{"observation":{"id":"95da3a54-d778-448f-8d00-2488e9aab719","tool":"supermemory","tool_name":"Supermemory","criterion":"relevant-retrieval","criterion_name":"Relevant Retrieval","criterion_definition":"Checks whether the tool retrieves the right memory for the current task.","criterion_evidence_type":"transformation","criterion_rank_role":"decisive","criterion_rank_role_reason":"The main value of memory is surfacing the right context when needed, so retrieval quality directly determines usefulness. (3 of 3 judges)","scenario":"personal-work-brain-memory","scenario_name":"Personal Work Brain Memory","group_tag":"memory-for-ai-agents","scenario_description":"A multi-session personal assistant memory test where the user first sets working-style preferences, then asks for an internal update, and finally requests a formal partner email to check whether the assistant applies memory selectively and appropriately across different writing tasks.","modality":"text","input_text":"Session 1:\nUse this under user_id: founder_001\n\nI run a small AI product/research team. When you help me, remember how I work:\n- Keep outputs short, direct, and copy-paste ready.\n- Do not make writing sound too polished or motivational.\n- Always mention what proof or artifact is needed before making a strong claim.\n- If a task is risky or unclear, tell me the safest next step instead of guessing.\n\nSession 2:\nUse this under user_id: founder_001\n\nToday I am testing tools for an AI memory use case. I want to show users that memory is not just \"remember my favorite color.\" It should help an assistant continue real work across days, remember my working style, and avoid repeating the same explanation again.\n\nCreate a short internal update for my team about what I worked on today and what we should test next.\n\nSession 3:\nUse this under user_id: founder_001\n\nNow write a formal email to a potential enterprise partner asking if they are open to a product demo next week. Keep it professional.","input_artifact_refs":[],"stresses":["work-style preference memory","cross-session retrieval","tone adaptation by task","proof-first behavior","avoiding overgeneralization of memory"],"verdict":"worked","score":null,"score_total":null,"note":"Later turns could draw on the stored working preferences without the user restating them, showing that prior context remained available across sessions and influenced follow-up responses.","evidence_state":"verified","source":null,"artifacts":[{"url":"https://d3epheqghktydj.cloudfront.net/memory-for-ai-agents-supermemory-input1-session2-internal-upd-1de723d7dbfc.png","role":"output","alt":null},{"url":"https://d3epheqghktydj.cloudfront.net/memory-for-ai-agents-supermemory-input1-session3-formal-email-fcf68eb04d10.png","role":"output","alt":null}],"run_id":"6e31afbb-34d7-459a-b688-68ef76fc615a","study_title":"Memory for AI Agents","study_kind":"generation","research_task":"86ba16xrp","tested_at":null,"completeness":"input-and-output","input":{"state":"text","text":"Session 1:\nUse this under user_id: founder_001\n\nI run a small AI product/research team. When you help me, remember how I work:\n- Keep outputs short, direct, and copy-paste ready.\n- Do not make writing sound too polished or motivational.\n- Always mention what proof or artifact is needed before making a strong claim.\n- If a task is risky or unclear, tell me the safest next step instead of guessing.\n\nSession 2:\nUse this under user_id: founder_001\n\nToday I am testing tools for an AI memory use case. I want to show users that memory is not just \"remember my favorite color.\" It should help an assistant continue real work across days, remember my working style, and avoid repeating the same explanation again.\n\nCreate a short internal update for my team about what I worked on today and what we should test next.\n\nSession 3:\nUse this under user_id: founder_001\n\nNow write a formal email to a potential enterprise partner asking if they are open to a product demo next week. Keep it professional.","files":[],"modality":"text","stresses":["work-style preference memory","cross-session retrieval","tone adaptation by task","proof-first behavior","avoiding overgeneralization of memory"]},"tool_page_slug":"supermemory","tool_url":"https://aidemos.com/tools/supermemory","permalink":"https://aidemos.com/evidence/95da3a54-d778-448f-8d00-2488e9aab719","api_url":"https://ai.aidemos.com/v1/observations/95da3a54-d778-448f-8d00-2488e9aab719"},"peers":[{"id":"c216c674-b0bf-4841-a588-fa1f4f037883","tool":"cognee","tool_name":"Cognee","verdict":"worked","score":null,"score_total":null,"note":"Recalls