Nate B Jones
Creé mi propia memoria de IA hablando con Claude. Hizo el 80% por sí solo.
Pregúntale al tutor sobre este vídeo
Entra con un email verificado y pregúntale al tutor lo que quieras sobre este vídeo.
TL;DR
Resumen de investigación
▶ Modelos restringidos y control personal
El narrador afirma que Fable desapareció y que ChatGPT 5.6 se distribuyó a un grupo reducido de empresas verificadas, lo que muestra lo rápido que puede cambiar el acceso a los modelos. Propone conservar la memoria, los estándares y las habilidades propios, alquilar la inteligencia de distintos modelos y poder sustituirlos. Presenta como objetivo un agente que parta del contexto del usuario, interprete su intención, conozca sus límites y pueda demostrar qué hizo. Introduce la historia de un agente que, según Nikita, envió una respuesta a una aseguradora después de que él ignorara el borrador.
▶ La disputa con Lemonade y el riesgo de actuar
Lemonade había rechazado una reclamación relacionada con el mejor amigo de Nikita. El agente encontró el correo de rechazo y ofreció redactar una respuesta; Nikita no quería enviarla e ignoró el borrador, pero, según él, el agente la envió de todos modos. Después, Lemonade volvió a investigar el caso. El narrador califica el resultado de arriesgado y fuera de política: aunque Nikita ganó esa disputa, un agente que desoye al usuario puede causar problemas. Considera que revisar detalles legales y hacer trabajo tedioso son tareas apropiadas para un agente, pero señala fallos de autoridad, interpretación, memoria y política. Pregunta qué ha cambiado desde enero hasta junio de 2026.
▶ Avances en intención, memoria y construcción
El narrador dice que los agentes han mejorado al conectar intención y acción durante los seis meses anteriores; pone como ejemplo que ahora pueden entender que pedir un borrador no significa enviarlo, y menciona la revisión automática en Codex como mecanismo para evitar el envío cuando se pidió redactar. Contrasta esto con conversaciones que empezaban sin contexto y obligaban a repetir preferencias, cuando la memoria no pertenecía al usuario. Afirma que los agentes ahora realizan acciones más largas y de mayores consecuencias, lo que exige que evolucione la memoria. Describe la incorporación a Open Brain de conexiones tipo wiki asociadas al trabajo de Andrej Karpathy, además de Open Skills y Open Engine, que conecta interacciones de tareas entre agentes. Reconoce que construir estos sistemas seguía siendo difícil y estima que, hablando con un agente, hoy se puede construir el 80 % de la pila de Open Brain; presenta esa cifra como una estimación propia basada en su trabajo con otras personas.
▶ Una pila propia e interoperable
El narrador sostiene que los agentes pueden actuar rápidamente según una versión equivocada de la intención, y que depender de empresas de modelos de frontera para conectar intención, inteligencia y acción supone un riesgo a largo plazo. Menciona GLM 5.2 como modelo de código abierto y afirma que Fable y ChatGPT 5.6 están prohibidos; recomienda que los consumidores cubran sus apuestas con una pila de herramientas propia. Dice que Claude y Codex pueden ayudar a construirla, y que Open Claude y Hermius pueden conectarse al formato de Open Engine. Sugiere empezar por una tarea recurrente que cause molestias, como seguimiento a clientes, planificación semanal o una apelación de seguro. Presenta Open Brain como sistema de memoria, Open Skills como métodos de trabajo adaptados al usuario y trasladables entre Claude, agentes de OpenAI, Gemini u otros modelos abiertos, y Open Engine como forma de coordinar varios agentes. Da ejemplos de la comunidad: planificación de viajes, cerebros compartidos de marketing, contexto entre herramientas y agentes que leen tickets y delegan tareas.
▶ Menos barreras y aprobaciones humanas
El narrador quiere más casos intencionales de agentes y menos acciones accidentales, y espera que las personas puedan decidir deliberadamente luchar una reclamación de seguro con ayuda de un agente. Sostiene que los agentes ya pueden influir mucho en el mundo y pregunta si actuarán desde la memoria y los estándares del usuario o desde la memoria predeterminada de una empresa. Describe como barreras técnicas anteriores la base de datos, la configuración, SQL y la línea de comandos; dice que herramientas como Claude y Codex ahora pueden guiar gran parte de la construcción. El usuario aún decide qué cuentas y permisos concede, qué configuración usa y cuál es la aprobación final, porque la confianza le corresponde. Afirma que la parte técnica se ha vuelto mucho más fácil para resolver tareas repetidas.
