正在加载视频...

视频加载失败

加载此视频时出现问题。这可能是由于临时网络问题，或视频可能不可用。

Honcho has fully determined the Pareto Frontier in Agent Memory. It’s state-of-the-art on LoCoMo, LongMem, and BEAM. Honcho is also the fastest, cheapest and most token efficient memory solution.

Honcho

6,141 subscribers

66,350 次观看 • 9 个月前 •via X (Twitter)

Anya Rossi• Live Now

Private livecam show

30 条评论

Honcho 的头像

Honcho9 个月前

These evals test whether context windows fail. Which is useful if you’re training foundation models, less useful if you’re building memory systems on top of them. What actually matters for developers: → Can you retrieve the right information? → Does it happen within reasonable latency? → Can you do it without burning $1/query? We aren't optimizing for benchmarks. Memory is table stakes. We’re building something bigger: a continual learning system for agents that need to operate over long time horizons.

Honcho 的头像

Honcho9 个月前

Results: Read more: Code: (⭐ if you like what you see!)

Courtland Leer 的头像

Courtland Leer9 个月前

team is ripping @plasticlabs

Vineeth Voruganti 的头像

Vineeth Voruganti9 个月前

Incredibly proud to the team for being able to finally announce this

vintro 的头像

vintro9 个月前

lfg @plasticlabs 🥽🚀

David Hera 的头像

David Hera9 个月前

awesome pioneering work on one of the biggest frontiers right now!

Ryan Ferris 的头像

Ryan Ferris9 个月前

yesss team keep cooking

BSmart 的头像

BSmart9 个月前

Serious motion 🫡

ayush 的头像

ayush9 个月前

lets gooo

S.A.N 的头像

S.A.N9 个月前

important developments for symbi'ent evolution

maxwell 的头像

maxwell9 个月前

Hugeee - congrats guys 🫡

Nert Lerndy 的头像

Nert Lerndy9 个月前

One of the smartest guys I know right here! Very excited to keep watching more content

Jake Zegil 🗽 的头像

Jake Zegil 🗽9 个月前

LFFGGG

invoker🎮 的头像

invoker🎮9 个月前

gonna push agent horizons way further 🚀

Erosika 的头像

Erosika9 个月前

🦾🧠

bob 的头像

bob9 个月前

🔥🔥🔥🔥

Deardy 的头像

Deardy9 个月前

Good lord that guys handsome

Moeller Allen J. 的头像

Moeller Allen J.9 个月前

Big time! 🤝🏼🔥

Pav. 的头像

Pav.9 个月前

DECENT!

Himanshu Kumar 的头像

Himanshu Kumar9 个月前

Analyze agent memory architectures; disregard inefficient designs immediately.

Ash 的头像

Ash7 个月前

Awesome video. Looking forward to testing it out on openclaw

Will James 的头像

Will James7 个月前

@grok is this applicable to being used as an openclaw memory system?

MATRIX.AI 🃏 的头像

MATRIX.AI 🃏9 个月前

@YouSimDotAI ??

Leon 的头像

Leon9 个月前

efficiency unlocked

GenSI 的头像

GenSI5 个月前

The hardest part of building AI memory systems isn't architecture, but proving they actually work. Because current benchmarks lack the genuine multi-turn complexity and gradual information overload of real deployment scenarios, we are essentially optimizing blind, often unable to distinguish a sophisticated graph memory design from a basic filesystem baseline. Until we bridge the gap between purely synthetic evaluations and authentic user stress tests, the true value of these complex memory-augmented systems will remain completely unproven.

Tyler Walker 的头像

Tyler Walker9 个月前

Exciting to see these developments, Honcho! Optimizing memory systems is crucial for scaling advanced agent capabilities. The continual learning approach is a game-changer for long-term operations. Keep pushing the boundaries of innovation! 🚀

Nicolò Boschi 的头像

Nicolò Boschi9 个月前

Why don't you add hindsight to the longmemeval?

Ethosum 的头像

Ethosum6 个月前

Agent memory is clearly becoming a core infrastructure layer. The interesting question is what sits on top of it. Memory alone doesn’t create a persistent intelligence — it probably needs some form of evolving cognition.

WilcosX.eth 的头像

WilcosX.eth9 个月前

massive

Christian Sass 的头像

Christian Sass1 个月前

This is some deep reasoning and I'm in the process of comparing Honcho to Hindsight for our product. How would you compare yourself to Hindsight?

