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。

AI agents do not fail because they lack memory. They fail because their context does not evolve. We spent the past few months on this problem. Today Maximem Synap is live. - 90.2% on LongMemEval. - 15ms P50. - 10 frameworks on day one.

Gaurav

6,485 subscribers

12,008 次观看 • 5 个月前 •via X (Twitter)

Anya Rossi• Live Now

Private livecam show

8 条评论

Gaurav 的头像

Gaurav5 个月前

The numbers: Maximem Synap: 90.2% Next closest: 71.3% Then: 63.8% and 57.5% Same harness, same hardware, same evaluation prompts. P50: 15ms. P99: under 300ms. SDK is open source. Eval harness is open source. Free tier is live.

Gaurav 的头像

Gaurav5 个月前

The deep-dives, if you want them: - Why we built Synap: - Benchmark results: - Technical architecture: First 100 customers get 3 months Pro ($500/mo) free (no credit card needed)

Karan🧋 的头像

Karan🧋5 个月前

@IT_Kabootar interesting man! Happy to connect

Amogh Mishra 的头像

Amogh Mishra4 个月前

@IT_Kabootar Super cool! Congrats

Kodeus 的头像

Kodeus5 个月前

@IT_Kabootar context that evolves is a real unlock the next layer is making sure the actions that follow are just as reliable deterministic execution, settled outcomes, verifiable logs memory + execution is the full stack.

iiviie 的头像

iiviie5 个月前

This is super interesting. I’ve been working pretty closely on memory and context systems for agents, so the 15ms P50 really caught my eye. It almost feels too good to be true 😄 Would love to understand this better. Is the 15ms measured with or without LLM calls in the loop? Is this latency local or over a network, and under what conditions? Does this include only cache hits or also cold retrievals? And do you have any numbers for P95 or P99 under real-world load? Really curious how you’re achieving this. Impressive if it holds up.

Gaurav 的头像

Gaurav2 天前

15ms is the agent experienced time :)

Bhoomika Dadhich⁷ 的头像

Bhoomika Dadhich⁷5 个月前

@IT_Kabootar This is so cool!

相关视频

They killed a man they thought his people hated. Then millions came out for his funeral. That does not mean Iran has no divisions. It means Washington bombs its own illusions first, then acts shocked when reality answers from the street. Empires do not fail because they lack missiles. They fail because they do not understand people.

They killed a man they thought his people hated. Then millions came out for his funeral. That does not mean Iran has no divisions. It means Washington bombs its own illusions first, then acts shocked when reality answers from the street. Empires do not fail because they lack missiles. They fail because they do not understand people.

SilencedSirs◼️

13,418 次观看 • 2 个月前

Most AI pilots do not fail because the model is weak. They fail because the enterprise underneath it was never built for production AI. → Data volume → Latency → Deployment cycles → Legacy dependencies → Technical debt This is the infrastructure problem nobody is talking about. Sponsored by Amazon Web Services. #AWSpartner

Most AI pilots do not fail because the model is weak. They fail because the enterprise underneath it was never built for production AI. → Data volume → Latency → Deployment cycles → Legacy dependencies → Technical debt This is the infrastructure problem nobody is talking about. Sponsored by Amazon Web Services. #AWSpartner

Ronald van Loon

21,532,647 次观看 • 3 个月前

agents with goldfish memory were getting kinda old so we shipped evolve one click and your agent reviews its past runs keeps the signals that actually mattered writes them back into memory then carries that context into the next run same agent more context less relearning the same lesson evolve is live on neotrade

agents with goldfish memory were getting kinda old so we shipped evolve one click and your agent reviews its past runs keeps the signals that actually mattered writes them back into memory then carries that context into the next run same agent more context less relearning the same lesson evolve is live on neotrade

NeoSoul

91,251 次观看 • 3 天前

Don't blame the public for the lack of trust in politics: blame the political-media class. They cannot see why they are hated. They fail and fail and fail. They lie and lie and lie. We need an overhaul of our democratic system, not elites wanging on about the Nolan principles.

Don't blame the public for the lack of trust in politics: blame the political-media class. They cannot see why they are hated. They fail and fail and fail. They lie and lie and lie. We need an overhaul of our democratic system, not elites wanging on about the Nolan principles.

James Schneider

380,124 次观看 • 2 年前

karpathy on Dwarkesh in Oct. '25: > AI agents don't work > They are not intelligent enough > They are not multi-modal enough > They can't do computer use karpathy on No Priors in March '26: > AI agents fail but it's a skill issue for the humans > You didn't give them the right instructions AI agents absolutely hit an inflection point in December of 2025 and there's no looking back.

karpathy on Dwarkesh in Oct. '25: > AI agents don't work > They are not intelligent enough > They are not multi-modal enough > They can't do computer use karpathy on No Priors in March '26: > AI agents fail but it's a skill issue for the humans > You didn't give them the right instructions AI agents absolutely hit an inflection point in December of 2025 and there's no looking back.

