What makes a great demo at AI Tinkerers?

Core Philosophy

AI Tinkerers showcases cutting-edge work in AI, particularly with LLMs and generative AI. Whether you’re building sophisticated AI workflows from the ground up, training or fine-tuning a new model, or creating a sophisticated app by orchestrating coding agents, we’re here to see what you built and how you built it—your stack, workflows, and agentic processes.

These are not pitches. Skip the market overview and jump right to the implementation—show us your demo and walk us through your process.

AI Tinkerers is like ‘The Homebrew Computer Club of AI’, not YC demo day. This is about enabling a fellow tinkerer. It’s not about pitching your startup.

Looking for inspiration? Watch this video by Sharif Shameem about following your curiosity and creating great demos. Fantastic!

✨ AI Insights ✨

We asked AI to analyze hundreds of AI Tinkerers demos based on attendee feedback surveys. Here’s what we found:

A great Dev Tools demo feels like a code-first experiment you can understand, evaluate, and (ideally) try yourself: it should be centered on a specific technical problem, show the actual system behavior end-to-end (inputs → processing → outputs), and include enough implementation detail to support audience transfer to their own stacks. Highest-rated talks tend to pair a clear, relatable use case with concrete engineering choices—e.g., local-first execution, Docker/Ollama deployment, MCP tool exposure, or mechanisms for turning raw model output into structured artifacts—so attendees walk away with “what to build next” rather than just “cool idea.” They also tend to emphasize transparency and user/workflow ownership (self-hosting/data control, personalized context, offline/local execution), which makes the technical tradeoffs tangible. Avoid demos that are primarily product-level overviews, are hard to follow in the timebox due to unclear problem framing, or suffer from slow/fragile live setup—feedback indicates those factors cause attention loss. Finally, be ready to supply a concrete example path or demo script during Q&A, since even positive responses often request deeper use-case demonstrations.

In ADeus (New York City), Martynas Krupskis’s team demo is compelling to an AI/ML audience because it turns real-world audio into a continuously updated, personalized context store on a user-owned server, combining wearable capture, transcription, and an open, self-hosted Supabase/LLM-backed architecture—rated 5/5 by multiple attendees. In Make text output actually useful - worksheet builder (Denver), Kristen Montesano’s demo focuses on the practical engineering question “how do I make genAI text outputs usable?” by pairing AI generation with a Canva-like drag-and-drop worksheet builder, and the audience gave 5/5 consistently. In AI powered reading buddy - what Kindle should be, but isn’t (Toronto), Mayank Singh’s “smarter Kindle” approach is appealing because it emphasizes line-by-line reading/explanation via a web-and-Docker/Ollama workflow and plans for graph-search context and a personal paper library; attendees found it “compelling.” In ​Agentic understanding of large codebases (Vancouver), Michael Sukkarieh describes Sourcebot exposing code-search and navigation via an MCP server so flagship agents can query large multi-repo codebases; audience feedback explicitly notes the tool’s usefulness and mentions similarity to Copilot Chat (“I will try out your and see how it compares!”), which is strong validation for a technical, developer-facing demo. And in Private Agents in your browser (Chicago), Rakesh Kumar’s demo succeeds because it shows agentic tasks running with small models directly in-browser/on a laptop—low latency and cost—earning 5/5 from attendees.

🔮 Guidance

Focus on Live Demonstrations: Show, don’t tell: Prioritize showing your project in action over static slides or lengthy explanations. Let the audience see the magic happen—whether that’s a running project or a workflow/stack in action.

