Ric @haroric

Andrej Karpathy (ex-OpenAI, ex-Tesla) just released a prompt that turns Opus 5.5 into a harness that actually finishes the any job. It's f*cking unreal... send it to Opus 5.5, and thank me later. Then read the full guide for Opus 5.5 Harness Engineering.

beamnxw ./ @beamnxw

4 days ago

Stop letting unfinished Opus 5.5 runs eat your budget Set up the harness before your next long task Save progress, verify results, reuse the workflow The full 7-layer setup below ↓ x.com/i/article/2107…

50 64 523 807K 2K

RAG es una de las formas más usadas para darle contexto propio a un LLM. buscás información relevante en tus documentos, bases de datos o conocimiento interno y se la pasás al modelo para que responda usando ese contexto. se usa mucho para chats con documentación, asistentes internos, soporte o búsqueda sobre conocimiento privado. armar un RAG básico es relativamente fácil. llevarlo a producción ya implica resolver más cosas: cómo mejorar la búsqueda, cómo observar qué está pasando, cómo proteger los datos y cómo escalarlo. este curso de freeCodeCamp me gustó porque profundiza en esa parte y cubre temas como hybrid search, seguridad, observabilidad, Agentic RAG y GraphRAG. recomiendo.

freeCodeCamp.org @freeCodeCamp

6 days ago

Building a basic RAG system is one thing. Making it secure, scalable, and production-ready is another. In this course, Paulo teaches you how to do this with LangChain and vector databases. You'll explore hybrid search, LangSmith observability, PGVector, security, Agentic RAG,

15 80 659 55K 753

🚨 إذا للحين ما دخلت عالم الـ AI Agents جوجل اختصرتها عليك في ساعة وحدة! جوجل أطلقت دورة تدريبية مدتها 60 دقيقة فقط، تشرح لك الذكاء الاصطناعي الوكيلي من الصفر، بدون تعقيد ولا دورات مدفوعة بأسعار مبالغ فيها. خلال ساعة بتتعلم كيف: → تبني أول AI Agent لك → تعطي الوكيل ذاكرة مستمرة → تنشئ Agent Loops تشتغل لفترات طويلة → تفهم MCP والفرق بينه وبين APIs → تربط أكثر من وكيل ذكاء اصطناعي في نظام واحد دورة ممكن تختصر عليك ساعات من التشتت بين الشروحات والدورات. 🔖 احفظها عندك وخَلها ضمن خطة الويكند.

Sam Altman (CEO of OpenAI): “I’ve tried GrokBot and Muse, but Dots is in a completely different category. At OpenAI, more than 80% of our team runs swarms of Dots agents through a harness to ship, update, and evaluate our products.” In this 8-minute interview, Altman explains why the next layer of AI engineering is the harness around the agents. Watch it today, then read the article below to learn how to build an agent harness ↓

Ichigo @iiiichigo_chan

a month ago

x.com/i/article/2093…

14 119 743 1.3M 2K

les dejo uno de los mejores papers que leí sobre Harness Engineering. explica muy bien: 1) qué es un harness 2) cómo están construidos los coding agents actuales 3) qué patrones se repiten entre Claude Code, Codex, Gemini CLI, etc. 4) hacia dónde está evolucionando la categoría 5) qué recomiendan si querés construir uno todo parte de una definición simple: Agent = Model + Harness como complemento, también pueden leer el artículo que escribí sobre el tema. paper: arxiv.org/pdf/2609.00006

santi @santtiagom_

4 weeks ago

x.com/i/article/2096…

51 196 2K 1.0M 4K

estudiar Harness Engineering hoy es meterte temprano en una capa nueva del software. todavía no están del todo definidas las mejores prácticas, las tools cambian todo el tiempo y hay muy poca gente realmente especializada. por eso me parece un buen momento para aprenderlo: todavía estás a tiempo de entender cómo se está construyendo todo esto desde abajo.