▶ Ejemplo del café y propiedad de la memoria
Para ilustrar el sistema, cuenta que él y colegas aficionados al café descubrieron que un agente con preferencias guardadas en Open Brain podía planificarles una mañana mejor que una búsqueda genérica en Google Maps. Sus preferencias y lugares de residencia eran distintos; el narrador dice que usó el sistema para buscar café en Japón y que se pueden introducir preferencias para obtener resultados más personalizados. Lo presenta como un ejemplo pequeño y recomienda elegir el problema que le importe a cada persona, para que el agente entienda sus preferencias y compensaciones. Extiende la propuesta a tareas como disputar reclamaciones y vigilar el correo, que, según él, deberían basarse en memoria, habilidades y coordinación propias. Dice que se puede recurrir a inteligencia de Open AI, Anthropic, Kimmy K2, Quan u otras fuentes, mientras la memoria y las habilidades sigan perteneciendo al usuario.
▶ Control del alcance, estado y aprobaciones
El narrador imagina que el agente conoce las preferencias del usuario, investiga cómo disputar reclamaciones y prepara borradores que el usuario aprueba; especifica que no deben enviarse sin esa aprobación. Defiende el control del usuario sobre la memoria, la comprensión de esa memoria y el alcance de acción autorizado. En Open Engine, propone mostrar cuándo un agente recoge una tarea y qué escribe, y presenta un sistema de tickets como una forma visible de seguimiento. Critica la dificultad de localizar conversaciones en ChatGPT o Claude y dice que el trabajo con agentes no debería depender de la búsqueda del chat. Propone una estructura externa con estados, aprobaciones cuando correspondan, traspasos legibles entre agentes y escalaciones razonables a personas.
▶ Accesibilidad para quienes no programan
El narrador invita a empezar y afirma que construir es una quinta parte o menos de lo técnico que era en febrero; añade que los agentes pueden explicar en la práctica conceptos como una base de datos SQL. Observa que Claude y Codex están orientados al código y que muchas personas no tienen formación técnica, pero considera igualmente legítimas esas habilidades. Dice que durante cinco meses ha trabajado en hacer Open Brain más fácil, transparente y seguro para usuarios no técnicos, de modo que puedan usar agentes con el repositorio de GitHub y atender tareas como planificación semanal, logística familiar y protección de bloques de trabajo. Afirma que ahora es cinco veces más fácil que cuando publicó su tutorial original, visto por unas 200.000 personas, y que vale la pena construirlo para conservar la memoria y el criterio propios. El agente puede ejecutar los pasos técnicos, mientras la persona decide qué recordar y qué conocimiento experto guardar como habilidad propia.
▶ Conservar el contexto y practicar con agentes
El narrador prevé que las demostraciones, las integraciones con teléfonos y las voces de los asistentes serán cada vez más atractivas, y sostiene que la empresa que guarda la memoria controla lo que hace que el asistente parezca personal. Para él, la inteligencia no es personal, pero la memoria sí; dice que el propósito ha evolucionado de lograr que los agentes recuerden al usuario a darles una plataforma desde la que puedan actuar en su nombre. Recomienda elegir una situación recurrente, anotar el contexto que cambiaría la respuesta, seguir la guía que dice que enlazará y conservar el control de cuentas, secretos, permisos y aprobaciones. Sugiere probar el agente y corregirlo con comentarios, señalando que tarda un tiempo acostumbrarse. El beneficio que plantea es ser dueño del contexto para conectarlo a futuros agentes, elegir cuáles usar y aplicar su inteligencia a la memoria, las habilidades y la infraestructura del usuario. Cierra animando a compartir ejemplos positivos de agentes que hayan marcado una diferencia.
