May 27, 2026
What to Build for the HackerNoon x Nosana Decentralized AI Hackathon
The next wave of AI products will not be simple chatbots sitting in a browser tab. They will be agents, workflows, tools, and systems that can reason, act, generate, classify, automate, monitor, and improve over time.
That is where decentralized GPU compute becomes interesting.
Many AI ideas need more than an API call. They need inference, orchestration, media generation, fine-tuning, data pipelines, real-time processing, or repeated autonomous execution. These are exactly the kinds of workloads builders can start exploring with Nosana.
For the HackerNoon x Nosana Decentralized AI Hackathon, the strongest projects will likely be the ones that show what happens when AI is not just a feature, but the core engine of a real product.
Here are some ideas worth building.
1. Agentic Systems
Agentic systems are one of the most exciting and evolving directions for AI builders, particularly within the OpenClaw ecosystem. The next generation of agents must move beyond simple Q&A. We are looking for systems that are end-to-end integrated with true agentic capabilities—systems that can autonomously plan, execute tasks, monitor results, and continuously improve the workflow as they go. This presents a major opportunity for builders to innovate by exploring new UIs for deploying and interacting with complex agent flows. These fully integrated systems can fundamentally change how users interact with technology, ultimately helping users control their digital lives. Examples of these systems could include:
- An AI research agent that gathers information, summarizes it, and turns it into a structured report
- A productivity agent that plans a user’s day based on deadlines, priorities, and calendar events
- A support agent that triages tickets, suggests responses, and escalates complex cases
- Agents that allow you to update and deploy a live website by talking to it via WhatsApp or Discord, or your
- A workflow agent that watches for a trigger and automatically runs the next steps
The key is to move beyond “AI gives an answer” and toward “AI completes a process.”
The key is to move beyond “AI gives an answer” and toward “AI completes a process.” A strong hackathon project could show an agent taking a goal, breaking it into steps, using tools, and producing a useful final outcome.
2. Agentic Orchestration Systems
Many AI workflows require more than one model or one agent.
One agent might search for context. Another might classify information. Another might generate content. Another might validate the output. The opportunity is to build orchestration systems that make these components work together.
This could be especially useful for:
- Multi-agent research workflows
- Automated content production pipelines
- AI-powered business operations
- Developer automation
- Data analysis workflows
- Web3 automation
A great project here would not just show one impressive AI output. It would show a system where multiple AI steps are coordinated reliably. The automatic composition of agents will set your project apart. How will your combine different agents to solve dynamic problems?
Think: “AI workflow engine,” not just “AI assistant.”
3. Generative Media Workflow Creators
Generative media is becoming one of the most powerful use cases for GPU compute.
A builder could create a workflow that turns a simple idea into a complete media asset. For example:
- Text prompt → image concepts → selected style → final visuals
- Blog post → social media graphics → short video script → video assets
- Product description → ad variations → visual creatives
- Podcast transcript → clips → captions → thumbnails
The strongest projects will make generative media feel practical, not just experimental.
Instead of only generating a single image or video, think about building a workflow creator that helps users move from idea to finished asset. How will your product leverage AI to automate and simplify the expression of ideas and creativity?
4. Video Generation Workflows
Video generation is one of the most compute-intensive and exciting AI categories.
A hackathon project could explore workflows such as:
- Text-to-video generation
- Image-to-video generation
- Short ad generation
- AI-generated product demos
- Social media video creation
- Educational video generation
- Automated video storyboarding
A simple but useful idea would be a tool where a user enters a topic, chooses a style, and gets a short video concept with scenes, voiceover, captions, and generated assets.
The opportunity is not only in generating video, but in making the full process easier.
Most people do not just need “a video.” They need an idea, a script, scenes, visuals, captions, and export-ready content.
5. Video Classification Workflows
Not every AI video project needs to generate video. Some of the most useful products will analyze video.
Builders could create systems that classify, tag, summarize, or detect patterns in video content.
Possible projects:
- A tool that analyzes long videos and finds the best clips
- A system that detects unsafe, low-quality, or off-brand content
- A sports video analyzer that identifies key moments
- A meeting video summarizer
- A security or monitoring workflow that flags unusual activity
- A creator tool that tags scenes by topic, emotion, speaker, or visual style
This category is powerful because it can serve creators, companies, educators, and developers working with large amounts of video.
