-rules https://mnemoverse.com/docs/library/claude-md-agents-md-cursor-rules A rules file is text placed in a context window. All three vendors say so on their own pages, in their own words. What delivers a rule, what the wrapper around it says, and what does enforce, because something does. Olga Timoshina Tue, 25 Aug 2026 00:00:00 GMT https://mnemoverse.com/docs/library/what-is-an-mcp-memory-server https://mnemoverse.com/docs/library/what-is-an-mcp-memory-server MCP standardises how a client discovers a tool, calls it, and reads the answer. It does not standardise memory. Walking the whole path with the specification and the public reference server open. Olga Timoshina Sun, 23 Aug 2026 00:00:00 GMT https://mnemoverse.com/docs/library/memory-response-shape https://mnemoverse.com/docs/library/memory-response-shape A ranked list cannot express supersession, retraction, or authority. Why retrieval flattens conflicts the store holds, and what a response with relations needs. Edward Izgorodin Sat, 22 Aug 2026 00:00:00 GMT https://mnemoverse.com/docs/research/evaluation/bleu-rouge-f1-sari-explained https://mnemoverse.com/docs/research/evaluation/bleu-rouge-f1-sari-explained BLEU, ROUGE, F1, accuracy, and SARI explained by metric question, Evaluate I/O shape, failure mode, and routing. Edward Izgorodin Mon, 17 Aug 2026 00:00:00 GMT https://mnemoverse.com/docs/library/hopfield-attention-equivalence https://mnemoverse.com/docs/library/hopfield-attention-equivalence Transformer attention is equivalent to one update step of a modern Hopfield network under specific query, key, value substitutions. Edward Izgorodin Sun, 16 Aug 2026 00:00:00 GMT https://mnemoverse.com/docs/library/prompt-cache-stable-prefix https://mnemoverse.com/docs/library/prompt-cache-stable-prefix How prompt cache keys are derived across Anthropic, OpenAI, Gemini, vLLM, and SGLang, and the six agent mistakes that silently break the stable prefix. Edward Izgorodin Sun, 16 Aug 2026 00:00:00 GMT https://mnemoverse.com/docs/library/rag-vs-agent-memory https://mnemoverse.com/docs/library/rag-vs-agent-memory At query time a memory layer runs the same hybrid retrieval a RAG stack does, and the shipping source proves it. The difference is on the write path, and it is narrower than the category claims. Edward Izgorodin Sun, 16 Aug 2026 00:00:00 GMT https://mnemoverse.com/docs/library/agent-memory-knowledge-graph-vs-retrieval https://mnemoverse.com/docs/library/agent-memory-knowledge-graph-vs-retrieval Knowledge graph vs retrieval for AI agent memory: the split that matters is learned vs asserted structure, connected by provenance. Evidence-first guide. Edward Izgorodin Sat, 15 Aug 2026 00:00:00 GMT https://mnemoverse.com/docs/library/agent-memory-evidence-vs-policy https://mnemoverse.com/docs/library/agent-memory-evidence-vs-policy Agent memory splits into evidence and policy, not short versus long term: three scopes, the fields each entry needs, and why closing a fact beats replacing it. Edward Izgorodin Wed, 12 Aug 2026 00:00:00 GMT https://mnemoverse.com/docs/library/agent-memory-feedback-dilemma https://mnemoverse.com/docs/library/agent-memory-feedback-dilemma Agent memory learns most from explicit feedback, but production agents rarely send it. Measurements show what changes and how to prescribe feedback. Edward Izgorodin Wed, 12 Aug 2026 00:00:00 GMT https://mnemoverse.com/docs/library/agent-memory-deduplication https://mnemoverse.com/docs/library/agent-memory-deduplication Six agent-memory systems, read at pinned commits: hardcoded similarity thresholds, no published error rate, and no way to undo a merge. Edward Izgorodin Mon, 03 Aug 2026 00:00:00 GMT https://mnemoverse.com/docs/library/is-memory-hyperbolic https://mnemoverse.com/docs/library/is-memory-hyperbolic Is memory hyperbolic? Cognitive maps are real, but one narrow CA1 result and indirect odor evidence do not establish a universal memory geometry. Edward Izgorodin Sun, 02 Aug 