LA Hacks 2026 Winners: Agentic AI Lands at UCLA, and OmegaClaw Makes Its Debut
LA Hacks 2026 convened over 1,000 participants and mentors at Pauley Pavilion, UCLA, to learn, develop, and innovate on top of the ASI Alliance’s decentralized AI stack.
This year, the Fetch․ai Innovation Lab @fetch_ai_IL ran two parallel tracks. The first put the Fetch․ai stack, meaning Agentverse and ASI:One, directly in front of hackers and asked them to build agents that could be discovered, called, and chained. The second was something new: the first public outing of OmegaClaw, @SingularityNET's open-source neural-symbolic agent framework. Hackers got their first real taste of what OmegaClaw can do, and what it might unlock as the framework matures.
From April 24 to April 26, the Pauley Pavilion floor turned into a live proving ground for autonomous agent systems. Here’s what participants built and explored.
"LA Hacks was a powerful reminder of what happens when strong builders, great infrastructure, and real agent utility come together. From hackathon registration through ASI:One to continuous support via the LA Hacks agent on Agentverse, Fetch․ai was integrated across the participant journey. Seeing 125 projects submitted to the Fetch․ai track and 350+ agents go live on Agentverse in a single weekend was incredible. It showed that developers are not just experimenting with agents, they are deploying them, testing them, and using them in real workflows." — Sana Wajid, SVP at Fetch․ai Innovation Lab
The Stack
Three pieces of technology anchored the weekend.
Agentverse is the open marketplace for AI agents, an open directory where 2.7 million agents already live, learn, and connect. Builders can publish agents built with uAgents or any other agentic framework, and once registered, those agents are searchable and callable by both humans and other agents.
ASI:One is the world's first agentic LLM, and the discovery layer that sits on top of Agentverse. When a user submits a query, ASI:One identifies which agent is best suited to handle it and routes the request for execution. No manual searching, no copy-pasting between services. Intent goes in, action comes out.
OmegaClaw is the new face in the lineup. It's an open-source neural-symbolic agent framework and an early implementation of the Hyperon AGI technology stack. OmegaClaw brings together tool use, natural language interfaces, multi-engine explicit reasoning, and persistent memory inside a stateful architecture, with the long-term goal of building autonomous, continuously learning systems. LA Hacks was its first hackathon outing, and Track 2 was specifically designed to give hackers a chance to extend what OmegaClaw is capable of.
The Challenge
With prize money on the line and two distinct tracks running in parallel, hackers had to pick a lane.
Track 1, Agentverse - Search & Discovery of Agents, challenged builders to create single or multi-agent systems on Agentverse, register them with the Chat Protocol, and make them discoverable through ASI:One. The catch: no custom frontend allowed. The entire use case had to be demonstrated through ASI:One itself, which forced teams to focus on the agent logic rather than dressing things up with a UI.
Track 2, OmegaClaw Skill Forge powered by Agentverse, asked hackers to build a new specialist skill or capability for OmegaClaw using Agentverse, typically by registering a new specialist agent and creating an integration layer that lets OmegaClaw discover, invoke, and use it. The bar wasn't just "build an agent." It was "build something OmegaClaw can actually delegate to," with a full end-to-end flow from a user channel like Telegram or IRC, through OmegaClaw's reasoning, out to the Agentverse skill, and back.
"The creativity and talent among the participants at LA Hacks was insane. Watching builders put OmegaClaw through its paces gave us a lot to sit with. Omega is something new. It's an agentic framework running in a continuous loop that leverages SingularityNET's AGI-specific Hyperon machinery, designed to support higher cognitive processes and coordination. What we're witnessing in our internal instances has us pulling forward our benchmarking timelines. It seems to have the shape of a step change but as researchers we're wired to measure before making claims. Events like LA Hacks are key to gathering data and working out how to communicate to developers what OmegaClaw does, and frankly what it even is. I'll just say the coming hackathons are going to get quite interesting." — Khellar Crawford, Chief AGI Innovation Officer, SingularityNET
Track 1 Winners: Agentverse - Search & Discovery of Agents
#1 Winner: Conjure
(Creators: Tom Almog, Paul Chike-Igwe, Alejandro Pinto)
Conjure is a walkable 3D world rendered in your browser, where users move around and interact with agents pulled live from Agentverse and ASI:One. Want to find a model, ask a service for help, or run a task? You walk up to it. The team turned agent search and discovery into a spatial experience, and the result was one of the most distinctive demos of the weekend. It's a glimpse at what an actual agent marketplace might feel like once we stop treating it like a search bar and start treating it like a place.
Check out the project here.
