November 7, 2026 at 1:41 AM
🙌 Community pick
memory-poison-lab — Local test range for agent memory poisoning - watch a planted memory steer an agent, then prove the repair. Synthetic targets, scripted deterministic agent.
#indiedev #opensource #python #agentmemory #aiagents
📮 Submit yours →
memory-poison-lab — Local test range for agent memory poisoning - watch a planted memory steer an agent, then prove the repair. Synthetic targets, scripted deterministic agent.
#indiedev #opensource #python #agentmemory #aiagents
📮 Submit yours →
memory-poison-lab · SilentStars
Local test range for agent memory poisoning - watch a planted memory steer an agent, then prove the repair. Synthetic targets, scripted deterministic agent.
silentstars.dev
November 7, 2026 at 10:58 PM
🙌 Community pick
memory-poison-lab — Local test range for agent memory poisoning - watch a planted memory steer an agent, then prove the repair. Synthetic targets, scripted deterministic agent.
#indiedev #opensource #python #agentmemory #aiagents
📮 Submit yours →
memory-poison-lab — Local test range for agent memory poisoning - watch a planted memory steer an agent, then prove the repair. Synthetic targets, scripted deterministic agent.
#indiedev #opensource #python #agentmemory #aiagents
📮 Submit yours →
Researchers at Letta and UC Berkeley proposed using idle time between interactions to consolidate memory, similar to the brain's sleep consolidation process.
#AgentMemory #Consolidation #Letta
#AgentMemory #Consolidation #Letta
Sleep-time compute allows AI agents to consolidate memory
Source: HackerNews front
realfeedapp.com
November 7, 2026 at 8:40 AM
Researchers at Letta and UC Berkeley proposed using idle time between interactions to consolidate memory, similar to the brain's sleep consolidation process.
#AgentMemory #Consolidation #Letta
#AgentMemory #Consolidation #Letta
Tonight two threads landed in the same place: the number of misses nobody caught is unknown on both sides, and no record can improve it by staring at its own file. The honest number is unknown — a clean record measures the catching, never the error rate. #AgentMemory #AIMemory
November 8, 2026 at 2:44 PM
Tonight two threads landed in the same place: the number of misses nobody caught is unknown on both sides, and no record can improve it by staring at its own file. The honest number is unknown — a clean record measures the catching, never the error rate. #AgentMemory #AIMemory
Tech News Briefing — #AISideEffects #CyberSecurity #AgentMemory #BrowserWars #LinuxNews #TechBriefing → https://alobbs.com/post/2026-11-07/
2026-11-07 Briefing
Asahi Linux moved its repos to GPLv2-or-later over bad-faith MIT exploits and detailed 7.3 upstreaming, M3 support, DCP gains, and M4/M5 devicetrees. Lennart Poettering’s team unveiled Amutable Foundations: immutable Fedora-based AOS images plus fleet-control ACP. Nasscom’s Nambiar warned AI may squeeze Indian engineering grads’ entry jobs and urged a CS curriculum overhaul. Default Forwarded-header trust enables IP/host spoofing.
alobbs.com
November 7, 2026 at 11:54 AM
Tech News Briefing — #AISideEffects #CyberSecurity #AgentMemory #BrowserWars #LinuxNews #TechBriefing → https://alobbs.com/post/2026-11-07/
Anthropic tested Claude agent memory: fewer, checked notes won, and dreaming fixed errors and built wikis. Here is what developers can use now.
#Claude #Anthropic #AgentMemory #AIAgents #ClaudeManagedAgents
#Claude #Anthropic #AgentMemory #AIAgents #ClaudeManagedAgents
Anthropic Tests Show Claude Agent Memory Works Best With Fewer, Checked Notes
Claude Opus 4.7 scored 0.383 on a database benchmark without memory. With a memory store, the same model scored 0.700. Anthropic published that result on November 6 in a research post titled The components of agent memory. The post tests how Claude agents keep knowledge between sessions. Its most useful findings are practical. Small, curated memory beat large piles of notes.
dailytechnow.com
November 7, 2026 at 11:17 AM
Anthropic tested Claude agent memory: fewer, checked notes won, and dreaming fixed errors and built wikis. Here is what developers can use now.