6. Screen Recording Agents
Screen recording agents are a practical and underrated category.
Imagine an agent that can watch a screen recording and understand what happened.
It could:
- Turn a product demo into documentation
- Convert a bug recording into a developer ticket
- Summarize a user testing session
- Create a tutorial from a screen recording
- Detect where a user got stuck in a workflow
- Generate onboarding materials from product walkthroughs
- Help users keep track of tasks they did during the day
This is especially interesting for SaaS teams, developers, support teams, and product managers.
A strong project could let someone upload a screen recording and receive a structured output: summary, steps, issues, screenshots, and suggested next actions.
7. Render 3D Environments for Games
AI and GPU compute can also support gaming workflows.
Builders could create tools that help game developers generate, render, or iterate on 3D environments.
Ideas include:
- AI-generated game environments from text prompts
- Procedural world-building tools
- 3D asset preview workflows
- NPC environment generation
- Game level concepting tools
- AI-assisted rendering pipelines
For indie game developers, this could reduce the time between idea and prototype.
A great hackathon project could show how AI helps a builder create a playable or visual game environment faster than a traditional workflow.
8. Fine-Tuning AI Workflows
Fine-tuning remains one of the most useful ways to make AI more specific.
Builders could create workflows that make fine-tuning easier for non-experts.
For example:
- Upload a dataset → clean it → format it → fine-tune a model
- Fine-tune an assistant for a specific company knowledge base
- Create a domain-specific writing assistant
- Build a customer support model trained on real support examples
- Fine-tune a model for legal, medical, educational, or developer documentation use cases
The strongest projects here will not just fine-tune a model. They will make the workflow simple, transparent, and useful.
A project that helps users prepare data correctly could be just as valuable as the model itself.
9. Self-Learning Agents
Self-learning agents are at the forefront of the OpenClaw vision, designed for truly continuous improvement. We envision systems that go beyond simple feedback loops by collecting all memories and interactions with the environment and the user daily. At the end of each day, the agent would retrain its own model, incorporating all of its collected memories. This mechanism ensures the agent is continually improving, simultaneously evolving its own personality to better align with its function and to deepen its understanding of the user. This advanced capability is foundational to end-to-end integrated agentic capabilities, enabling the exploration of new UIs and allowing these systems to fundamentally help users control their digital lives.
This could include:
- An agent that learns a user’s preferences over time
- A trading or research agent that reviews its own past decisions
- A coding agent that remembers project-specific patterns
- A personal productivity agent that adapts to how a user works
- A support agent that improves from resolved tickets
The important part is not claiming that the agent is fully autonomous or magically intelligent. The goal is to show a feedback loop.
A good project could demonstrate how an agent stores past results, evaluates what worked, and changes its next action based on that history.
10. RAG Data Pipelines
Retrieval-augmented generation is still one of the most practical AI patterns.
But many RAG projects fail because the pipeline around the model is weak. The data is messy. The documents are not chunked well. Retrieval is poor. The answers are not grounded.
Builders can create better RAG workflows for:
- Company documentation
- Research papers
- Legal documents
- Medical knowledge bases
- Developer docs
- Customer support centers
- Internal team knowledge
A strong hackathon project could focus on the full pipeline:
Document upload, parsing, chunking, embedding, retrieval, answer generation, source citation, and feedback.
That kind of project is immediately useful.
11. Voice and Audio Agents
Voice is becoming one of the most natural interfaces for AI.
Builders could explore:
- Voice assistants for specific professions
- Meeting transcription and summarization agents
- Audio note-to-task workflows
- Voice-based customer support
- AI podcast editors
- Speech-to-speech agents
- Language learning agents
- Audio classification tools
A practical example: upload a long audio file and get a transcript, summary, action items, speaker highlights, and social media clips.
Voice and audio projects are especially strong because they connect AI to real human behavior. People already record meetings, calls, notes, interviews, lectures, and podcasts. AI can help turn that raw audio into something useful.
12. Vision, OCR, and Document Workflows
There is huge potential in AI systems that understand images, PDFs, screenshots, scans, and documents.
Builders could create:
- Invoice processing tools
- Receipt extraction workflows
- PDF summarizers
- Contract review assistants
- ID or form processing tools
- Screenshot-to-task agents
- Medical document organizers
- Research paper analysis tools
The best projects will go beyond simple OCR.