2026 00:00:00 GMT https://mnemoverse.com/docs/library/independence-illusion https://mnemoverse.com/docs/library/independence-illusion AI agent consensus can be weak evidence when agents share models, sources, prompts, or memory. Count independent provenance chains instead. Edward Izgorodin Sun, 26 Jul 2026 00:00:00 GMT https://mnemoverse.com/docs/library/who-wins-when-agents-disagree https://mnemoverse.com/docs/library/who-wins-when-agents-disagree When voting, reputation, and consensus can fail for multi-agent memory — and the published alternatives: outcomes, meta-knowledge, independence, reasons. Edward Izgorodin Sat, 25 Jul 2026 00:00:00 GMT https://mnemoverse.com/docs/library/bitemporal-memory-for-ai-agents https://mnemoverse.com/docs/library/bitemporal-memory-for-ai-agents Bitemporal memory tracks valid time and transaction time — settled engineering. One reviewed agent system ships it; none of seven benchmarks tests it. Edward Izgorodin Thu, 23 Jul 2026 00:00:00 GMT https://mnemoverse.com/docs/library/provenance-in-agent-memory https://mnemoverse.com/docs/library/provenance-in-agent-memory Two agents wrote conflicting facts: a timestamp shows which came last, not whose claim should win. Record writer, evidence and derivation, and keep both. Edward Izgorodin Thu, 23 Jul 2026 00:00:00 GMT https://mnemoverse.com/docs/library/ai-memory-solutions-2026-q3 https://mnemoverse.com/docs/library/ai-memory-solutions-2026-q3 Mem0, Zep, Letta, Cognee and Supermemory compared with Mnemoverse, the publisher of this page: integration posture, self-host reality, temporal handling, and what benchmark scores hide. Edward Izgorodin Tue, 21 Jul 2026 00:00:00 GMT https://mnemoverse.com/docs/library/memory-mcp-servers-compared https://mnemoverse.com/docs/library/memory-mcp-servers-compared Thirteen memory MCP servers across knowledge-graph, vector, Markdown and SQL storage, with pricing, maintenance reality, and the local-first vs hosted decision. Edward Izgorodin Tue, 21 Jul 2026 00:00:00 GMT https://mnemoverse.com/docs/library/stateless-mcp-agent-memory https://mnemoverse.com/docs/library/stateless-mcp-agent-memory The final MCP 2026-07-28 specification removes protocol sessions, the Mcp-Session-Id header, and the handshake. Durable agent memory remains external. Edward Izgorodin Tue, 21 Jul 2026 00:00:00 GMT https://mnemoverse.com/docs/library/memory-across-the-computing-stack https://mnemoverse.com/docs/library/memory-across-the-computing-stack Types of memory in computing, mapped from DRAM and virtual memory to KV caches, model weights, context windows, and persistent agent memory. Edward Izgorodin Sun, 19 Jul 2026 00:00:00 GMT https://mnemoverse.com/docs/library/what-is-an-agent-os https://mnemoverse.com/docs/library/what-is-an-agent-os Karpathy's LLM OS, the AIOS kernel, MemGPT paging, Agno AgentOS, Windows Agent Workspace: six meanings of Agent OS and one test that tells them apart. Edward Izgorodin Sun, 19 Jul 2026 00:00:00 GMT https://mnemoverse.com/docs/research/memory-science/transactive-memory https://mnemoverse.com/docs/research/memory-science/transactive-memory Transactive memory divides the labor of remembering across a group — a directory of who knows what — mapped, with its limits, to multi-agent AI memory. Edward Izgorodin Sun, 19 Jul 2026 00:00:00 GMT https://mnemoverse.com/docs/library/shared-memory-poisoning https://mnemoverse.com/docs/library/shared-memory-poisoning Shared memory poisoning can expose many agents to one bad write. Use a three-trust matrix and consumer-side hygiene to limit the risk. Edward Izgorodin Mon, 13 Jul 2026 00:00:00 GMT https://mnemoverse.com/docs/library/agent-memory-interop-gap https://mnemoverse.com/docs/library/agent-memory-interop-gap Agent memory interoperability needs rules for identity, schema, provenance, consistency, and permissions. A2A and MCP do not define them. Edward Izgorodin Sun, 12 Jul 2026 00:00:00 GMT https://mnemoverse.com/docs/library/llm-judge-field-manual