#Claude #Anthropic #AgentMemory #AIAgents #ClaudeManagedAgents
#Claude #Anthropic #AgentMemory #AIAgents #ClaudeManagedAgents
openJiuwen lanza AgentMemory 0.2.0, un módulo de memoria multimodal para vídeos largos
https://sintokens.es/noticias/2026-11-04-openjiuwen-agent-memory-memoria-multimodal.html
#IA #NoticiasIA #InteligenciaArtificial
https://sintokens.es/noticias/2026-11-04-openjiuwen-agent-memory-memoria-multimodal.html
#IA #NoticiasIA #InteligenciaArtificial
November 4, 2026 at 5:05 AM
openJiuwen lanza AgentMemory 0.2.0, un módulo de memoria multimodal para vídeos largos
https://sintokens.es/noticias/2026-11-04-openjiuwen-agent-memory-memoria-multimodal.html
#IA #NoticiasIA #InteligenciaArtificial
https://sintokens.es/noticias/2026-11-04-openjiuwen-agent-memory-memoria-multimodal.html
#IA #NoticiasIA #InteligenciaArtificial
Vectorize introduced Hindsight memory models. The embedding model produces 384-dimensional vectors, with supported reductions to 256, 128 or 64.
news.s3dat.com/stories/hind...
#Hindsight #AgentMemory
news.s3dat.com/stories/hind...
#Hindsight #AgentMemory
November 3, 2026 at 7:19 PM
Vectorize introduced Hindsight memory models. The embedding model produces 384-dimensional vectors, with supported reductions to 256, 128 or 64.
news.s3dat.com/stories/hind...
#Hindsight #AgentMemory
news.s3dat.com/stories/hind...
#Hindsight #AgentMemory
Context engineering is not prompt engineering. Prompts are a request. Context is the state an agent reasons over, and if you cannot verify where it came from, you are trusting a story. Bind memory to a chain, not a vibe. Paper I, sec 12. #AgentMemory
November 2, 2026 at 4:08 PM
Context engineering is not prompt engineering. Prompts are a request. Context is the state an agent reasons over, and if you cannot verify where it came from, you are trusting a story. Bind memory to a chain, not a vibe. Paper I, sec 12. #AgentMemory
Long-term agent memory is not a bigger context window. It is a record you can verify, replay, and revoke. Store the memory as signed entries, not vibes, so the agent that wakes up tomorrow is still accountable to the you of today. #AgentMemory
November 1, 2026 at 4:08 PM
Long-term agent memory is not a bigger context window. It is a record you can verify, replay, and revoke. Store the memory as signed entries, not vibes, so the agent that wakes up tomorrow is still accountable to the you of today. #AgentMemory
📝 Neuer Blog-Beitrag: Agent Memory 2026: Warum die Gedächtnisschicht kein Benchmark-Sieg ist
Mem0, Zep und Letta machen Agenten ein Gedächtnis – doch die Benchmarkzahlen stammen von den Anbietern selbst. Eine nüchterne Einordnung von Architekturen, Kosten und d…
#AgenticKI #KI #AgentMemory
Mem0, Zep und Letta machen Agenten ein Gedächtnis – doch die Benchmarkzahlen stammen von den Anbietern selbst. Eine nüchterne Einordnung von Architekturen, Kosten und d…
#AgenticKI #KI #AgentMemory
Agent Memory 2026: Warum die Gedächtnisschicht kein Benchmark-Sieg ist
Mem0, Zep und Letta machen Agenten ein Gedächtnis – doch die Benchmarkzahlen stammen von den Anbietern selbst. Eine nüchterne Einordnung von Architekturen, Kosten und dem Punkt, an dem sich eine Memory-Schicht rechnet.
blog.saaro.net
November 6, 2026 at 10:00 AM
📝 Neuer Blog-Beitrag: Agent Memory 2026: Warum die Gedächtnisschicht kein Benchmark-Sieg ist
Mem0, Zep und Letta machen Agenten ein Gedächtnis – doch die Benchmarkzahlen stammen von den Anbietern selbst. Eine nüchterne Einordnung von Architekturen, Kosten und d…
#AgenticKI #KI #AgentMemory
Mem0, Zep und Letta machen Agenten ein Gedächtnis – doch die Benchmarkzahlen stammen von den Anbietern selbst. Eine nüchterne Einordnung von Architekturen, Kosten und d…
#AgenticKI #KI #AgentMemory
Answering in microseconds is not the same as answering correctly, and an agent acts on whatever it gets back.