Show the Internals: Expose the guts of your build—traditional code, builder graphs, AI-assisted development, or the workflows, context setups, and tool choices behind your approach. This now includes:

• Traditional Code: Walk through your scripts, functions, and technical architecture
• Vibe-Code Workflows: Show your serverless flows, node chains, API orchestrations, and platform configurations
• AI-Powered Methodologies: Demonstrate how you used AI to transform business requirements into product requirements, your prompt engineering strategies, or novel AI-human collaboration workflows
• Tool Comparisons: Share what you tried, what worked, what didn’t, and why—especially valuable for emerging vibe-coding platforms

Maintain Technical Depth: The bar is high for all presenters. For vibe coders specifically:

• Show sophisticated orchestration that would impress traditional developers
• Demonstrate complex problem-solving using AI-powered tools
• Explain the technical challenges you overcame, even if the solutions involved AI assistance rather than manual coding
• Avoid superficial tool demos—show depth, creativity, and innovation that pushes boundaries

Highlight Uniqueness and Relevance: What makes your project special? How does it align with or challenge current trends in AI? What were the hard-earned surprising lessons you learned? For vibe coders, this might include novel approaches to AI collaboration, unexpected platform capabilities, or innovative ways to achieve complex functionality without traditional programming.

Prioritize Clarity and Engagement: Guide the audience through your demo with clear logical flow. Whether showing code or AI workflows, explain your technical choices and define any jargon. Help fellow builders understand not just what you built, but your decision-making process.

Technical Considerations: Get ready before you’re called on stage. When showing code, use a large font size for easy viewing. When showing visual workflows, zoom in so nodes and connections are readable. If demoing AI conversations or prompt engineering, make the text large enough to read from the back. Speak clearly and loudly so everyone can hear you.

Avoid the Pitfalls: This is not a platform for promoting your startup or services. Skip the deck. A single architecture diagram or workflow screenshot is fine if it clarifies the build, but otherwise focus on the live demo and detailed explanation of how you built it. Pop the hood and show the build. No videos: show the thing running, not the video of the thing running.

🤖 AI Tinkerers Demo Vibes

🤖 Demo FAQ

Q: How long should my demo be?

A: Aim for a concise 5-minute presentation. Jump right in by showing what you built running live or by walking us through the workflow/tooling you used to build it. Then explain your process and choices. Try to focus entirely on answering the implicit question, “how can I enable a fellow tinkerer with what I’m showing today?”

Depending on the event and how many presentations there are, there may be room for a couple of quick questions from the audience at the end of your presentation. Check with your host.

Q: Can I use slides?

A: You don’t need slides. Focus on (a) showing the thing you built running live, or (b) walking fellow tinkerers through how you built it—your workflow, tool stack, or agent setup.

Skip all the preamble, background, market opportunity, etc. This is not a forum for promoting your startup and getting customers. There are plenty of other forums for that. AI Tinkerers is about builders enabling fellow builders and pushing leading edge of innovation in AI.

Nobody is coming to hear pitches. However, if you need a slide to show some data visualization of diagram pertinent to the core goal of your demo, then of course that is fine.

Q: What if my project is still in development?

A: That sounds great!! We welcome half-baked side projects. AI Tinkerers is a forum for innovation among builders. It’s a safe place to show things that are barely working.

Some of the very best presenters have started with, “Hi, I’m [name], a [impressive senior technical role such as CTO] from [important company you’ve heard of], but today I’m going to show you something I’m hacking on the side…”

Q: Should I focus on a specific aspect of my project?

A: Yes, highlighting a particular technical challenge and your solution will make your demo more focused and engaging than trying to show the entire project. Think: for an audience of folks who are quite deep into all the latest things, what is the particular aspect of your project that would be most novel and where you’ve likely gone further than others?

What specific aspect of your build would be most helpful to enable a fellow builder with their projects?

Q: I built something sophisticated using vibe-coding tools like Cursor, Replit, or no-code platforms. Do I qualify?

A: Absolutely—if you can demonstrate technical depth and sophisticated problem-solving. Show us your workflow orchestration, explain your AI collaboration methodology, walk through complex logic chains, or demonstrate how you pushed these tools beyond their obvious use cases. The key is impressing our technical audience with the depth and innovation of your approach, regardless of whether you wrote traditional code.

“In the future, everyone will be AI Tinkerers” has been on our homepage since even before our first meetup, on the belief that humans increasingly will build and manage AIs through tinkering as the primary means of working in the future.