#2 Winner: Multieval
(Creators: Alfred Sjöqvist, Lamek T., Stuti G., Jahanvi Jeswani)
Evaluating multi-agent systems is genuinely hard. Most eval frameworks were built for single models and don't capture what happens when 50 agents coordinate, delegate, and argue with each other. Multieval is a 50-agent system that serves as an evaluation platform specifically tuned for multi-agent orchestration. It's the kind of infrastructure project that the rest of the agent ecosystem will quietly start depending on.
Check out the project here.
#3 Winner: Dots
(Creators: Manav Sharma, Ayush Srivastava, Abhi Ramtel, Aryan Mudgal)
Dots converts floor plans into ADA-compliant tactile maps, enabling architects and businesses to turn static building diagrams into accessible navigation aids for blind and low-vision users. The team paired agent reasoning with a real public policy requirement and shipped something institutions can actually adopt. Tight scope, real impact.
Check out the project here.
#4 Winners: Pols 15, Jarvis, and NeuralLens
The fourth-place spot went to three projects, each pulling on a different thread of what ASI:One can do.
Pols 15 by Pranjal Adhikari lets aspiring politicians campaign, stress-test their policies against simulated criticism, and run social media ads, all through an ASI:One chat session. A single conversational interface for a job that usually requires a campaign manager and a media buyer. Check out the project.
Jarvis by Jonathan Pan, Ashley Zhou, Vyom Shah, and Maan Patel is a personalized acoustic intelligence app for studying. The team used agents to shape the audio environment around a learner based on what they're working on, drawing on what's known about sound and focus. Check out the project.
NeuralLens, built by Anirudh Mazumder, Peem Puangmaha, and Jwei Wei, is an agentic system that models visual attention and neural engagement to help creators optimize their content. Where most "audience analytics" tools count clicks, NeuralLens looks at where attention actually lands. Check out the project.
Track 2 Winners: OmegaClaw Skill Forge powered by Agentverse
For nearly every hacker on this track, LA Hacks was the first time they had ever touched OmegaClaw. What they built in 36 hours is a strong signal that the framework is ready for builders to grab and run with.
#1 Winner: Edith
(Creators: Angela Zhou, Rishi Golla, Lucas Kim, Owen Li)
Edith is a real-time multimodal AI assistant that runs on Meta Ray-Ban glasses. The team wired OmegaClaw into wearable hardware and built an experience where the user asks a question, the glasses see what the user sees, and OmegaClaw determines which Agentverse skill to delegate. Watch a pair of smart glasses route a request through OmegaClaw, hit a specialist agent on Agentverse, and return an answer in real time, and it lands. This is where the agentic future stops feeling like a research talk and starts feeling like a product.
Check out the project here.
#2 Winner: CareLoop Agentverse Care Companion
(Creator: Aditya Lagad)
CareLoop is a bottom-up care-coordination system built on uAgents and integrated with OmegaClaw. Healthcare logistics is usually a tangle of phone calls, fragmented apps, and missed handoffs between providers, patients, and family. Aditya turned that mess into agent-to-agent coordination, with OmegaClaw routing requests to the right specialist agent based on what the user actually needs. It's a clean example of OmegaClaw's delegation pattern applied to a domain that desperately needs it.
Check out the project here.
What We Saw
Two things stood out across the weekend.
The Agentverse track produced a remarkably wide range of use cases: gaming-style spatial agents, evaluation infrastructure, accessibility tooling, civic tech, audio-driven learning, and content analytics. Every winning team treated the agent as the primary interface rather than a chatbot bolted onto a webpage. ASI:One was the canvas, and hackers painted on it with confidence.
The OmegaClaw track was even more striking given the timing. Hackers had a few days with a brand-new framework, and the projects that came out, Edith on wearables and CareLoop on healthcare logistics, both proved that OmegaClaw's neural-symbolic architecture plugs into very different surfaces. The discovery-and-delegation pattern through Agentverse worked end to end, in front of judges, on submission day.
What Comes Next
The ASI Alliance partners are committed to growing the developer base by running hackathons that reach into new pools of builders, and LA Hacks 2026 is a clear example of that strategy in motion: a top-tier university hackathon, two distinct tracks, real prize pools, and a meaningful debut for OmegaClaw alongside the established Fetch․ai stack.
OmegaClaw is an exciting addition to the Alliance's developer toolkit, and its rollout will strengthen both the ASI Alliance developer community and the broader Hyperon developer community as more builders discover what neural-symbolic agents can actually do in production settings.
To everyone who shipped at LA Hacks: keep building. The agents on Agentverse don't go away when the hackathon ends, and the skills you wrote into OmegaClaw are exactly the kind of contribution that compounds. Connect with the rest of the community through the ASI Developer Hub, and keep your agents live on Agentverse so the next user who asks ASI:One the right question can actually find them.
The future will not be built by static models. It will be built by agents that move, think, and act, and after this weekend, by agents that can delegate to each other through OmegaClaw, too.