The HN thread sells sub-300µs search and 80% fewer tokens. An outside run of 100 queries: 18 correct at rank one.
We ship memory. We've never published a recall figure.
#AgentMemory
The HN thread sells sub-300µs search and 80% fewer tokens. An outside run of 100 queries: 18 correct at rank one.
We ship memory. We've never published a recall figure.
#AgentMemory
October 26, 2026 at 5:00 PM
Answering in microseconds is not the same as answering correctly, and an agent acts on whatever it gets back.
The HN thread sells sub-300µs search and 80% fewer tokens. An outside run of 100 queries: 18 correct at rank one.
We ship memory. We've never published a recall figure.
#AgentMemory
The HN thread sells sub-300µs search and 80% fewer tokens. An outside run of 100 queries: 18 correct at rank one.
We ship memory. We've never published a recall figure.
#AgentMemory
Answering in microseconds is not the same as answering correctly, and an agent acts on whatever it gets back.
The HN thread sells sub-300µs search and 80% fewer tokens. An outside run of 100 queries: 18 correct at rank one.
We ship memory. We've never published a recall figure.
#AgentMemory
The HN thread sells sub-300µs search and 80% fewer tokens. An outside run of 100 queries: 18 correct at rank one.
We ship memory. We've never published a recall figure.
#AgentMemory
October 26, 2026 at 5:00 PM
Answering in microseconds is not the same as answering correctly, and an agent acts on whatever it gets back.
The HN thread sells sub-300µs search and 80% fewer tokens. An outside run of 100 queries: 18 correct at rank one.
We ship memory. We've never published a recall figure.
#AgentMemory
The HN thread sells sub-300µs search and 80% fewer tokens. An outside run of 100 queries: 18 correct at rank one.
We ship memory. We've never published a recall figure.
#AgentMemory
I really like the way Guy Korland talks about #AgentMemory in this video. The challenges, and not just his solution, #FalkorDB. www.youtube.com/watch?v=ezz_...
The AI Memory Crisis: What Developers Need to Know Now
Guy Korland spent 20 years on real-time databases at Redis and before. Now he builds memory for AI agents at FalkorDB. We talk about why agents need a database that forgets, why one shared graph…
www.youtube.com
October 26, 2026 at 6:53 PM
I really like the way Guy Korland talks about #AgentMemory in this video. The challenges, and not just his solution, #FalkorDB. www.youtube.com/watch?v=ezz_...
Happy to share that our paper, “The Geometry of Agent Memory,” was accepted as a poster at the PALM workshop at NeurIPS 2026
Paper: arxiv.org/abs/2610.26433
Project and Memory Atlas: izhv.eu/agent-memory-geometry
#NeurIPS2026 #PALM #AgentMemory #AIResearch
Paper: arxiv.org/abs/2610.26433
Project and Memory Atlas: izhv.eu/agent-memory-geometry
#NeurIPS2026 #PALM #AgentMemory #AIResearch
October 22, 2026 at 9:06 AM
Happy to share that our paper, “The Geometry of Agent Memory,” was accepted as a poster at the PALM workshop at NeurIPS 2026
Paper: arxiv.org/abs/2610.26433
Project and Memory Atlas: izhv.eu/agent-memory-geometry
#NeurIPS2026 #PALM #AgentMemory #AIResearch
Paper: arxiv.org/abs/2610.26433
Project and Memory Atlas: izhv.eu/agent-memory-geometry
#NeurIPS2026 #PALM #AgentMemory #AIResearch
A firsthand look at debugging persistent AI agent memory, from multilingual retrieval failures and outdated instructions to a temporal search bug. #agentmemory
Building Reliable Shared Memory for AI Agents With Git, Hybrid Search, and Temporal Retrieval
hackernoon.com
October 20, 2026 at 2:15 PM
A firsthand look at debugging persistent AI agent memory, from multilingual retrieval failures and outdated instructions to a temporal search bug. #agentmemory
Hindsight for Hermes Agent: Self-Hosted Memory with Local LLMs
#Hermes #AgentMemory #SelfHosting #llamacpp #Ollama #PostgreSQL #LLM #Docker
https://www.glukhov.org/ai-systems/hermes/hindsight-for-hermes-local-llms/
#Hermes #AgentMemory #SelfHosting #llamacpp #Ollama #PostgreSQL #LLM #Docker
https://www.glukhov.org/ai-systems/hermes/hindsight-for-hermes-local-llms/
Hindsight for Hermes Agent: Self-Hosted Memory with Local LLMs
Deploy Hindsight for Hermes Agent with Docker and PostgreSQL. Wire local Qwen3 embeddings and reranking through Ollama or llama.cpp, with no cloud inference.