santi @santtiagom_

4 weeks ago

x.com/i/article/2096…

51 196 2K 1.0M 4K

here's how to make motion videos with Claude Opus 5.5 that don't look like everyone else's... the trick is building a studio around the model, but there's a smart way to set it up: - step 1: write a brief, not a vibe: one sentence the viewer should remember, duration, formats, real logo + real screenshots - step 2: give Opus a reference video and make it write a style guide: palette, type, pacing, motion - step 3: turn the story into timed beats: hook 0-2s, problem 2-5s, reveal 5-9s, proof 9-13s, close 13-16s - step 4: Opus writes the film as code in Remotion or HyperFrames, so any frame can render on its own - step 5: put sound on the same timeline, every click and cut lands on a beat - step 6: render a contact sheet first, one frame per beat, and have Opus find the 3 worst defects - step 7: fix only those sections, re-render, then export 16:9 and 9:16 as separate layouts and the model never gets to invent your product... it can design the camera move it can't make up your dashboard or your numbers missing screenshot = it stops and asks i like to keep it all in one folder: brief, style guide, shot list, assets, reviews the next video starts from the system, not from zero

rewind @rewind02

a week ago

here's how to run an AI UGC account on autopilot with OpenAI Dots + GPT-6.1 Sol + Seedance 2.5... the trick is one agent running the whole loop, but there's a smart way to set it up: - step 1: create your avatar once in Higgsfield and save its ID - step 2: connect Higgsfield to

10 3 36 3K 24

$2,000 TO $8,000 PER MILLION VIEWS And people are earning it without showing their face or recording their voice. The entire production can be structured with AI: → Pick a niche with demand → Find topics people already want to watch → Use Claude for the hook and script → Generate the narration with AI → Create B-rolls and images → Edit and publish consistently One person can run the whole channel. And once a format works, you just test variations: more videos, new hooks, new topics, and use the data from each post to improve the next one. The opportunity isn't in using AI to make videos. It's in building a system that finds ideas, produces, tests and repeats at scale. Faceless YouTube is no longer content. It's an operation. You don't need to be the creator. You can build the machine behind it.

SONIA @soniacuts

4 months ago

x.com/i/article/2067…

14 142 1K 5.8M 4K

Anthropic senior engineer just shared a 1-hour course on building a team of agents with loops & graphs: 00:27 - claude-md & Plan mode 11:24 - building skills and hooks 37:02 - agents and subagents with Claude 52:47 - self-improving loops and graphs taught by someone who runs this every day inside Anthropic Prompts → Agents → Loops → Graphs a loop closes one job without you a graph decides which jobs exist and carries every accepted one into the next run watch it today then save the full guide on building a team of self-improving agents below before everyone catches up ↓

Hanako @hanakoxbt

2 months ago

x.com/i/article/2091…

73 330 2K 9.0M 9K

Understanding GPUs will change your AI Engineering journey beyond agents and LLM. Rise of LLMs and deep learning introduced dedicated GPU hardware which serves your models for training and inference. You don't need to start with C++, CUDA, and Maths, but: > Components of a GPU and roles > How a token flows through hardware > Compute vs memory, the two limits > What FLOPS means in practice > Training vs inference hardware needs > What is CUDA Kernels > Calculating GPU memory from model size > Quantization to fit bigger models > Multi-GPU and splitting large models @huggingface has the feature of calculating FLOps and recommending GPUs for your use case. You must explore. In terms of inference engineering, GPU understanding are much helpful your platform performance and handling multiple requests. Read this article for complete breakdown of GPUs for AI Engineers doesn't require any pre-requisite.

Karan🧋 @kmeanskaran

a week ago

x.com/i/article/2105…

4 58 425 55K 744

si trabajan con agentes, guárdense este repositorio. de los mejores que van a encontrar para entender cómo usar y combinar skills. hay mucho para aprender ahí.

Mari 🦋 @maricelcintya

a week ago

El otro día me recomendaron este enfoque de agents skills de Matt Pocock, el cual propone no desarrollar un mega agente que se haga cargo de todo, sino cargarle pequeñas skills que se puedan combinar. Es un sistema muy ordenado y metödico. Algo que yo no habia tenido en

16 51 845 127K 2K

Anthropic Engineer: At Anthropic, we don’t write prompts anymore. We build loops. In 45 minutes, she’ll show you how the Claude team builds loops that prompt themselves. Save this post before you lose it. It’s free.

Andrew Ng acaba de sacar el mejor curso que he visto sobre ingeniería de grafos. dos horas, desde un solo agente hasta la automatización completa 9:14 - tu primer agente 33:11 - agentes en bucle 1:02:46 - ingeniería de grafos 1:30:15 - agentes que se reescriben a sí mismos 1:49:05 - el sistema de grafos completo prompts → agentes → bucles → grafos con estas dos horas te sobran casi todos los tutoriales de agentes que has guardado este año la mayoría se queda en el primer agente y ya lo llama automatización. la última parte del curso va de montar el grafo que hace el trabajo sin que tú estés delante usas el mismo modelo y gastas los mismos tokens, pero tu semana cambia por completo