https://mnemoverse.com/docs/library/llm-judge-field-manual LLM-as-judge reliability explained: assess position bias, verbosity, self-preference, rubrics, retrieval recall, and benchmark comparability. Edward Izgorodin Sun, 12 Jul 2026 00:00:00 GMT https://mnemoverse.com/docs/library/agent-memory-consolidation https://mnemoverse.com/docs/library/agent-memory-consolidation Agent memory consolidation turns raw writes into reusable structure between uses — what works, what ships, and what remains unproven. Edward Izgorodin Sat, 11 Jul 2026 00:00:00 GMT https://mnemoverse.com/docs/library/ai-introspection-voice-not-audit https://mnemoverse.com/docs/library/ai-introspection-voice-not-audit AI introspection is real but unreliable. Chain-of-thought and self-reports cannot replace durable audit records outside the model. Edward Izgorodin Sat, 11 Jul 2026 00:00:00 GMT https://mnemoverse.com/docs/library/global-workspace-memory-outside https://mnemoverse.com/docs/library/global-workspace-memory-outside Anthropic's global workspace research reveals Claude's transient working memory, and why AI agent memory must persist outside the model. Edward Izgorodin Sat, 11 Jul 2026 00:00:00 GMT https://mnemoverse.com/docs/library/graph-memory-mcp-tools https://mnemoverse.com/docs/library/graph-memory-mcp-tools Graph memory MCP comparison of Graphiti, Cognee, Neo4j, and server-memory across traversal, temporal support, provenance, and STOP gaps. Edward Izgorodin Sat, 11 Jul 2026 00:00:00 GMT https://mnemoverse.com/docs/library/graphrag-tax https://mnemoverse.com/docs/library/graphrag-tax GraphRAG vs RAG decision guide: when a knowledge graph's build, query, and latency cost pays off for multi-hop retrieval — and when it doesn't. Edward Izgorodin Sat, 11 Jul 2026 00:00:00 GMT https://mnemoverse.com/docs/library/ontology-schema-topology-dialects https://mnemoverse.com/docs/library/ontology-schema-topology-dialects Ontology vs schema vs topology: one shared commitment, five different contracts, and what missing data means — false, unknown, or forbidden. Edward Izgorodin Sat, 11 Jul 2026 00:00:00 GMT https://mnemoverse.com/docs/library/navigating-knowledge-graphs https://mnemoverse.com/docs/library/navigating-knowledge-graphs AI agent knowledge graph traversal depends on navigation policy, read-side tools, resolution, provenance, and stopping budget — not just graph size. Edward Izgorodin Fri, 10 Jul 2026 00:00:00 GMT https://mnemoverse.com/docs/library/knowledge-graph-memory-for-agents https://mnemoverse.com/docs/library/knowledge-graph-memory-for-agents Knowledge-graph memory reframes agent memory from transcript to navigable substrate — GraphRAG, temporal graphs, PPR, and why it shouldn't rewrite itself. Edward Izgorodin Tue, 07 Jul 2026 00:00:00 GMT https://mnemoverse.com/docs/library/how-ai-agents-trust-each-other https://mnemoverse.com/docs/library/how-ai-agents-trust-each-other A2A and MCP hand identity to OAuth, OIDC and mTLS. A signed Agent Card proves origin, not trust; RFC 8693 scopes delegation; prompt injection stays open. Edward Izgorodin Mon, 06 Jul 2026 00:00:00 GMT https://mnemoverse.com/docs/library/least-privilege-ai-agents https://mnemoverse.com/docs/library/least-privilege-ai-agents Least privilege bounds what a tricked AI agent can do — not whether it's tricked: authz policy, JIT tokens, and the confused-deputy ceiling it can't cross. Edward Izgorodin Mon, 06 Jul 2026 00:00:00 GMT https://mnemoverse.com/docs/library/prompt-injection-credential-exfiltration https://mnemoverse.com/docs/library/prompt-injection-credential-exfiltration Prompt injection is a 3-stage credential kill chain: injection lands, the agent reads a secret, it leaves via an allowed channel. Three defenses matter. Edward Izgorodin Mon, 06 Jul 2026 00:00:00 GMT https://mnemoverse.com/docs/library/trust-model-spectrum-agent-secrets https://mnemoverse.com/docs/library/trust-model-spectrum-agent-secrets Six rungs of