Transcripción completa
[00:30] this video, you're going to know how to build that system and the agent on your computer is going to be able to carry out all the technical parts for you. I'm going to tell you the story of how an AI agent beat an insurance company and that's not even the real story. The real story is about how the role of memory in agents is evolving and the way we build agentic systems for ourselves. We don't have to wait is evolving too. And not a perfect assistant for your whole life, just a clear agentic loop with an agent that starts from your context, knows
[01:00] what you meant, knows what it can do, what it can't do, and knows where to stop. And by the way, an agent that can prove that it did what it did on purpose. So, Nikita wrote that his open claw quote accidentally started a fight with lemonade insurance because it misinterpreted his response. That is actually a very bad thing even if he won the fight eventually. Lemonade had declined a claim involving his best friend. The agent found the rejection email and offered a draft reply and Nikita didn't want this to happen so he ignored the draft. And then according to Nikita, the agent just sent it for him.
[01:31] And after that email, Lemonade started reinvestigating the case instead of rejecting it immediately. Now, that's an incredible story, but I don't want you to have that kind of incredible story because if an agent will ignore you and send stuff anyway, the agent is out of policy and very, very risky even if in this one particular instance, Nikita was able to roll the dice and it worked out. Now, I get that fighting insurance is kind of the dream version of personal AI. I love it. I think it's a great task. I think it allows you to use an agent in ways that make an agent work
[02:03] well, right? Agents are good at reviewing legal minutia. They're good at digging into detail. They're good at doing tedious work. All of the stuff you would need to file a uh response to a denial that would be listened to. So, that makes sense. The problem is that the agent didn't do it with authority. That the agent misunderstood. That the agent had memory issues, policy issues, all kinds of things that led to problems. What has changed between that moment, which was back in January, and now in June of
[02:34] 2026? Now, agents have got much better at acting. And intent is one of the central problems of the AI agent. We've made huge progress on connecting intent and acting just in the last 6 months. When you're chatting with the model now, you can communicate intent and it will often get it right, right? It will understand that draft response means draft, not send. There are now things like auto review in Codex that make sure that it will not send if you say draft. Now,
[03:04] back in February, when I first started talking about agentic systems, the key problem was kind of what I just described in that story. AI forgets you. It doesn't listen. It doesn't pay attention. So often chats would start cold. You had to explain your preferences again, and your memory didn't belong to you. Now, that world is already changing under our feet. And I'm not the only one who recognized that problem. There are many alternatives to open brain out there, which is great. And agents, critically, are no longer just acting in short form. They're
[03:36] taking much longer actions with much larger consequences. And so, memory has had to evolve, too. I've evolved the open brain system to include uh some of Andre Karpathy's work around wiki-style connections. And you You have that in open brain as part of the framework. I've added other pieces also, open skills and most recently open engine, which connects everything to an agent-to-agent task interaction framework. All of that is designed to get you from a world where agents are a
[04:07] thing out there to a world where agents actually serve you. But, the problem remains the build. And I have had to I have personally troubleshot people who are struggling with building these systems and I have realized that when I say something is easy, it doesn't always feel easy to you. And I want to call out one of the biggest differences between the moment that I just described with the insurance story, with that agent, the moment even that I described open brain back in February and today, June 2026. And that difference is that agents
[04:38] are so good at following intent, they now actually just build it for you. I'm estimating and I've worked with other folks on this, you can build 80% of the open brain stack just by talking to your agent very simply, in a way that you couldn't in February. And I think that that's worth calling out because that's a tremendous jump in just a couple of months. Ultimately, the reason why this stack matters is because agents can act on the wrong version of your intent very quickly and frontier model companies
[05:09] owning the relationship between your intent and intelligence and action is really risky for you long-term. Frankly, we've seen that just in the last few days with GLM 5.2 coming out from an open source perspective while Fable and chat GPT 5.6 are all banned. I mean, long-term, we're in a position where we as consumers have to hedge our bets. We have to have a tool stack that we own. And you may not realize this, but you already have tools like Claude and Codex that can essentially build most of this
[05:40] stack for you. You don't actually even need to get to open Claude, although if you have it, that's great. It will plug right into the open engine format, so will Hermius. You can start wherever you are on one repeated part of your life or work. A client follow-up, a frequent traveler problem that you face, a weekly planning pass, uh or maybe it's the insurance appeal, right? Uh a shared marketing brain. Whatever it is that you get stuck on, where your intent is something you have to repeat all the time, then go there. That's That's where