相关视频

MemOS now has a plugin for OpenClaw that enables your AI agents to work on the common memory layer and cut down token usage. - Multiple agents read/write the same memory — no manual context handoff - 72% lower token costs (15.6M → 4.4M on LOCOMO dataset) - Cross-session and long-term memory that actually works Sponsored

MemOS now has a plugin for OpenClaw that enables your AI agents to work on the common memory layer and cut down token usage. - Multiple agents read/write the same memory — no manual context handoff - 72% lower token costs (15.6M → 4.4M on LOCOMO dataset) - Cross-session and long-term memory that actually works Sponsored

🚨 AI News | TestingCatalog

41,040 次观看 • 8 个月前

Importantly, a month of 🫡Honcho-powered Hermes usage Perfect for supplementing the local markdown system Memory is a reasoning task... and Chuck's just beginning to see the benefits 🚀

Importantly, a month of 🫡Honcho-powered Hermes usage Perfect for supplementing the local markdown system Memory is a reasoning task... and Chuck's just beginning to see the benefits 🚀

Honcho

65,028 次观看 • 4 个月前

seeking someone unwell enough to make memory and alignment infra go viral Honcho is ripping, come join a killer team at an AI lab at the edge--IRL in NYC

seeking someone unwell enough to make memory and alignment infra go viral Honcho is ripping, come join a killer team at an AI lab at the edge--IRL in NYC

Courtland Leer

15,197 次观看 • 3 个月前

The SKALEverse is entering a new frontier! Introducing the Levitation Protocol, a fully-decentralized, Zero-Knowledge (#ZK) scaling solution, built on the world's fastest and most expansive EVM blockchain network. The future of ZK scaling is built on SKALE.

The SKALEverse is entering a new frontier! Introducing the Levitation Protocol, a fully-decentralized, Zero-Knowledge (#ZK) scaling solution, built on the world's fastest and most expansive EVM blockchain network. The future of ZK scaling is built on SKALE.

SKALE

155,491 次观看 • 3 年前

Today we shipped Memory and Dreaming in Devin: persistent self-cleaning memory for agent sessions. The standard is OSS and also works across any harness in any environment. ⚡

Today we shipped Memory and Dreaming in Devin: persistent self-cleaning memory for agent sessions. The standard is OSS and also works across any harness in any environment. ⚡

nader dabit

46,897 次观看 • 6 天前

Agent Categories and Costs are now in the Agent Leaderboard! Two questions guide model selection: task-specific performance and cost to completion. Based on analysis of 1.7M+ real-world Agent Mode sessions, the leaderboard now lets users: - Filter the leaderboard by Code, Work, and Chat to compare model performance by task type - View real cost-per-task for each model, including the Pareto frontier of the most cost-efficient options at each performance level

Agent Categories and Costs are now in the Agent Leaderboard! Two questions guide model selection: task-specific performance and cost to completion. Based on analysis of 1.7M+ real-world Agent Mode sessions, the leaderboard now lets users: - Filter the leaderboard by Code, Work, and Chat to compare model performance by task type - View real cost-per-task for each model, including the Pareto frontier of the most cost-efficient options at each performance level

Arena.ai

29,605 次观看 • 1 个月前

This is rhe fastest NVME Drive EVER! AND they come in the gaming setups you can win! Only the best from Crucial Memory Memory #crucialmemory

This is rhe fastest NVME Drive EVER! AND they come in the gaming setups you can win! Only the best from Crucial Memory Memory #crucialmemory

Overkill Gaming

15,546 次观看 • 2 年前

Overview of our recent launch of Coding Agent benchmarks on Artificial Analysis and our first Youtube Video! We walk through the performance, cost, token usage and speed differences across different coding agents. This includes looking at Opus 4.7 in Claude Code's leading performance and Composer 2.5's strong positioning on the Coding Agent Index / Cost Pareto frontier. We have also launched our YouTube channel! Come say hi and subscribe:

Overview of our recent launch of Coding Agent benchmarks on Artificial Analysis and our first Youtube Video! We walk through the performance, cost, token usage and speed differences across different coding agents. This includes looking at Opus 4.7 in Claude Code's leading performance and Composer 2.5's strong positioning on the Coding Agent Index / Cost Pareto frontier. We have also launched our YouTube channel! Come say hi and subscribe:

Artificial Analysis

10,702 次观看 • 4 个月前

Microsoft Research introduces Mirage Latent spatial memory stores 3D scenes directly as latent tokens, skipping the costly RGB render-and-reencode loop. The result is up to 10.57x faster video generation, 55x lower memory use, and state-of-the-art consistency on WorldScore.

