Devpost’s Post

Took part in the aipromptchamp.com this week in Chicago. Couldn’t help but think back to my first hackathon at Liberty Mutual 12 years ago. Back then we had 🪚: • 2 days + 1 night and a team of 3 • Built a simple app and… it crashed during the demo 😅 • No pipelines; no cloud dev environments • A lot of effort → very small output This time 🤖 : • 4 hours only • Spun up GitHub Codespaces on a tablet • Used ChatGPT + Claude to ideate, stress-test, and build • Built and deployed via GitHub → Netlify ⏱️ But the real difference wasn’t just speed. The quantum of output in 4 hours (1 person) was significantly more than what we achieved in that hackathon with a team over ~36 hours. And now I spent less than 25% of the time in an IDE. ⬅️ That’s the shift. The developer role is changing. The responsibilities of product, architecture, and engineering are collapsing into one. Less code, more of: • frame the problem • think and design in user journeys • understand the data • ORCHESTRATE Clarity of thought matters more than coding, tech stacks and infrastructure. What I ended up building would have been the mvp of a small project for a few months. 🏥 A patient navigation system for a healthcare disruption scenario (hospital closure → thousands of displaced patients). A patient dashboard to find alternative care providers as well as a health care navigator dashboard with alerts, capacity views and 311 case triaging. Data from 13 hospitals, 10 clinics, 10 demographic zones, ER trends, displacement data, and live alerts —> all stitched into a working system. Are software teams embracing this faster than their business counterparts? 🤔 Not fast enough; there is an existential crisis. Too fast; there is a risk and controls nightmare. Either way, if you are not on claude code every weekend like it is a dota LAN party, you are at risk. Started but overwhelmed? Lets exchange notes. Thanks to Ramsey Jamoul, AI Prompt Championship for the event and great feedback from the panel judges James Evans, Justin Massa, Rohan Subramanian.

Join us this saturday for the first Macau edition of Cafe Cursor! Organized jointly with Macao Startup Club & Macau Data Lab/HK Data Lab We hear a lot about Foundational Models like Claude Opus/GPT 4.0 but have you wondered how to integrate them into your daily coding flows, without completely giving up the reins of control to a AI agent that could potentially drop your DB tables? Thats where Cursor comes in, an AI-powered code editor/IDE that helps you write, understand, and improve code using natural-language prompts, while you can see and review real time what the code and files the AI is writting. Sing up via the link on luma!

Yesterday wasn’t just a hackathon. It was a glimpse into the future of building. We participated in an internal hackathon powered by Fwdslash AI — and honestly, the experience was something else. What stood out the most? People who don’t usually code… were building real, working solutions. In just a few hours: - Ideas turned into prototypes - Problems turned into products - Curiosity turned into confidence I worked on building an AI agent focused on generating high-quality frontend LinkedIn content — combining: - Content intelligence - Image generation - Structured automation And the best part? You don’t need to be an “expert” anymore to build something impactful.   You just need the right tools and the willingness to explore. Huge shoutout to the team for organizing this and creating such an open environment to experiment and build. This is just the beginning 🚀 #AI #Hackathon #Frontend #Innovation #Fwdslash #BuildInPublic

Hallo Amsterdam! 👋 #Testing isn't just about scripts anymore; it’s about solving the "undebuggable." We’re gathering in-person on March 25th to tackle the industry’s toughest challenges and the future of smart machine testing. 🎙️ The Sessions: 1. Debugging the Undebuggable: A Deep Dive into Test Failures and Root Cause Analysis Speaker: David Burns, Head of Developer Advocacy, BrowserStack David will pull back the curtain on advanced root cause analysis, showing you how to find the "why" behind the most elusive bugs and flaky tests in your pipeline. 2. The Art of Breaking Smart Machines Speaker: Adonis Celestine, Senior Director, Global Automation, Applause As machines get smarter, our testing strategies must evolve. Adonis explores the nuances of testing #AI-driven and "smart" systems to ensure quality in an automated world. 📅 March 25 | 5:30 PM CET | 📍 In-Person Meetup The Agenda: - Root Cause Analysis: Moving beyond flaky tests to find the actual "why" behind failures. - Smart Testing: Mastering the technical nuances of breaking and testing AI-driven systems. - Community: In-person networking with the Amsterdam tech scene. 👉 Register here : https://lnkd.in/e_eTWXBv Huge shout out to our partners at spriteCloud for hosting this session and supporting the #community! 🙌