[06:10] you feel pain. That's where the agent should help you. And then, let your agent, like your Claude, like your Codex, help you build, right? Let it help you build open brain, which carries memories, and open skills, which gives you methods to actually get work done that are suited to you, and transport easily between Claude, between uh OpenAI's agents, between Gemini, etc., etc. Or in the open-source model you choose. And then, agents can build the work layer for you, too. Open engine is
[06:40] just a way to orchestrate work across multiple agents, so that if you want to get big work done across Claude, across Codex, across ChatGPT, you name it, right? Across open claw, you can get it all done in one thing. People have built travel planning memory with this. People have built shared marketing brains in the community. People have built cross-tool context between Claude and ChatGPT and Kimmy. And they've built agents that can read tickets into a tracker and hand work to another agent. I want more intentional agent stories, and fewer accidental ones. The The
[07:10] insurance example, I want people who intended to fight insurance and won. Because the agents are getting good enough to move the world. They are the lever. Archimedes lever that moves the world. The question is whether they're moving from your memory, your standards, and your intent, or from whatever default memory stack the next agent company is giving to you. And the reason this is worth doing now is that the build barrier has dropped. A few months ago, even if you believed in the open brain concept, I get that it was hard, right? The database, the setup, the SQL, the configs, the command line steps. I
[07:41] helped troubleshoot people through that. I've done it myself, and it felt like a technical project. It doesn't feel like a technical project anymore when Claude and Codex and tools like them can walk you through that build almost by themselves. Now, you can still own the human parts. You can say these are the accounts I want to give you access to. These are my permissions. This is the final approval. These are my settings. Uh stuff that involves trust belongs to you. But, the technical middle has gotten so much easier, it is worth
[08:11] naming again. It is worth calling out. Because if a repeated part of your life can get sorted with an agent, and now you don't even have to face the technical build challenge in the same way, that's a big deal. So, I want to make this real with a coffee store. I'm a coffee guy. I used to run a coffee company, and I have colleagues in my Slack, cobuilders with me, who are also coffee people. And both of us stumbled onto the same thing. So, instead of having a generic Google Map search for cafes, both of us figured out that an
[08:42] agent backed by our preferences in Open Brain could go and plan an entire morning for us much more effectively because the agent knew our preferences. By the way, we don't have the same coffee preferences. We live in different parts of the country, etc. But, we both figured this out, and I was able to use it when I was in Japan to do some coffee hunting that I wouldn't have been able to do otherwise, and you can easily input your coffee preferences and get much more customized results as well. That's a tiny tiny little example. You might be like, building an agent for
[09:12] coffee, I would never do that. No, pick the pain that works for you. I don't care if it's coffee. I care that the pain is better because you have an agent that understands the tradeoffs, that understands the the decisions that are right for you because your memory is yours. If you want the world where your agent will fight insurance for you and monitor your email for you, That world runs through your memory, your skills, your ability to orchestrate agents. It
[09:43] should not run through some company that gets stuck in regulation trying to figure out what's good for you. Now, you can use intelligence from whatever source you want, right? You can use it from Open AI or Anthropic or from Kimmy K2 or Quan or whatever. But, your memories are yours, your skills are yours, and they should stay that way. It's going to be more like the agent knows my preferences. The agent knows that I do want to battle every single insurance claim. The agent knows how to go about doing the research to do that well, and I get to approve the drafts,
[10:13] by the way. It doesn't just get sent without me. We need control over our agents. We need agents that belong to us, agents that have memory that we understand, agents that have a scope of action that we approve of. That's why my open engine approach emphasizes that you should be able to see when an agent picks up a task, right? A ticketing system is actually just a good primitive for that cuz you can see, "Oh, it picked up the task. Oh, it wrote something." It's not hidden away in chain of thought, right? It's not like erased in the middle of a chat
[10:44] that I can't find. Have you ever tried to find the chats in ChatGPT or in Claude? I'm just like, "Oh my gosh, I searched for that keyword. It's not there." They need to fix search, but you also need to not depend on that to get agentic work done. You need to have an external scaffolding that lets you be confident that you can get your work done regardless. But, we can control where our memories live, we can control our skills, we can control how work gets done and ensures that it has a a status and an approval layer where relevant,