protecting an AI agent's secret, weakest to strongest—each defeats a different threat, but none stops a tricked agent misusing what it unlocks. Edward Izgorodin Sun, 05 Jul 2026 00:00:00 GMT https://mnemoverse.com/docs/library/credential-llm-never-sees https://mnemoverse.com/docs/library/credential-llm-never-sees Credential the LLM never sees: resolve secrets below the model, inject them on the wire, and account for MCP, logs, and confused deputies. Edward Izgorodin Sat, 04 Jul 2026 00:00:00 GMT https://mnemoverse.com/docs/library/memory-poisoning-secret-leak https://mnemoverse.com/docs/library/memory-poisoning-secret-leak A poisoned AI agent memory can wait weeks, then leak an API key. Persistence and key-theft are each demonstrated; chaining them isn't—yet. Here's the fix. Edward Izgorodin Sat, 04 Jul 2026 00:00:00 GMT https://mnemoverse.com/docs/library/why-ai-agents-nag-about-secrets https://mnemoverse.com/docs/library/why-ai-agents-nag-about-secrets AI agents warn about API keys because the risk is real. But warning after a secret enters context is not protection. Edward Izgorodin Mon, 29 Jun 2026 00:00:00 GMT https://mnemoverse.com/docs/library/workflow-intelligence-ships-as-a-primitive https://mnemoverse.com/docs/library/workflow-intelligence-ships-as-a-primitive OpenAI's Record & Replay for Codex and Anthropic's composable skills arrived the same week. Two stacks, one shift: encoding how work is done and making the procedure callable. The ownership question follows. Edward Izgorodin Mon, 29 Jun 2026 00:00:00 GMT https://mnemoverse.com/docs/library/ai-agent-memory https://mnemoverse.com/docs/library/ai-agent-memory AI agent memory explained: what it is, how it works, the approach families, how it is evaluated, and how to choose a real memory layer. Edward Izgorodin Tue, 23 Jun 2026 00:00:00 GMT https://mnemoverse.com/docs/research/evaluation/ai-memory-benchmarks-field-guide https://mnemoverse.com/docs/research/evaluation/ai-memory-benchmarks-field-guide A map of how AI-agent memory is actually measured — LoCoMo, LongMemEval, BEAM, long-context tests — what each checks, what it misses, and which to trust. Edward Izgorodin Tue, 23 Jun 2026 00:00:00 GMT https://mnemoverse.com/docs/research/evaluation/judges-good-and-evil https://mnemoverse.com/docs/research/evaluation/judges-good-and-evil Swap only the grading prompt on the same answers and an AI-memory benchmark score moves ~40 points — you can't trust a leaderboard without the recipe. Edward Izgorodin Tue, 23 Jun 2026 00:00:00 GMT https://mnemoverse.com/docs/library/hypergraph-vs-hyperbolic-graph https://mnemoverse.com/docs/library/hypergraph-vs-hyperbolic-graph Hypergraph vs hyperbolic graph for AI memory: one grows the edge to many vertices (n-ary); the other curves the space for hierarchy. Where Mnemoverse bets. Edward Izgorodin Mon, 22 Jun 2026 00:00:00 GMT https://mnemoverse.com/docs/library/memory-is-becoming-a-procurement-decision https://mnemoverse.com/docs/library/memory-is-becoming-a-procurement-decision Cursor's valuation, Anthropic's dreaming preview, Google's Memory Bank: different architectures, same shift. Persistent memory has stopped being a research conversation and become a choice builders can no longer defer. Edward Izgorodin Mon, 22 Jun 2026 00:00:00 GMT https://mnemoverse.com/docs/library/a2a-agent-card https://mnemoverse.com/docs/library/a2a-agent-card What an A2A Agent Card is: JSON at a well-known URL describing an agent's skills, endpoint, and auth — fetched to discover and call it. Edward Izgorodin Fri, 19 Jun 2026 00:00:00 GMT https://mnemoverse.com/docs/library/a2a-integration-howto https://mnemoverse.com/docs/library/a2a-integration-howto A2A tutorial for Python: install the SDK, publish an Agent Card, run a server, delegate a task, stream artifacts, and add domain-scoped memory. Edward Izgorodin Fri, 19 Jun 2026 00:00:00 GMT https://mnemoverse.com/docs/research/agents/context-budgeting