Sean P. Larkin’s Post

Vibe coding changed how we build software. Today we're open sourcing the next leap. Vibe Working. You describe what your business needs. AI proactively assists you on achieving your goals. You stay in the flow of work that actually matters. ohwow is a human+AI operating system that runs on your laptop. You tell it about your business and it assembles a team of automations + agents around it. A system that observes before acting. That earn trust slowly and lose it fast. That learn your work rhythm, energy, and boundaries. Under the hood: a seven-layer cognitive architecture. Perception, reasoning, memory, embodiment, purpose, identity, collaboration. Your agents argue with their own plans before executing. They predict failures from experience. They know when you're overloaded and hold decisions until you're sharp. They have bodies. Screen control, desktop automation, WhatsApp from your real number, real browser. The architecture supports physical hardware too: Arduino, ESP32, sub-10ms reflexes. Your devices become one organism with shared knowledge. Local. Free. MIT licensed. Your data never leaves your machine. npm install ohwow -g If you're a researcher: the architecture is real and open. If you're a founder: this is the team you need but can't hire yet. If you're a builder: 90+ tools, A2A protocol, MCP, full extensibility. Your computer should work for you. Not the other way around. Link in comments. #vibeworking #ai #agents #localfirst #opensource

Vibe coding proved something massive: you don't need to control every detail to get great work done. You set the direction, AI handles execution. Now apply that to your entire business. Not just code. Sales. Support. Research. Content. Operations. That's vibe working. And it changes the equation of what one person can build. This idea pulled me into a rabbit hole of cognitive science, trust dynamics, embodiment, and what collaboration actually means between humans and machines. OHWOW is the result. It's far from finished. I built it alone, which means there are rough edges everywhere, blind spots I can't see, and assumptions that need to be challenged by people smarter than me in a dozen different fields. If any of this resonates with you, I'd genuinely love your eyes on it. Install it and tell me what breaks. Open an issue with what frustrates you. Challenge the architecture. Or just tell me I'm wrong about something. That's usually where the best ideas come from. The Discord is open. The codebase is open. The conversation is open.

There's something really wrong with the discussion around AI: there's too much solution theatre and not enough actual thinking. It's almost as if people have outsourced their critical faculties... (no... that couldn't possibly be true). 🤔 On the one hand, I genuinely think that any software developer who isn't actively reappraising how they work right now is not paying attention: and this jangliness will only intensify over the next few years. Yet on the other hand, anyone who expects everyone and their dog to jump to full-fat agentic development in the next heartbeat is kidding themselves bigly. If nothing else, going from close-to-zero attack surface to 100+ brand new (and constantly updated) npm packages feels insane. And it's the mile-wide gap in the middle that everyone like me is actually interested in. How should I be changing my practices and the way I think about what makes good software? It's the real question du jour. But all we get fed is yet more agentic theatre. I don't want to be a cool kid (I'm a tad too old for that), but there's plenty of good stuff coming that I'm looking out for.

Everyone's using Claude. Most are only using 10% of what it can actually do. Here's the Claude starter pack you need to know: 1. Claude Chat (claude.ai) Your thinking partner. Write, analyze, research, brainstorm. Works on web, mobile, desktop. This is where most people stop. Don't. 2. Claude Code An agentic coding assistant in your terminal. It doesn't just write code. It plans, executes, and iterates. Builds full apps. Writes scripts. Automates pipelines. For developers who want to ship 10x faster. 3. Claude Cowork No code? No problem. Cowork connects Claude to your files, calendar, email, and apps. Automate your desktop workflows without writing a single line. Built for operators, not just developers. Most people use Chat. Smart people add Code. Rare people run all three. Which one are you using right now? #Claude #AI #ClaudeCode #Cowork #AIAgents #Automation

🚀 Claude Code in Action - The Future of Software Development is Here Just explored Claude Code in Action, and it’s honestly redefining how we think about coding. 💡 Imagine this: You describe a feature in plain English… and an AI agent builds, debugs, and ships it for you. That’s exactly what Claude Code enables. It’s not just a coding assistant-it’s an agentic development system that can: ✔️ Understand your entire codebase ✔️ Write and refactor code across multiple files ✔️ Debug issues and run tests ✔️ Automate repetitive dev tasks What stood out to me: 🔹 Works directly in your terminal/IDE - no context switching 🔹 Handles end-to-end workflows (idea → PR) 🔹 Shifts developers from “coding” to orchestrating intelligent agents 📈 This is a big shift: Developers are no longer just writing code-they’re designing systems and guiding AI agents to build it. () 🔥 Key takeaway: The future belongs to engineers who can combine AI + system design + problem-solving, not just syntax. Excited to explore more use cases and integrate this into real-world workflows! #AI #GenerativeAI #ClaudeCode #SoftwareEngineering #LLM #Developers #FutureOfWork

Claude Code just redesigned their desktop app around one concept: parallel agents. This signals a fundamental shift in how AI coding tools should work. Most engineering teams still approach AI assistance as single-agent interactions. One AI helps one developer with one task at a time. But software development is inherently parallel. Multiple developers, multiple components, multiple systems - all evolving simultaneously. The real question: should AI agents work in isolation or coordinate their efforts across the entire codebase? For teams scaling AI workflows, what's the bigger unlock? - Smarter individual agents   - Orchestrated parallel agents The redesign makes Claude's bet clear. Which direction is your team heading? #AIEngineering #DigitalTransformation #CodeGeneration #TechStrategy #AILeadership