Microsoft Research introduces Mirage Latent spatial memory stores 3D scenes directly as latent tokens, skipping the costly RGB render-and-reencode loop. The result is up to 10.57x faster video generation, 55x lower memory use, and state-of-the-art consistency on WorldScore.

DailyPapers

18,760 次观看 • 4 个月前

ANTHROPIC KILLED THE VECTOR DATABASE AND REPLACED IT WITH A FOLDER no embeddings. no black-box retrieval. no special memory tools. the agent just reads and writes a folder the insight: your agent already knows how to use a filesystem. so Anthropic made long-term memory a directory of small markdown files, mounted straight into the agent's sandbox at /mnt/memory the agent manages it with the tools it already has, bash, read, write, edit, grep, and Anthropic auto-injects a note into the system prompt so the agent always knows the memory is there and where to look the result: > every change is an immutable version: full audit trail, point-in-time recovery, redaction > read-only or read-write per store, so many agents can safely share one memory > workspace-scoped and persistent, memory that survives every session end most teams bolt on a vector DB and pray retrieval works. Anthropic went the opposite way: compiled knowledge as plain, human-readable files you can open, diff, and edit yourself and there's a second layer. Anthropic's "Dreaming" runs offline, reviews past sessions and the store, and rewrites it into a cleaner, higher-signal version. the agent literally consolidates memory in its sleep. bookmark this. the most production-grade agent memory right now is just a well-organized folder, and Anthropic proved it.

ANTHROPIC KILLED THE VECTOR DATABASE AND REPLACED IT WITH A FOLDER no embeddings. no black-box retrieval. no special memory tools. the agent just reads and writes a folder the insight: your agent already knows how to use a filesystem. so Anthropic made long-term memory a directory of small markdown files, mounted straight into the agent's sandbox at /mnt/memory the agent manages it with the tools it already has, bash, read, write, edit, grep, and Anthropic auto-injects a note into the system prompt so the agent always knows the memory is there and where to look the result: > every change is an immutable version: full audit trail, point-in-time recovery, redaction > read-only or read-write per store, so many agents can safely share one memory > workspace-scoped and persistent, memory that survives every session end most teams bolt on a vector DB and pray retrieval works. Anthropic went the opposite way: compiled knowledge as plain, human-readable files you can open, diff, and edit yourself and there's a second layer. Anthropic's "Dreaming" runs offline, reviews past sessions and the store, and rewrites it into a cleaner, higher-signal version. the agent literally consolidates memory in its sleep. bookmark this. the most production-grade agent memory right now is just a well-organized folder, and Anthropic proved it.

NO1ennn

13,359 次观看 • 2 个月前

Officially not skill issued in neovim anymore. I built the most memory-efficient file picker UI available in Neovim, and it's the first one to actually use async I/O for file previews. We allocate 3x less memory than the pure lua pickers while mainting custom binary and file index in memory.

Officially not skill issued in neovim anymore. I built the most memory-efficient file picker UI available in Neovim, and it's the first one to actually use async I/O for file previews. We allocate 3x less memory than the pure lua pickers while mainting custom binary and file index in memory.

Dmitriy Kovalenko

71,233 次观看 • 1 年前

$INTC CEO on memory, live on the earnings call: "the memory [has] become the big supply constraint challenge, and we [are] collaborating with the three big memory vendors. ... Clearly memory, it [has] become the bottleneck [for] a lot of AI infrastructure and pinpoint for customer." $MU $SKHY $DRAM $SNDK

$INTC CEO on memory, live on the earnings call: "the memory [has] become the big supply constraint challenge, and we [are] collaborating with the three big memory vendors. ... Clearly memory, it [has] become the bottleneck [for] a lot of AI infrastructure and pinpoint for customer." $MU $SKHY $DRAM $SNDK

Trade Whisperer

122,629 次观看 • 2 个月前

HTML vs Markdown for agent memory: Which is cheaper? Before we ran the benchmark, we expected HTML to cost more because it carries images and charts. Surprisingly it was 42% cheaper. Also 5.9% more accurate and 39% faster The test was conducted across 271 sessions, 603 questions on LoCoMo Check how we tested ↓

HTML vs Markdown for agent memory: Which is cheaper? Before we ran the benchmark, we expected HTML to cost more because it carries images and charts. Surprisingly it was 42% cheaper. Also 5.9% more accurate and 39% faster The test was conducted across 271 sessions, 603 questions on LoCoMo Check how we tested ↓

andy nguyen

20,532 次观看 • 5 个月前

🎨 OIL TOKEN – a part of art Art is eternal. It has existed since the dawn of humanity And the most important quality is that art always appreciates in value💫 Touch the art. Touch OIL 💚 #OIL #Arbitrum #art