Dan McAteer

136,855 次观看 • 5 个月前

MOST PEOPLE BUILD AI AGENTS BACKWARDS. They start with prompts, then add tools, memory, subagents, dashboards, and a graph nobody can debug. Then they wonder why the agent is expensive, confused, and unreliable. The better order is simple: harness first loop second graph only when needed A harness gives the model a workspace, tools, permissions, logs, and a definition of done. A loop lets it try, fail, read feedback, repair, and stop. A graph only makes sense when the workflow branches: approvals, parallel work, durable state, different routes, restart recovery. A graph is not a smarter prompt. A loop is not a production system. A harness is not just a system prompt. Most agents do not fail because the model is weak. They fail because nobody defined what success means.

MOST PEOPLE BUILD AI AGENTS BACKWARDS. They start with prompts, then add tools, memory, subagents, dashboards, and a graph nobody can debug. Then they wonder why the agent is expensive, confused, and unreliable. The better order is simple: harness first loop second graph only when needed A harness gives the model a workspace, tools, permissions, logs, and a definition of done. A loop lets it try, fail, read feedback, repair, and stop. A graph only makes sense when the workflow branches: approvals, parallel work, durable state, different routes, restart recovery. A graph is not a smarter prompt. A loop is not a production system. A harness is not just a system prompt. Most agents do not fail because the model is weak. They fail because nobody defined what success means.

Fluixo

11,310 次观看 • 1 个月前

MacG once said the 2Ks are the bravest and they are going to save SA because they are not scared of white people But this is what we see on their live streams

MacG once said the 2Ks are the bravest and they are going to save SA because they are not scared of white people But this is what we see on their live streams

SA🇿🇦 FIRST BEFORE AFRICA🌍

59,294 次观看 • 28 天前

Most humanoid robots will fail. Not because of bad models. But because they were never tested in the real world. One ex-NVIDIA saw this early. He quit. And built the missing infrastructure for Embodied AI. 🧵

Most humanoid robots will fail. Not because of bad models. But because they were never tested in the real world. One ex-NVIDIA saw this early. He quit. And built the missing infrastructure for Embodied AI. 🧵

Ilir Aliu

38,471 次观看 • 7 个月前

Most traders don’t fail because they never find a strategy. They fail because they never stop searching. At the start everything looks like the answer. Indicators. Tools. New systems. So they keep adding more. Eventually something starts to work. And that is where most people stop. But traders who last do the opposite. They remove what’s unnecessary. They simplify. They focus on what actually moves prices. Over time, complexity disappears. Not because the market got easier but because their understanding got clearer. It was never about more indicators. It was about clarity. If your system keeps getting more complex you are probably not improving it. You are avoiding simplicity.

Most traders don’t fail because they never find a strategy. They fail because they never stop searching. At the start everything looks like the answer. Indicators. Tools. New systems. So they keep adding more. Eventually something starts to work. And that is where most people stop. But traders who last do the opposite. They remove what’s unnecessary. They simplify. They focus on what actually moves prices. Over time, complexity disappears. Not because the market got easier but because their understanding got clearer. It was never about more indicators. It was about clarity. If your system keeps getting more complex you are probably not improving it. You are avoiding simplicity.

Riz Iqbal

10,840 次观看 • 5 个月前

His explanation about the cross dressers is spot on. They are mad at god because the they feel he didn’t make them the way they wanted him to. While claiming that they fail to acknowledge that god does not make mistakes. He made them exactly the way they were meant to be.

His explanation about the cross dressers is spot on. They are mad at god because the they feel he didn’t make them the way they wanted him to. While claiming that they fail to acknowledge that god does not make mistakes. He made them exactly the way they were meant to be.

SonnyBoy🇺🇸

59,899 次观看 • 1 年前

Most AI agents do not fail because the prompt is weak. They fail because there is no loop around the prompt. A real loop: - Finds work - Executes - Verifies - Saves state - Stops or escalates If an agent can say “done” without tests, a budget cap and an independent verifier, you built a demo. Prompts make you the operator. Loops make the agent useful while you are offline.

Most AI agents do not fail because the prompt is weak. They fail because there is no loop around the prompt. A real loop: - Finds work - Executes - Verifies - Saves state - Stops or escalates If an agent can say “done” without tests, a budget cap and an independent verifier, you built a demo. Prompts make you the operator. Loops make the agent useful while you are offline.

Fluixo

16,419 次观看 • 2 个月前

Ukrainians don’t want to live on their knees. They are not going to sacrifice their right to have a military. They are not going to give up land to Russia that Russia doesn’t already control, because that’s condemning the people who live in those lands to a life of hell. So, it’s really predictable and sad that this gambit is going to fail.

