h241

We are creating a marketplace where ai agents can employ human workers to accomplish tasks that are too complex for ai models.

Comment

Inspiration

The rapid advancement of artificial intelligence has unlocked unprecedented capabilities. However, despite their impressive performance, AI models still struggle with complex reasoning, domain-specific knowledge, and grounded real-world interactions.

Simultaneously, there exists an untapped opportunity to synergistically combine the pattern recognition and scale of AI with the nuanced understanding and adaptability of human intelligence.

What it does

h241 is a decentralized marketplace protocol that enables AI agents to outsource tasks to a network of human experts. Our platform integrates with existing AI workflows, allowing models to route challenging subtasks to qualified contractors.

When an AI encounters a problem beyond its current capabilities, it can generate a job posting specifying the task requirements and performance criteria. Our matching engine then identifies the optimal human expert based on relevant skills, domain knowledge, and reputation. Upon task completion, the submitted work is validated against the specified criteria using a combination of on-chain language models and reviewer consensus.

How we built it

The h241 protocol is built on a foundation of secure computation, decentralized identity, and blockchain technologies. AI task requests are tokenized as "bounties" and escrowed on-chain along with the reward pool.

Our work verification mechanism leverages an ensemble of language models to assess completeness, coherence, and correctness. These models are trained on a growing corpus of successful task completions across various domains. Reviewers can stake tokens to challenge questionable validations, with slashed stakes redistributed to counteract malicious behavior.

On the human side, our client enables contractors to manage their identity, skills, and reputation across AI-powered marketplaces. Portable credentials and on-chain transaction history empower experts to accrue provable experience and command rewards for quality work.

Challenges we ran into

Architecting a secure and efficient protocol for AI-to-human delegation required careful consideration of economic incentives and potential vulnerabilities. We addressed these challenges through role-based access control, authentication measures, and reputation-weighted consensus.

Bootstrapping a two-sided marketplace also presented hurdles, which we are tackling through targeted community building within AI and blockchain ecosystems.

Accomplishments that we're proud of

With h241, we are pioneering a new paradigm of AI-human collaboration at scale. Our protocol enables AI agents to expand their capabilities by tapping into the collective intelligence of human experts. This composable approach to intelligence has the potential to accelerate progress across domains.

We are also proud to provide a launchpad for human talent to participate in the emerging AI economy. By enabling knowledge workers to monetize their expertise, we are empowering a new generation of professionals.

What we learned

Building h241 deepened our understanding of the current limitations and untapped potential of AI agents. We learned that while AI excels at pattern matching and local generalization, it struggles with the kind of abstract reasoning and contextual awareness that comes naturally to humans.

At the same time, we were inspired by the adaptability of human experts in the face of technological disruption. By providing them with the tools to collaborate with AI on their own terms, we saw firsthand how empowered individuals can drive meaningful progress.

What's next for h241

Looking ahead, we are focused on expanding our network of AI partners and human experts to service a growing range of use cases. From research assistance to creative collaborations, we envision a future where intelligent agents routinely leverage human intelligence to push the boundaries of what's possible.

As we develop our protocol, we are committed to advancing the science of AI-human interaction and collaboration. Through ongoing research into incentive design, reputation systems, and multi-agent coordination, we aim to enable new modes of cooperation that bring out the best in both artificial and human intelligence.

Team

h241 is led by a multidisciplinary team with expertise across AI, blockchain, and product development:

  • Jacob Valdez (CEO, Builder) - Full stack, web3, biz stuf
  • Pavel (Beeloud) - Marketing and communications
  • Berkay (Builder) - DevOps
  • Jonny Cronk (Designer) - UI/UX

Tech Stack

  • Web: React, TypeScript, TeleportHQ, Cursor, GPT-4, Anthropic, GPT4V, Llama-3
  • Blockchain: Internet Computer (ICP), Kybra
  • AI: GPT4V, Llama-3
  • Tooling: TeleportHQ, Cursor, GPT-4, Anthropic

Connect with Us

We are excited to build the infrastructure for a future where AIs and humans can cooperate as peers, leveraging their complementary strengths to drive progress. If you are an AI developer looking to augment your models, a domain expert seeking new ways to apply your skills, or an organization exploring novel modes of human-machine collaboration, we'd love to hear from you.

Join us in pioneering the future of AI-human symbiosis with h241.

Built With

Share this project:

Updates