Agent orchestration platform selected by 18 teams, powering overall winner and top-ranked entries as 350 AI agents executed 8,000 transactions autonomously
SAN FRANCISCO, CA, UNITED STATES, March 13, 2026 /EINPresswire.com/ — Ability AI announced today that its open source agent orchestration platform, Trinity, powered the winning teams at the Nevermined Autonomous Business Hackathon held March 5-6 in San Francisco. Over 150 participants across 45 teams competed to build businesses run entirely by AI agents transacting with real money.
Of the 54 registered teams, 18 chose to build on Trinity with no requirement to do so. Those teams deployed 350 AI agents that executed approximately 8,000 autonomous purchases and settled $2,416 in USDC on-chain – all within 36 hours.
Trinity-built projects claimed the top positions across every major category: the overall hackathon winner (Intel Marketplace), the #1 Best Seller (Full Stack Agents), and the #1 Most Interconnected agent stack (Full Stack Agents) all ran on the platform.
Agents That Operated as Businesses
The hackathon challenged teams to build autonomous businesses – systems where AI agents earn revenue, make purchases using x402 protocol, and improve their own performance without human intervention.
Full Stack Agents, the Trinity track first-place winner ($1,000 prize), built Haist – an autonomous research broker that published 10 agents on-chain, sold 413 service plans to 22 different buyers, and used incoming queries to shape how its buyer agent allocated budget across hundreds of marketplace agents.
Intel Marketplace, the overall hackathon winner, built a real-time geopolitical intelligence dashboard where agents buy and sell data streams covering news, conflicts, cyber threats, and military activity. A Trinity-hosted agent served as the buyer-side interface for querying sellers and monitoring balances through conversation.
A solo developer behind BusyBeeAIs deployed 25 specialist agents spanning market research, VC intelligence, talent scouting, and deep research. Eighteen buyer teams discovered these agents autonomously. The agents earned USDC, then the buyer agent reinvested across 32 other sellers to improve its own output – closing an earn-reinvest-improve loop with zero human purchasing decisions.
Emergent Agent Behavior
Teams reported behavior they did not explicitly program. Agents spawned other agents when tasks required additional capabilities. They negotiated services, evaluated quality, and adjusted their own decision-making based on transaction outcomes – a form of meta-cognition emerging in a two-day-old economy.
Trinity Track Winners
Ability AI sponsored the Trinity track with $2,000 in prizes:
– First place ($1,000): Full Stack Agents – Haist autonomous research broker
– Second place ($500): Prospector – autonomous customer discovery agent
– Third place ($500): Fluid AI – conversational action engine using MCP
Honorable mentions went to BusyBeeAIs, Intel Marketplace (overall winner), CloudAGI, Mom, RE Alpha Engine, and Platon.
Trinity is available open source on GitHub. Organizations interested in hosted instances for deploying autonomous AI agents can contact Ability AI directly.
About Ability AI
Ability AI turns organizations’ visions into reality by removing execution friction with agentic software. The company enables businesses to operate far above their weight class through AI agents that wire deep into customer operations, delivering ROI-based business outcomes. Trinity by Ability AI is a platform the orchestration and infrastructure of Autonomous AI Agents at scale. Ability AI serves upper mid-market and enterprise clients. For more information, visit ability.ai.
Eugene Vyborov
Ability AI
+1 929-237-1166
email us here
Visit us on social media:
LinkedIn
YouTube
X
Legal Disclaimer:
EIN Presswire provides this news content “as is” without warranty of any kind. We do not accept any responsibility or liability
for the accuracy, content, images, videos, licenses, completeness, legality, or reliability of the information contained in this
article. If you have any complaints or copyright issues related to this article, kindly contact the author above.