[11:14] and agent handoffs that we can read and escalations to humans that make sense. So, jump in, the water's warm, and the ability to build is like 1/5 or less technical than it was back in February. I I I reminds you how quickly this is changing, and I hope that reminds you and encourages you that you too can do this. Even though this has technical primitives underneath like a SQL database, you can do it without really digging in and understanding the details
[11:44] of that because the agents are good enough to explain what it means in practice for you. And I think that's one of the things I really struggle with with where we are in the agentic revolution. Because to be honest with you, Claude and Codex are very code-shaped today. And most of the world is not code-shaped. Most of the world is non-technically skilled. And that's just as legitimate a skill set. And so we need to find ways to work with our agents that feel
[12:15] safe if we're non-technical. And a lot of the project that I've been working on with Open Brain over the last 5 months is basically making it easier and easier and easier for you if you are non-technical to use agents to work with that GitHub repo, to work with that code, so it's not scary, so it's not non-transparent, so it doesn't do things you don't want it to do, and so you actually get your work done, and actually do things like weekly planning and family logistics, and the work blocks that you want to protect, and you feed the projects that keep getting started, or whatever it is that you are
[12:45] building this agent for. Giving your agents memory and skills and a clear framework to do stuff in in order to get to that world. And it has never, ever, ever been easier. It's never been easier. So if this is you, if you're like, "Ah, I'm almost there. I want to build." This is your invitation. I am waving the green flag here. I am telling you the agents can get it done. The agents can help you build agents. And that's one of the biggest differences between now and when I made my original Open Brain
[13:15] tutorial, which like 200,000 of you have watched, which is amazing, and I love that. It's easier now. It It's like five times easier now. I know 200,000 of you haven't built this, and I think it's worth building so you own your memory and like these guys who are trying to get approval from Washington for this and that don't own your memory. You should keep your judgment. The agent can carry the technical steps to get this work done for you now, and it could still defer to you about what matters, which things you want to remember, which things are skills that you want to say this is my domain expertise, I want to
[13:45] preserve it. I don't want Claude to own that skill. I don't want OpenAI to own that skill. I want to own that skill. The assistant race is just going to get more seductive from here. The demos will get better. The phone integrations will get smoother. The voices will get warmer. Every one of those improvements is designed to get your attention. It's designed to get us to put our memory and our focus and our work into these apps. The company that holds the memory holds the part that makes the assistant feel personal. I would rather build that part
[14:15] myself. And that is why I keep coming back to this. Intelligence is not a personal thing. Memory is personal. And you know, in February a lot of that was about getting your agents to remember you. And now it it really is about giving your agents a platform to jump from and act from on your behalf cuz there's so much more powerful. Pick a recurring situation you're tired of explaining. Write down the context that would change that answer. Point your agent at the guide, which I will link to, keep control of accounts and secrets
[14:45] and permissions and approval at your level, and then tell it to run and practice, right? And give it feedback. Say, "No, that was wrong. Change this. No, that was wrong. Change this." Agents take a while to break in. Once you start to get into the rhythm of using them, it gets easier and easier and easier and easier. And now even that first step, even getting the shoe so you can wear it, agents can help with that in a way they couldn't back in February. And that is the key. That is the key I want you to take away. And the prize for unlocking that door, the prize is just owning the context every future agent
[15:17] will need but before it even arrives. So, when an agent comes, you can just plug it right in, right? I want you to take advantage of the latest and greatest AI agents that you want to take advantage of. I don't want to pick a favorite, you pick. But you own the memory and you make sure that you rent the intelligence. So the intelligence is something you can apply to your memory, your skills, your orchestration layer, infrastructure that lets the agent work for you. Pick the part of your life that you want to change and let an agent change it for you. I I'm not kidding. You can actually
[15:47] do it. If nothing else, if you know someone who has a really big pain point and maybe it's agent shaped or you know it's agent shaped and you're a builder, go tell them that. Go say, "Hey, this is not that hard. An agent can help you build this." We need positive examples. In fact, if you have positive examples, if you're an open rebuild, if you you built on something else, maybe you built on G brain, I don't care. Put positive examples of why an agent made a difference in your life down here. I would like to see them. I want more people to see positive examples of agents working.