Built this prototype conversational maps during the Mappls Hackathon, 24 hours, first place🥇 (with Pranay Jalan & ANSH NOHRIA). Just talk to the map. Research places, plan routes, edit them, navigate, etc. all through conversation. Under the hood, we built an MCP server that exposes 8 Mappls API tools as callable functions for an AI agent, so the model can actually do things on the map, not just talk about them. We also threw together a ranking engine that weights different factors based on what you're asking for, so if you're looking for a quick lunch spot vs. planning a weekend trip, the suggestions feel more relevant. It gave noticeably better results than vanilla search, though honestly there's a lot of room to make it smarter given more time. That part alone could be a full project. Check out the quick demo video below. Github repo: https://lnkd.in/g_jRxBJT

A few weeks ago, we ran a company-wide hackathon where we gave everyone three hours and unlimited access to AI development tools to build whatever they wanted. 2 of 3 winners weren't engineers. I've been watching AI developments accelerate on a weekly basis. Everyone hears about what's happening, but it's harder to understand how it actually translates to your day-to-day work. We wanted our entire team to experience what's possible firsthand, instead of just reading about it in newsletters. One of our investors graciously provided Lovable credits to everyone at @Streamline AI to make this happen (huge shoutout to Sara Barek!). The instruction was simple: build what you think our application should look like. No constraints or limitations, but just three hours to create something. Non-technical team members went from zero to prototypes in ways that would have been impossible even six months ago. It was amazing to watch the amount of creativity coming to life. But we also learned something equally important. The demos were buggy. They looked great on the surface, but they weren't production ready. You can absolutely get someone from zero to a working prototype quickly now. But would you run your most confidential, critical legal data through it? Probably not yet. Especially considering we also support in-house teams at highly regulated companies, this was an important reminder. We also found clear deficiencies with vibe coding integrations and gaps in system architecture. Even in 2026, to build products people can trust, you need people who understand how to architect systems correctly, how to make data flow properly, how to build something that won't break when it matters most. The whole point of the exercise was to get everyone thinking bigger and differently. That part definitely worked. But understanding both the potential and the real limitations matters just as much.

This is why we build. 🗽 On Tuesday in NYC, we co-hosted a hackathon with Cartesia alongside some of the most passionate, creative developers we've had the pleasure of meeting. The range of ideas made one thing clear: there is no limit to what voice AI can do. Congrats to Koushik Vasa and his project, Sonus, on taking home first place on the night! Sonus is a clinical voice agent that interviews hospital patients before their doctor arrives and generates a structured SOAP note in real time. Our team really appreciated his attention to detail not just on the agent, but across the entire data lifecycle. Check out the live demo here: https://lnkd.in/gv9zyAqZ Some other incredible projects that we're still thinking about: 🛒 Cartsy. A voice agent that takes your grocery list, roasts your food choices, and tells you exactly where to shop. 🤖 An intelligent personal assistant that can help with any task, even some nefarious ones (we don't endorse, we just admire the ambition). 🚨 An AI-powered 911 voice agent built to make emergency response smarter and faster. So how were people able to develop a voice agent so quickly? Composer. Our new alpha feature lets you text-prompt your way into creating, debugging, and testing voice agents. Check it out in the Vapi dashboard. Thank you to every developer, vibe coder, and builder who came out and built with us in New York. We hope to see you at the next one.

I've noticed something talking to friends, family, and fellow business owners: The desire to automate is there. The curiosity about AI is real. But nobody knows where to start. They're spending hours on tasks that could run themselves. They know it. They just don't know how to get there — and most resources assume you're already a developer. And I decided to do something about it. I'm hosting my first ever in-person meetup in Kingston, NY — a hands-on working session where we actually build automations together, live in the room. No slides. No theory. Just you, your laptop, and one process from your business you want off your plate. I've spent years building AI systems and automation workflows in production. Now I want to bring that into a room with people who are ready to take their first step. April 16th. Kingston, NY. Only 6 spots. 🔗 Link in the comments.

What happens when a product manager 👩💻 goes to an AI Agent 🤖 hackathon in Silicon Valley? A lot of AI-assisted coding magic ✨ With the help of Sandeep (go team Northeastern University!) we used Codex to build an LLM-assisted lead extraction tool to help with sales prospecting at Plantime. Enter a city and vertical -> get contact info in <1min. As a bonus we added a multi-agent architecture with master router + specialist agents which could be used in the future to extend the tool: https://lnkd.in/eR2NtFHf The more I do this, the more I get excited to build more stuff... 🛠️