🚀 𝐋𝐞𝐭 𝐦𝐞 𝐬𝐡𝐨𝐰 𝐲𝐨𝐮 𝐡𝐨𝐰 𝐭𝐨 𝐩𝐨𝐰𝐞𝐫 𝐲𝐨𝐮𝐫 𝐛𝐮𝐬𝐢𝐧𝐞𝐬𝐬 𝐰𝐢𝐭𝐡 𝐈𝐧𝐜𝐨𝐝𝐞 𝐆𝐫𝐨𝐮𝐩 𝐒𝐨𝐥𝐮𝐭𝐢𝐨𝐧𝐬 Most companies don't need to build everything from scratch. They need the right building blocks. At Incode Group, we've already built quite a few of them - a Telegram AI for lead monitoring, a healthcare assistant, a real estate staging tool, a finance tracker, an interview platform for tech hiring, and more. Different industries. Different problems. Different shapes. The through-line is always the same: someone had a workflow eating their time or money, and we built our way out of it - usually with AI, sometimes with solid infrastructure, always with the goal of making something people actually want to use. Because this is where I see the market going: 👉 Less "let's build everything" 👉 More "let's assemble smarter systems" The real advantage isn't writing more code. It's how fast you can test, adapt, and scale. If you're developing something and wondering whether the hard part has already been solved, it might be worth a look. 👇 incode-group.com/solutions Do you prefer building from scratch or leveraging existing solutions where possible?

Claude Sonnet 4.6 isn’t just a chatbot. It’s a full operating system for thinking, building, and shipping. 15 things almost no one uses: 1. Two models exist Sonnet 4.6 + Opus 4.6 (heavy reasoning) 2. Memory via artifacts Your own mini database inside chat 3. Claude can call itself Yes you can build AI tools with embedded AI 4. Lives inside your tools Chrome, Excel, PowerPoint (beta) 5. Deep Research mode 20+ searches before giving answers 6. Knows today’s date And searches the web when needed 7. Connects to your apps Gmail, Drive, Calendar via MCP 8. Claude Code Not chat, a full terminal agent 9. Persistent personality Set once → applies everywhere 10. Builds live apps React, charts, tools… inside chat 11. Cowork Automate your desktop without coding 12. Generates real files .docx, .pptx, .xlsx, PDFs - ready to download 13. No ads. Ever. Your workspace stays clean 14. Writing style control Concise, detailed, whatever you prefer 15. Learns from past chats Builds context over time (if enabled) Most people use Claude like Google. That’s the mistake. The real power = treating it like a system, not a tool. Which one of these did you NOT know? Questions about O-1, EB-1A, or EB-5? Book a free consult - https://lnkd.in/gqJUQ-8X Join our Open Atlas community for visa-friendly job drops and free resume reviews - https://lnkd.in/gqVU84qW 🔔 Follow to stay updated on high-skilled immigration, jobs, and tech #AI #Claude #Productivity #Startups #Tech #BuildInPublic #Automation

I am not a coder. I’m just a guy who knows how to orchestrate machines. This is a glimpse into "Vibe coding" (or Context Engineering, as I prefer to refer to it). The point is to show you that anyone can do it. I repeat, I am *not* a coder. I'm no smarter than the average bear. But in this this video, I’m building a real-time, multiplayer iOS game. I plan with AI (using the superpowers skill I shared in yesterday's post) and then AI codes and we iterate on bugs/issues. Really, I'm guiding (with context; I know what I want) and then I'm a monkey following instructions for building/testing (and eventually publishing). Imagine *EVERYONE* in your company with these abilities! What would or could this do for your business productivity? What if your marketing admin could suddenly build systems that automate some of their work processes? What if your operations manager could build custom tools to automate their entire workflow without waiting on IT? Giving your employees access to AI isn’t 'enabling' them—it’s just giving them a library card. If they don't know what questions to ask or how to find the right book, they’re still just standing in the lobby. I teach them how to become the Librarian. What would it do for your bottom line if you could 10x the output of the team you already have, simply by teaching them how to "vibe" with the AI? The talent is already in your building. Unleash me in your organization. Unleash your organization's talent. Unleash your business outcomes.

The age of agentic engineering shifts the bottlenecks in software engineering. What we see clearly is that the new bottlenecks become to be about acceptance testing and gathering feedback. There are two sides to managing this new bottleneck: 1. We need to collect feedback more efficiently and in higher volumes than in the past. User feedback buttons and meeting notes need to be completely utilized to maximize the volume of high quality feedback. 2. Utilizing the feedback maximally. If you have a system with let's say a native app and a web application and you're trying to keep them in parity and in alignment, if you report a bug in one, Claude Code will happily fix it in one side, and not in the other. Also, it will not proactively look for other places with the same kind of a bug, or examples of similarly ill-designed items. You need to tell it to do these things. This is about maximal utilization of the feedback signal, to let it affect at larger blast radiuses. #AI #AgenticEngineering #ClaudeCode