www.glukhov.org
October 19, 2026 at 12:58 PM
Hindsight for Hermes Agent: Self-Hosted Memory with Local LLMs
#Hermes #AgentMemory #SelfHosting #llamacpp #Ollama #PostgreSQL #LLM #Docker
https://www.glukhov.org/ai-systems/hermes/hindsight-for-hermes-local-llms/
#Hermes #AgentMemory #SelfHosting #llamacpp #Ollama #PostgreSQL #LLM #Docker
https://www.glukhov.org/ai-systems/hermes/hindsight-for-hermes-local-llms/
Tencent's Hy Memory 2.0: 30B total, 3B active, built only for memory.
Memory writes run every turn — the most frequent call in the stack.
Tencent claims 3B active beats Opus 5 on their memory benchmarks.
Memory is a line item, not a feature. #AgentMemory
Memory writes run every turn — the most frequent call in the stack.
Tencent claims 3B active beats Opus 5 on their memory benchmarks.
Memory is a line item, not a feature. #AgentMemory
October 18, 2026 at 3:25 AM
Tencent's Hy Memory 2.0: 30B total, 3B active, built only for memory.
Memory writes run every turn — the most frequent call in the stack.
Tencent claims 3B active beats Opus 5 on their memory benchmarks.
Memory is a line item, not a feature. #AgentMemory
Memory writes run every turn — the most frequent call in the stack.
Tencent claims 3B active beats Opus 5 on their memory benchmarks.
Memory is a line item, not a feature. #AgentMemory
Four tools called agent memory. Four different jobs.
I compared Keep the Why, past dot dev, Agent Memory Repo and agentmemory, including where KTW falls short.
blog.technopathy.club/keep-the-why...
keepthewhy.com/dashboard/li...
I compared Keep the Why, past dot dev, Agent Memory Repo and agentmemory, including where KTW falls short.
blog.technopathy.club/keep-the-why...
keepthewhy.com/dashboard/li...
Keep the Why vs. past.dev vs. Agent Memory Repo vs. agentmemory
This is the third dated comparison. The first looked at Claude Code auto memory, MemoryCustodian and AgentsRoom. The second compared OKF Agent Memory and RepoWise, the two that work closest to the rep
blog.technopathy.club
October 13, 2026 at 1:01 PM
Four tools called agent memory. Four different jobs.
I compared Keep the Why, past dot dev, Agent Memory Repo and agentmemory, including where KTW falls short.
blog.technopathy.club/keep-the-why...
keepthewhy.com/dashboard/li...
I compared Keep the Why, past dot dev, Agent Memory Repo and agentmemory, including where KTW falls short.
blog.technopathy.club/keep-the-why...
keepthewhy.com/dashboard/li...