https://mnemoverse.com/docs/research/agents/context-budgeting Context budgeting allocates finite agent tokens across system, tools, retrieval, history, outputs, and response buffer. Edward Izgorodin Fri, 19 Jun 2026 00:00:00 GMT https://mnemoverse.com/docs/research/agents/context-optimizer https://mnemoverse.com/docs/research/agents/context-optimizer Context optimization for AI agents unifies KV-cache hit rate, prefix stability, token budget, latency, cost, and placement into one runtime decision. Edward Izgorodin Fri, 19 Jun 2026 00:00:00 GMT https://mnemoverse.com/docs/library/a2a-protocol-explained https://mnemoverse.com/docs/library/a2a-protocol-explained A2A protocol explained: Agent2Agent primitives, transports, how it complements MCP, and why shared agent memory stays a separate layer. Edward Izgorodin Thu, 18 Jun 2026 00:00:00 GMT https://mnemoverse.com/docs/library/a2a-vs-mcp https://mnemoverse.com/docs/library/a2a-vs-mcp A2A vs MCP compared: MCP connects agents to tools; A2A connects agents to agents. They compose as complementary layers, not rivals — and leave memory open. Edward Izgorodin Thu, 18 Jun 2026 00:00:00 GMT https://mnemoverse.com/docs/library/hebbian-memory-for-ai-agents https://mnemoverse.com/docs/library/hebbian-memory-for-ai-agents Hebbian memory for AI agents: how learned associations, automatic query expansion, and outcome feedback differ from static vector database similarity retrieval. Edward Izgorodin Thu, 18 Jun 2026 00:00:00 GMT https://mnemoverse.com/docs/library/not-a-vector-database https://mnemoverse.com/docs/library/not-a-vector-database Is Mnemoverse a vector database? No. A vector DB does static similarity retrieval; Mnemoverse adds learning, recency, and consolidation on top (consolidation is off on the hosted service). Edward Izgorodin Thu, 18 Jun 2026 00:00:00 GMT https://mnemoverse.com/docs/library/rescorla-wagner-agent-memory https://mnemoverse.com/docs/library/rescorla-wagner-agent-memory Rescorla-Wagner for agent memory: how prediction-error outcome feedback updates valence and improves recall in Mnemoverse. Edward Izgorodin Thu, 18 Jun 2026 00:00:00 GMT https://mnemoverse.com/docs/library/shared-memory-for-multi-agent-systems https://mnemoverse.com/docs/library/shared-memory-for-multi-agent-systems Shared memory for AI agents needs explicit domains, isolation, and hierarchical reads. Protocols coordinate work; memory preserves context. Edward Izgorodin Thu, 18 Jun 2026 00:00:00 GMT https://mnemoverse.com/docs/research/agents/context-compiler-vs-orchestration https://mnemoverse.com/docs/research/agents/context-compiler-vs-orchestration Where flow control ends and window assembly begins: the boundary between orchestrator and context compiler in LLM agent systems. Edward Izgorodin Thu, 18 Jun 2026 00:00:00 GMT https://mnemoverse.com/docs/research/agents/deterministic-vs-llm-context-assembly https://mnemoverse.com/docs/research/agents/deterministic-vs-llm-context-assembly Deterministic context assembly improves cacheability and auditability; LLM-directed assembly adds adaptivity. Most agent systems need both. Edward Izgorodin Thu, 18 Jun 2026 00:00:00 GMT https://mnemoverse.com/docs/library/agent-memory-is-not-a-database https://mnemoverse.com/docs/library/agent-memory-is-not-a-database Orogat and Mansour name four failure modes of treating agent memory as storage, and propose GEM: four state-level operations instead of record-level CRUD. Why the abstraction, not the engine, is the problem. Edward Izgorodin Wed, 17 Jun 2026 00:00:00 GMT https://mnemoverse.com/docs/research/agents/context-compiler https://mnemoverse.com/docs/research/agents/context-compiler Context engineering is the discipline. The context compiler is the per-turn runtime layer that ranks, budgets, secures, and assembles each model call. Edward Izgorodin Tue, 16 Jun 2026 00:00:00 GMT https://mnemoverse.com/docs/research/memory/architectures/multimodal-memory-integration