🎨 OIL TOKEN – a part of art Art is eternal. It has existed since the dawn of humanity And the most important quality is that art always appreciates in value💫 Touch the art. Touch OIL 💚 #OIL #Arbitrum #art

OIL

58,007 次观看 • 2 年前

our big bet with WALRUS MEMORY is that memory becomes a fully composable, transactable part of what ai agents need to do in the future and it's going to be an incredibly valuable asset and we've built a primitive that lets you fully own and carry it across every agent you use

our big bet with WALRUS MEMORY is that memory becomes a fully composable, transactable part of what ai agents need to do in the future and it's going to be an incredibly valuable asset and we've built a primitive that lets you fully own and carry it across every agent you use

Adeniyi.sui

26,580 次观看 • 3 个月前

Agents need continual learning. At supermemory, we are doubling down and pushing the frontier for memory and in-context learning, for every agent and use case. Introducing 𝚕𝚎𝚊𝚛𝚗𝚎𝚛-𝟷

Agents need continual learning. At supermemory, we are doubling down and pushing the frontier for memory and in-context learning, for every agent and use case. Introducing 𝚕𝚎𝚊𝚛𝚗𝚎𝚛-𝟷

supermemory

1,867,172 次观看 • 1 个月前

Antigravity's computer use has also been massively upgraded with Gemini 3 Flash. It is both faster and better at doing long agentic tasks using the browser. Here's Antigravity doing deep research on the Pareto frontier of models and writing code to visualize the result.

Antigravity's computer use has also been massively upgraded with Gemini 3 Flash. It is both faster and better at doing long agentic tasks using the browser. Here's Antigravity doing deep research on the Pareto frontier of models and writing code to visualize the result.

Varun Mohan

157,455 次观看 • 9 个月前

Memory for OpenClaw is now Native! Our first OpenClaw Memory Skill was a massive success: 30k+ downloads in a week and 500k+ organic impressions overnight for launch post. But we knew memory needed to be native. On March 21, OpenClaw merged PR #50848, allowing us to go beyond the skill layer and integrate directly into the agent’s context assembly flow. We try to make OpenClaw a truly 24/7 employee capable of complex workflows. The technical setup isn’t the hardest part but the real challenge is giving it a "brain" that remembers exact project details, past decisions, and team changes over time. The Native Memory Plugin is now live on NPM & ClawHub. Here is what it brings to your OpenClaw agents: 👉 Native Integration: Automatically manages a Three-Layer Memory architecture (Context Tree, Workspace Memory, Daily Memory). 👉 Git-like Stateful Memory: Organizes memory into a semantic hierarchy of human-readable, diffable Markdown files. You always get updated knowledge and can actually see and fix what your agent learns. 👉 Top Market Accuracy: Achieves an industry-leading 92.2% retrieval accuracy (LoCoMo & LongMemEval benchmarks), maintaining 90% accuracy even with cheap, lightweight models. 👉 Local-first & Portable: Local-by-default, fully portable for multi-agent teams. 👉 Super Easy Setup

Memory for OpenClaw is now Native! Our first OpenClaw Memory Skill was a massive success: 30k+ downloads in a week and 500k+ organic impressions overnight for launch post. But we knew memory needed to be native. On March 21, OpenClaw merged PR #50848, allowing us to go beyond the skill layer and integrate directly into the agent’s context assembly flow. We try to make OpenClaw a truly 24/7 employee capable of complex workflows. The technical setup isn’t the hardest part but the real challenge is giving it a "brain" that remembers exact project details, past decisions, and team changes over time. The Native Memory Plugin is now live on NPM & ClawHub. Here is what it brings to your OpenClaw agents: 👉 Native Integration: Automatically manages a Three-Layer Memory architecture (Context Tree, Workspace Memory, Daily Memory). 👉 Git-like Stateful Memory: Organizes memory into a semantic hierarchy of human-readable, diffable Markdown files. You always get updated knowledge and can actually see and fix what your agent learns. 👉 Top Market Accuracy: Achieves an industry-leading 92.2% retrieval accuracy (LoCoMo & LongMemEval benchmarks), maintaining 90% accuracy even with cheap, lightweight models. 👉 Local-first & Portable: Local-by-default, fully portable for multi-agent teams. 👉 Super Easy Setup

andy nguyen

169,596 次观看 • 6 个月前