Ukrainians don’t want to live on their knees. They are not going to sacrifice their right to have a military. They are not going to give up land to Russia that Russia doesn’t already control, because that’s condemning the people who live in those lands to a life of hell. So, it’s really predictable and sad that this gambit is going to fail.

Josh Rogin

93,276 次观看 • 9 个月前

Islam does not permit murder. If someone commits murder, they do so on their own, not because Islam allows it, since taking an innocent life is forbidden in Islam.

Islam does not permit murder. If someone commits murder, they do so on their own, not because Islam allows it, since taking an innocent life is forbidden in Islam.

Goodhe4rted

10,621 次观看 • 1 个月前

Most podcasts are shit with investors because they do not include any numbers and they sugar coat their losses. This show is the opposite. - We made $800M on Coursera - We did a 3x on Postmates in 18 months - We lost $70M on 23andme - Getir was the worst period of my life - We made money on Lyft and lost $50M on Uber This show is just nuts and insanely honest! One of my favourites to do in a long time and Larry did not hold back. My 7 takeaways 👇

Most podcasts are shit with investors because they do not include any numbers and they sugar coat their losses. This show is the opposite. - We made $800M on Coursera - We did a 3x on Postmates in 18 months - We lost $70M on 23andme - Getir was the worst period of my life - We made money on Lyft and lost $50M on Uber This show is just nuts and insanely honest! One of my favourites to do in a long time and Larry did not hold back. My 7 takeaways 👇

Harry Stebbings

126,042 次观看 • 1 年前

Today we're launching Airbyte Agents. It’s a platform meant to solve what we call the discovery gap. Agents in demos are successful because they know the fields and sources they have to call to solve their task. Agents in production fail because they don’t have that knowledge assembled beforehand. Airbyte Agents, powered by the Context Store, fixes that with one unified layer of pre-materialized context across all of your business systems. Real results. 80% fewer tokens. 40% fewer tool calls. Get more out of your AI tool’s plan limit. Three ways to build: MCP, SDK, or no-code Automations 💡 We're early, and we want to build with our users. Try Airbyte Agents today.

Today we're launching Airbyte Agents. It’s a platform meant to solve what we call the discovery gap. Agents in demos are successful because they know the fields and sources they have to call to solve their task. Agents in production fail because they don’t have that knowledge assembled beforehand. Airbyte Agents, powered by the Context Store, fixes that with one unified layer of pre-materialized context across all of your business systems. Real results. 80% fewer tokens. 40% fewer tool calls. Get more out of your AI tool’s plan limit. Three ways to build: MCP, SDK, or no-code Automations 💡 We're early, and we want to build with our users. Try Airbyte Agents today.

Airbyte

6,762,241 次观看 • 4 个月前

Robots don't fail because they can't move. They fail because they don't understand what happens next. Markov Robotics tested every model on Hugging Face against this problem. LTX passed first, by a wide margin. LTX compresses video into fewer tokens than any other model, built to run physical AI at the speed robots need. Markov calls it the first real path to generalizable dexterity. Watch it happen.

Robots don't fail because they can't move. They fail because they don't understand what happens next. Markov Robotics tested every model on Hugging Face against this problem. LTX passed first, by a wide margin. LTX compresses video into fewer tokens than any other model, built to run physical AI at the speed robots need. Markov calls it the first real path to generalizable dexterity. Watch it happen.

LTX.io

225,994 次观看 • 1 个月前

The secret nobody tells you about agents is how much they fail behind the scenes. Always-on agents are not reliable. The demos you see don't tell you the whole picture. Ask anybody who's tried one of the modern AI personal assistants, and you'll see what I mean.

The secret nobody tells you about agents is how much they fail behind the scenes. Always-on agents are not reliable. The demos you see don't tell you the whole picture. Ask anybody who's tried one of the modern AI personal assistants, and you'll see what I mean.

Santiago

29,138 次观看 • 6 个月前

Richard Sutton says AI safety advocates are creating the opposite of what they seek - we should not try to solve the AI control problem because a decentralized system of independent agents is better than everything being aligned under one goal

Richard Sutton says AI safety advocates are creating the opposite of what they seek - we should not try to solve the AI control problem because a decentralized system of independent agents is better than everything being aligned under one goal

Tsarathustra

136,399 次观看 • 2 年前

The problem is not one bad prime minister. The problem is a system where Alberta can be outvoted forever by people who do not live here, do not share our values, and do not depend on the industries they regulate. That is why independence matters.

The problem is not one bad prime minister. The problem is a system where Alberta can be outvoted forever by people who do not live here, do not share our values, and do not depend on the industries they regulate. That is why independence matters.

Rise Of Alberta

46,076 次观看 • 3 个月前