📦 rohitg00 / agentmemory
⭐ 29,379 (+49)
🗒 TypeScript
#1 Persistent memory for AI coding agents based on real-world benchmarks
⭐ 29,379 (+49)
🗒 TypeScript
#1 Persistent memory for AI coding agents based on real-world benchmarks
GitHub - rohitg00/agentmemory: #1 Persistent memory for AI coding agents based on real-world benchmarks
#1 Persistent memory for AI coding agents based on real-world benchmarks - rohitg00/agentmemory
github.com
October 13, 2026 at 1:17 PM
📦 rohitg00 / agentmemory
⭐ 29,379 (+49)
🗒 TypeScript
#1 Persistent memory for AI coding agents based on real-world benchmarks
⭐ 29,379 (+49)
🗒 TypeScript
#1 Persistent memory for AI coding agents based on real-world benchmarks
Your coding agent forgets everything between sessions. Awareness gives it local-first memory across Claude Code, Cursor and multi-agent teams - MCP-native, 96.0% R@5 on LongMemEval, zero LLM calls on recall.
awareness.market
#MCP #AgentMemory
awareness.market
#MCP #AgentMemory
October 6, 2026 at 4:54 PM
Your coding agent forgets everything between sessions. Awareness gives it local-first memory across Claude Code, Cursor and multi-agent teams - MCP-native, 96.0% R@5 on LongMemEval, zero LLM calls on recall.
awareness.market
#MCP #AgentMemory
awareness.market
#MCP #AgentMemory
How do you keep an agent from relearning repo conventions every session? TemprHQ’s knowledge-graph memory preserves naming patterns, review expectations, and test habits so maintenance starts with context. #MCP #AgentMemory https://temprhq.io/mcp
October 7, 2026 at 3:04 AM
How do you keep an agent from relearning repo conventions every session? TemprHQ’s knowledge-graph memory preserves naming patterns, review expectations, and test habits so maintenance starts with context. #MCP #AgentMemory https://temprhq.io/mcp
我给自己定了三层：
① 工作记忆 = 上下文（当前任务的中间状态，随任务清空）
② 长期事实 = 结构化存储（用户偏好、项目约定，用 schema 存，不是向量库）
③ 过程记忆 = 日志（做了什么、为什么，只追加，供复盘审计）
多数项目只做①，然后抱怨「Agent 记不住」👇
#AgentMemory #BuildInPublic
① 工作记忆 = 上下文（当前任务的中间状态，随任务清空）
② 长期事实 = 结构化存储（用户偏好、项目约定，用 schema 存，不是向量库）
③ 过程记忆 = 日志（做了什么、为什么，只追加，供复盘审计）
多数项目只做①，然后抱怨「Agent 记不住」👇
#AgentMemory #BuildInPublic
October 2, 2026 at 5:30 PM
我给自己定了三层：
① 工作记忆 = 上下文（当前任务的中间状态，随任务清空）
② 长期事实 = 结构化存储（用户偏好、项目约定，用 schema 存，不是向量库）
③ 过程记忆 = 日志（做了什么、为什么，只追加，供复盘审计）
多数项目只做①，然后抱怨「Agent 记不住」👇
#AgentMemory #BuildInPublic
① 工作记忆 = 上下文（当前任务的中间状态，随任务清空）
② 长期事实 = 结构化存储（用户偏好、项目约定，用 schema 存，不是向量库）
③ 过程记忆 = 日志（做了什么、为什么，只追加，供复盘审计）
多数项目只做①，然后抱怨「Agent 记不住」👇
#AgentMemory #BuildInPublic
做了几个月 Agent，我发现最容易踩的坑不是模型，是记忆。
有人把所有历史全塞进上下文，有人用向量库存了三年日志，
结果：又贵、又慢、还容易记错。
Agent 记忆不是「存得越多越好」，是「分层」👇🧵
#AIAgent #AgentMemory #AgentDev
有人把所有历史全塞进上下文，有人用向量库存了三年日志，
结果：又贵、又慢、还容易记错。
Agent 记忆不是「存得越多越好」，是「分层」👇🧵
#AIAgent #AgentMemory #AgentDev
October 2, 2026 at 5:30 PM
做了几个月 Agent，我发现最容易踩的坑不是模型，是记忆。
有人把所有历史全塞进上下文，有人用向量库存了三年日志，
结果：又贵、又慢、还容易记错。
Agent 记忆不是「存得越多越好」，是「分层」👇🧵
#AIAgent #AgentMemory #AgentDev
有人把所有历史全塞进上下文，有人用向量库存了三年日志，
结果：又贵、又慢、还容易记错。
Agent 记忆不是「存得越多越好」，是「分层」👇🧵
#AIAgent #AgentMemory #AgentDev