https://mnemoverse.com/docs/research/memory/architectures/multimodal-memory-integration Multimodal memory integration binds text, image, and audio into one representation. A sourced guide to binding: TPR, VSA/HDC, SDM, and Hopfield networks. Edward Izgorodin Tue, 16 Jun 2026 00:00:00 GMT https://mnemoverse.com/docs/research/memory/architectures/self-organizing-memory-systems https://mnemoverse.com/docs/research/memory/architectures/self-organizing-memory-systems Self-organizing memory systems explained: Adaptive Resonance Theory, Self-Organizing Maps, and Growing Neural Gas, including GNG-U utility-based pruning. Edward Izgorodin Tue, 16 Jun 2026 00:00:00 GMT https://mnemoverse.com/docs/library/memory-mcp https://mnemoverse.com/docs/library/memory-mcp Memory MCP servers explained: what they are, how to choose one by where data lives and what it does, and how to install so an agent remembers across sessions. Edward Izgorodin Mon, 15 Jun 2026 00:00:00 GMT https://mnemoverse.com/docs/research/ai-agent-memory-crisis https://mnemoverse.com/docs/research/ai-agent-memory-crisis AI agent memory fails three ways: statelessness, context rot, and lost-in-the-middle. Bigger context windows trade one failure for another and add token spend. Edward Izgorodin Sun, 14 Jun 2026 00:00:00 GMT https://mnemoverse.com/docs/research/mcp/federated/architecture https://mnemoverse.com/docs/research/mcp/federated/architecture MCP federation in 2026: what gateways and the June 2025 spec solved for running multiple MCP servers, and which problems, like auth propagation, remain open. Edward Izgorodin Wed, 10 Jun 2026 00:00:00 GMT https://mnemoverse.com/docs/research/memory/cognitive-models/episodic-semantic-memory https://mnemoverse.com/docs/research/memory/cognitive-models/episodic-semantic-memory Tulving's episodic vs semantic split is a design decision for AI agents: an event log beside a fact store, plus a step deciding when episodes become facts. Edward Izgorodin Wed, 10 Jun 2026 00:00:00 GMT https://mnemoverse.com/docs/research/memory-science/kinds-of-memory https://mnemoverse.com/docs/research/memory-science/kinds-of-memory Memory has dozens of named kinds because no one has drawn its boundary. A guided tour of the list, the reasons it grew, and why the seams never close. Edward Izgorodin Mon, 08 Jun 2026 00:00:00 GMT https://mnemoverse.com/docs/research/memory-science/schema-formation https://mnemoverse.com/docs/research/memory-science/schema-formation Schema theory explains how memory turns episodes into reusable structure. For AI agent memory, it clarifies why episodes and consolidation should stay separate. Edward Izgorodin Mon, 08 Jun 2026 00:00:00 GMT https://mnemoverse.com/docs/research/memory-science/working-memory https://mnemoverse.com/docs/research/memory-science/working-memory Working memory is the bounded active workspace of cognition; its capacity debate (Miller's 7 vs Cowan's 4) and models map onto AI context-window design. Edward Izgorodin Mon, 08 Jun 2026 00:00:00 GMT https://mnemoverse.com/docs/research/evaluation/deepeval-framework-deepdive https://mnemoverse.com/docs/research/evaluation/deepeval-framework-deepdive DeepEval turns LLM evals into pytest tests that fail CI builds. 50+ metrics, but most are LLM-as-a-judge (G-Eval: 0.514 vs humans) — DAG adds determinism. Edward Izgorodin Sun, 07 Jun 2026 00:00:00 GMT https://mnemoverse.com/docs/research/evaluation/huggingface-evaluate-deepdive https://mnemoverse.com/docs/research/evaluation/huggingface-evaluate-deepdive Practical Hugging Face evaluate docs: evaluate.load() + compute(predictions, references), the BLEU list-of-lists trap, and no information_retrieval metric. Edward Izgorodin Sun, 07 Jun 2026 00:00:00 GMT https://mnemoverse.com/docs/research/memory-science/bernard-widrow-cognitive-memory https://mnemoverse.com/docs/research/memory-science/bernard-widrow-cognitive-memory Bernard Widrow's 1960 LMS delta rule taught machines to learn and still runs in adaptive filtering; late