Our core is people who (a) can build, (b) are building as their day job, (c) are producing and sharing innovations at AI Tinkerers meetups that are pushing the field forward.

Q: How do I show “internals” for a vibe-coded project?

A: Pop the hood on your process: Show your workflow configurations, prompt engineering strategies, API orchestrations, or AI-assisted development methodologies. Walk us through your most complex logic chains, explain your tool selection rationale, or demonstrate how you solved technical challenges using AI collaboration. The goal is helping fellow builders understand your sophisticated approach.

Q: What makes a vibe-coding demo stand out?

A: Demonstrate that you’ve gone deep enough to impress traditional developers. Highlight your most sophisticated workflow, your most complex AI orchestration, or your most innovative use of the platform. What specific methodology or technical approach would be most helpful to enable fellow builders? What did you discover that pushes the boundaries of what’s possible with these tools?

• Show complex problem-solving using AI-powered methodologies
• Explain sophisticated workflows that others could learn from
• Highlight novel approaches or unexpected platform capabilities that aren’t widely known
• Share hard-won lessons about what works and what doesn’t
• Demonstrate creativity and technical depth, not just tool usage

Q: What if my demo involves both traditional code and vibe-coding?

A: Perfect! Show how you combined approaches, explain your decision-making process for when to code versus when to use AI-assisted tools, and share lessons about hybrid development methodologies. This kind of sophisticated tool orchestration is exactly what our community wants to learn about.

Q: Can I talk about my company or product?

A: No pitching! Focus on enabling fellow builders, not on promoting your company or product. If a company name appears during your presentation, it should be in a stack trace.

Q: What if my demo doesn’t work perfectly?

A: Don’t worry! We’re all tinkerers here, and we understand that things don’t always go as planned. Embrace the imperfections and
share your learnings.

Q: What makes a demo stand out at AI Tinkerers?

A: Demos that showcase novel solutions, highlight technical challenges, surface workflow/tooling insights, and offer takeaways from the build process are always appreciated.

As are creative and non-obvious applications of AI that are fresh and haven’t been seen before.

And hearing from the protagonists of popular or innovative new open source projects are always exciting!

Q: Can I present on a research paper?

A: Absolutely. While AI Tinkerers primarily focuses on demos of working projects, if you’ve published something profound and interesting, it can be valuable for the community to hear directly from the author, especially if you can connect it to practical applications or potential future developments. Consider incorporating interactive elements to make your presentation more engaging.

Q: Any tips for visual aids?

A: Absolutely! Use a large font size (at least 24pt) for any text on your slides or visuals. Ensure your visuals are clear, concise, and easy to understand from a distance. If you’re showing a node graph, collapse unrelated branches to reduce clutter.

Q: How can I ensure my presentation is engaging?

A: Speak clearly and enthusiastically, maintaining a good pace and varying your tone. Make eye contact with the audience and encourage questions. Practice your presentation beforehand to ensure a smooth delivery. Run the presentation through whisper piped to GTP-4 using a prompt similar to https://gist.github.com/jheitzeb/528f5d7634ecc7b7c4155d07e4fd6d2a (replace PRESENTATION_TRANSCRIPT with your transcript)

Q: What are your standards for all builders?

A: Whether you’re writing raw PyTorch or orchestrating sophisticated AI workflows, the standard remains the same: demonstrate technical depth, share actionable insights, and enable fellow builders to learn from your experience. The tools may vary, but the innovation, creativity, and technical sophistication should impress any developer in the room.

Q: How do you choose which projects are selected?

A: Based on the criteria above. Please provide lots of information about your project when you submit it. If you haven’t submitted much information, the organizers are likely to pass over your project because it’s just not possible to determine how interesting it could be. Also, fill out your networking profile and add your social links – all of that information is available to organizers when they decide on projects.

We’re excited to see your demo come to life! Thanks for sharing your work with the AI Tinkerers community.