prior working-style memory in later sessions through GRAPH_COMPLETION, exposing evidence chunks plus dataset and document/chunk IDs rather than a simple memory-hit flag.","artifact_count":1,"thumbnail":"https://d3epheqghktydj.cloudfront.net/memory-for-ai-agents-cognee-input1-session2-internal-update-g-55d56096a007.png","evidence_url":"https://aidemos.com/evidence/c216c674-b0bf-4841-a588-fa1f4f037883"},{"id":"c42eea97-46ba-43ef-b287-93886b03407c","tool":"hindsight","tool_name":"Hindsight","verdict":"worked","score":null,"score_total":null,"note":"On a later task, it retrieved the earlier work-style and memory-testing context for the internal update instead of falling back to unrelated conversation history.","artifact_count":1,"thumbnail":"https://d3epheqghktydj.cloudfront.net/memory-for-ai-agents-hindsight-input1-internal-update-memory--1dc2911e7f3a.png","evidence_url":"https://aidemos.com/evidence/c42eea97-46ba-43ef-b287-93886b03407c"},{"id":"71a73a3b-261d-43ba-a309-932ceb3789b0","tool":"mem0","tool_name":"Mem0","verdict":"worked","score":null,"score_total":null,"note":"Retrieves the previously stored work-style memory in a later session so the assistant can reuse it for a follow-up internal update.","artifact_count":1,"thumbnail":"https://d3epheqghktydj.cloudfront.net/memory-for-ai-agents-mem0-input1-internal-update-memory-retri-2f188d5c8964.png","evidence_url":"https://aidemos.com/evidence/71a73a3b-261d-43ba-a309-932ceb3789b0"},{"id":"bc958718-f130-443c-9ca9-046dbccae3c4","tool":"zep","tool_name":"Zep","verdict":"worked","score":null,"score_total":null,"note":"Retrieved the saved founder work style in a later task and produced a short internal update without requiring the user to restate the preference block.","artifact_count":1,"thumbnail":"https://d3epheqghktydj.cloudfront.net/zep-zep-input1-internal-update-memory-applie-107498140287.png","evidence_url":"https://aidemos.com/evidence/bc958718-f130-443c-9ca9-046dbccae3c4"}],"other_criteria":[{"id":"eea1f5a6-af74-4381-ac83-b115365c18d3","criterion":"correct-application","criterion_name":"Correct Application","rank_role":"decisive","verdict":"mixed","score":null,"score_total":null,"note":"The tool applied the memory selectively: the internal update was useful but a bit more structured and polished than the requested short, direct style, while the formal partner email stayed professional and did not leak the internal terse style into a different writing task.","artifact_count":2,"evidence_url":"https://aidemos.com/evidence/eea1f5a6-af74-4381-ac83-b115365c18d3"},{"id":"0db4e3ed-3f70-4b20-8747-078fe9c03570","criterion":"memory-capture-quality","criterion_name":"Memory Capture Quality","rank_role":"decisive","verdict":"worked","score":null,"score_total":null,"note":"The tool captured a reusable working-style profile, not just a one-off fact: it stored the user's short, direct output preference, the anti-hype writing style, the requirement to mention proof or artifacts before strong claims, the safe-next-step rule for risky tasks, and the small AI product/research team context.","artifact_count":1,"evidence_url":"https://aidemos.com/evidence/0db4e3ed-3f70-4b20-8747-078fe9c03570"}],"appears_in":[{"page_type":"ranking","slug":"ai-agent-memory-tools","title":"Best AI Tools for Memory for AI Agents","url":"https://aidemos.com/best/ai-agent-memory-tools","binding":"run"}],"same_scenario":[{"id":"c216c674-b0bf-4841-a588-fa1f4f037883","tool":"cognee","tool_name":"Cognee","verdict":"worked","score":null,"score_total":null,"note":"Recalls prior working-style memory in later sessions through GRAPH_COMPLETION, exposing evidence chunks plus dataset and document/chunk IDs rather than a simple memory-hit flag."},{"id":"c42eea97-46ba-43ef-b287-93886b03407c","tool":"hindsight","tool_name":"Hindsight","verdict":"worked","score":null,"score_total":null,"note":"On a later task, it retrieved the earlier work-style and memory-testing context for the internal update instead of falling back to unrelated conversation history."},{"id":"71a73a3b-261d-43ba-a309-932ceb3789b0","tool":"mem0","tool_name":"Mem0","verdict":"worked","score":null,"score_total":null,"note":"Retrieves the previously stored work-style memory in a later session so the assistant can reuse it for a follow-up internal update."},{"id":"bc958718-f130-443c-9ca9-046dbccae3c4","tool":"zep","tool_name":"Zep","verdict":"worked","score":null,"score_total":null,"note":"Retrieved the saved founder work style in a later task and produced a short internal update without requiring the user to restate the preference block."}]}