A useful workflow could extract text, understand structure, classify the document, summarize the content, and send the output to another tool or database.
This is a great category for building something practical and easy to demo.
13. Web3 AI Orchestration Systems
Web3 and AI are starting to overlap in more meaningful ways.
Builders can explore AI systems that interact with on-chain data, wallets, smart contracts, DAOs, or decentralized applications.
Possible ideas:
- AI agents that monitor on-chain activity
- DAO proposal summarization agents
- Smart contract analysis workflows
- DeFi research agents
- Wallet activity classification tools
- On-chain alerting systems
- Autonomous agents that trigger actions based on blockchain data
The strongest projects will make Web3 easier to understand or use.
Instead of building something complex for complexity’s sake, focus on a real problem: monitoring, explaining, classifying, summarizing, or automating.
14. AI-Enabled Developer Tooling
Developers are one of the clearest audiences for useful AI tools.
Builders could create:
- Code review agents
- Documentation generators
- Test generation workflows
- Bug reproduction agents
- Error explanation tools
- Repository analysis agents
- Pull request summarizers
- Developer onboarding assistants
A strong project could connect to a GitHub repository and produce a useful output: code quality notes, security warnings, missing tests, documentation gaps, or architecture summaries.
The best developer tools save time immediately.
If a builder can create something that removes repetitive work from a developer’s day, that is a strong hackathon direction.
15. Personal Agents That Manage Digital Lives
Personal agents are one of the most relatable AI categories.
People do not only need AI for work. They need help managing the overwhelming amount of information, tasks, messages, plans, and decisions in daily life.
Ideas include:
- A personal admin agent
- A travel planning agent
- A finance organization agent
- A health routine assistant
- A digital decluttering agent
- A personal knowledge manager
- A calendar and task prioritization agent
- An inbox summarization and response assistant
The opportunity is to build something that feels like a real assistant, not another app to manage.
A strong project could help users reduce mental load by turning messy inputs into clear next steps.
16. Trading Agents
Trading agents are a popular and challenging category.
Builders could create AI systems that analyze market data, track signals, summarize market conditions, or assist with strategy research.
Possible projects:
- A market research agent
- A DeFi trading signal assistant
- A portfolio monitoring agent
- A risk analysis workflow
- A news and sentiment analysis agent
- An on-chain trading research assistant
- A backtesting assistant
The strongest projects should be careful with positioning. Rather than promising profits, focus on research, monitoring, analysis, and decision support.
A good trading agent should help users understand information faster, compare signals, and evaluate risk.
17. Code Review Agents
Code review agents are one of the most immediately useful AI developer tools.
A project could analyze pull requests and identify:
- Bugs
- Security issues
- Missing tests
- Poor documentation
- Performance concerns
- Repeated code
- Style inconsistencies
- Risky changes
A more advanced version could explain the impact of the change, suggest fixes, and generate test cases.
This is a strong hackathon idea because the value is easy to show. Upload code, run the review, and receive actionable feedback.
What Makes a Strong Hackathon Project?
The best projects will not necessarily be the most complex.
They will be the ones that clearly show:
- A real problem
- A real product
- A specific target demographic
- A working AI workflow
- A reason GPU compute matters
- A simple user experience
- A clear before-and-after result
A project does not need to solve the entire future of AI. It just needs to make one workflow faster, smarter, cheaper, more autonomous, or easier to use.
That could be a tool for creators. A system for developers. An agent for traders. A pipeline for documents. A workflow for video. A personal assistant. A Web3 automation layer.
The opportunity is to build something that shows AI as infrastructure for action.
Build What AI Will Actually Need Next
The next generation of AI products will need more than chat interfaces.
They will need compute. They will need orchestration. They will need workflows. They will need agents that can process data, generate media, analyze content, interact with tools, and improve over time.
That is why the HackerNoon x Nosana Decentralized AI Hackathon is a strong opportunity for builders.
It is a chance to explore what AI can become when it is not limited to a single prompt, a single model, or a single centralized infrastructure layer.
Build the workflow.
Build the agent.
Build the system that turns AI from an answer machine into something that can actually get work done.
Useful Links
Stay Updated with Nosana
Get the latest insights on AI infrastructure, GPU launches, and network innovations — all in one place