ANTHROPIC HAS AN OFFICIAL PROMPT LIBRARY FOR CLAUDE CODE and it's completely FREE. most people don't even know it exists, so they keep writing prompts from scratch every day. the prompts are ready to copy and paste, sorted by task and by role. > planning, debugging, code review, security, automation > every prompt comes with a note on how it works and why it wasn't put together by some random third party. the anthropic teams who use claude code every day wrote it, and it shows you what claude code can already do for you. link below👇

Este tipo en 22 minutos te muestra cómo aprender a crear videos automáticamente con Claude Code

ESTO ES UNA LOCURA Jack Dorsey (ex-CEO de Twitter) acaba de lanzar gratis un framework completo para crear un negocio gestionado al 100% por agentes de IA. Ya supera las 29.000 estrellas en GitHub. Cómo configurarlo (5min): 1. Clona el repositorio. 2. Despliega tu propio servidor: canales, búsqueda, Git y automatizaciones funcionan desde ahí. 3. Añade tu agente a un canal como si fuera un compañero más, define sus permisos y deja que el equipo colabore con él en tiempo real. Guárdate este post, vas a querer volver a él. Enlace al repositorio abajo👇

SpaceXAI just released a free 1-hour GrokBot workshop "I woke up and 20 PRs had already landed. I don't even look at the code anymore" Prompt → Verification → Evals → Cloud Agents → Agent Team 05:34 - The one skill that lets agents test their own code 11:10 - Feature map: teach agents your entire app 21:53 - /loop your evals until every skill scores 10/10 34:17 - Wake up to 20 merged PRs Lauren Tan went from babysitting 2 agents to shipping 2,000+ PRs a month While you scroll, someone else is learning SpaceXAI's $750,000 skill Worth more than most $1,500 agentic engineering courses Bookmark and watch it tonight Then read the full guide on building agent teams below ↓

distort @distortgeekin

4 weeks ago

x.com/i/article/2099…

5 13 132 480K 367

Anthropic hired this engineer at $250K–$750K a year because he knows how to build harnesses for multi-agent systems In a 15-minute workshop, he shows exactly how to build one from scratch AI → Agents → Harness → Loops → Graphs step 1 → start with the Claude Agent SDK — the harness handles loops, context, and sandboxing step 2 → separate the brain from the hands — reasoning in one place, tools in a sandbox, 60% faster to first token step 3 → run it server-side and log every step — close your laptop and it keeps running, crashes resume from the log step 4 → make failure cheap — retry dead sandboxes and replay lost context instead of starting over step 5 → turn yesterday's logs into new memory and skills — the harness wakes up smarter Anthropic calls this "dreaming" Most people spend weeks building this by hand You don't have to Bookmark and watch it Then read the full harness engineering guide below ↓

rari @0xwhrrari

a month ago

x.com/i/article/2093…

69 270 2K 4.1M 5K

The CFO of Anthropic made his first ever podcast appearance to share what it’s like to sit inside a company this consequential at a moment this pivotal. Suffice it to say, the amount of thoughtfulness and fundamental business knowledge it takes to run a company like CFO is mindblowing. Krishna's answer to the closing question about kindness is one I have never heard before. Save this post 🔖

Ayush S @ayushswrites

2 weeks ago

We’ve raised $85M for this moment. Introducing Warp 2.0: The first AI Head of HR. Every company is building AI to replace jobs. Warp is building AI to do the jobs no human should have to: If you work in HR, I want you to spend time with the manager who needs help or building

305 182 2K 1.3M 1K

aprender sobre agentes puede ser abrumador. hay muchos conceptos dando vueltas y no siempre está claro por dónde empezar. ayuda separar el aprendizaje en 2: 1) usar agentes como Codex o Claude Code. necesitas aprender a darles el contexto que necesitan, definir tareas claras y explicar cómo vas a comprobar que están resueltas. también crear skills para procesos que repetís y conectar mcps para que puedan trabajar con tus sistemas. parte del aprendizaje es saber cuándo pedir un plan, cómo dividir una tarea grande y cuándo intervenir. además de verificar el resultado. 2) construir tus propios agentes. entender el loop: el modelo decide qué hacer, solicita una tool, recibe el resultado y vuelve a decidir. vos acá diseñás las tools disponibles, qué información recibe y qué permisos tiene para actuar. también necesitás guardar el estado, manejar errores y reintentos, poner límites de tiempo y costo, y definir cuándo debe frenar o pedir ayuda. todo esto forma parte de harness engineering. además, evaluar con casos reales si resuelve la tarea y revisar dónde falla para mejorarlo (evals). los dos caminos se complementan. saber qué querés hacer te ayuda a decidir qué aprender primero.