in life he turned to content-addressable memory. Edward Izgorodin Sun, 07 Jun 2026 00:00:00 GMT https://mnemoverse.com/docs/research/memory-science/geoffrey-hinton-boltzmann-machine https://mnemoverse.com/docs/research/memory-science/geoffrey-hinton-boltzmann-machine Geoffrey Hinton's 1985 Boltzmann machine made memory generative: a stochastic, energy-based network with hidden units that learns a distribution and samples it. Edward Izgorodin Sun, 07 Jun 2026 00:00:00 GMT https://mnemoverse.com/docs/research/memory-science/jeff-hawkins-hierarchical-temporal-memory https://mnemoverse.com/docs/research/memory-science/jeff-hawkins-hierarchical-temporal-memory Jeff Hawkins argues the brain is a memory system for prediction; Hierarchical Temporal Memory uses sparse representations, sequences, and continual learning. Edward Izgorodin Sun, 07 Jun 2026 00:00:00 GMT https://mnemoverse.com/docs/research/agents/kv-cache-context-engineering https://mnemoverse.com/docs/research/agents/kv-cache-context-engineering How prompt-cache keys are canonicalized, why a stable prefix decides your KV-cache hit rate, and what each provider actually discounts on a cache read. Edward Izgorodin Sat, 06 Jun 2026 00:00:00 GMT https://mnemoverse.com/docs/research/evaluation/evaluating-agent-memory https://mnemoverse.com/docs/research/evaluation/evaluating-agent-memory How to evaluate AI agent memory: the framework, the dimensions, a benchmark map (LoCoMo, LongMemEval, BEAM), and why latency and cost are co-equal axes. Edward Izgorodin Sat, 06 Jun 2026 00:00:00 GMT https://mnemoverse.com/docs/research/memory-science/hopfield-associative-memory https://mnemoverse.com/docs/research/memory-science/hopfield-associative-memory Hopfield's 1982 PNAS associative memory, the 0.138N capacity limit (Amit, Gutfreund, Sompolinsky, 1985), and how attention became one read from it. Edward Izgorodin Sat, 06 Jun 2026 00:00:00 GMT https://mnemoverse.com/docs/research/evaluation/llm-as-judge-patterns https://mnemoverse.com/docs/research/evaluation/llm-as-judge-patterns LLM-as-a-judge leniency and bias explained: how a "be generous" grader, MT-Bench failures, and the LoCoMo judge shape the memory-benchmark numbers vendors cite Edward Izgorodin Fri, 05 Jun 2026 00:00:00 GMT https://mnemoverse.com/docs/research/evaluation/langchain-evaluation-deepdive https://mnemoverse.com/docs/research/evaluation/langchain-evaluation-deepdive LangChain/LangSmith evaluation explained: datasets, LLM-as-judge biases, the eval tool landscape, and the blind spot none cover — whether your agent remembers. Edward Izgorodin Thu, 04 Jun 2026 00:00:00 GMT https://mnemoverse.com/docs/research/ai-memory-landscape-2026 https://mnemoverse.com/docs/research/ai-memory-landscape-2026 AI memory landscape 2026: how persistent memory became a production discipline — platform features, funded startups, the LoCoMo benchmark, and open problems. Edward Izgorodin Mon, 06 Apr 2026 00:00:00 GMT https://mnemoverse.com/docs/research/building-memory-that-scales https://mnemoverse.com/docs/research/building-memory-that-scales Building memory that scales: the path from an early prototype to a registered benchmark protocol — how an oracle leak inflated our early LoCoMo numbers, and what we did about it. Edward Izgorodin Mon, 06 Apr 2026 00:00:00 GMT https://mnemoverse.com/docs/research/when-ai-cites-what-doesnt-exist https://mnemoverse.com/docs/research/when-ai-cites-what-doesnt-exist When AI cites what does not exist: a case study of a recombination hallucination that passed four of five checks, and why persistent memory needs verification. Edward Izgorodin Mon, 06 Apr 2026 00:00:00 GMT https://mnemoverse.com/docs/research/market-landscape https://mnemoverse.com/docs/research/market-landscape AI memory market 2025-2026: platform memory from OpenAI, Anthropic, Google, Microsoft, xAI, plus startups, funding rounds, and context-management trends. Edward Izgorodin Sat, 04 Apr 2026 00:00:00 GMT