EP 86: This Kid Built A Fusion Reactor In His Bedroom Using AI

Episode Summary

Executive Summary: The episode spotlights teenager/Waterloo math student Hadza, whose viral post showed him using AI tools, especially Claude Sonnet, to organize and accelerate building a demo nuclear fusor. The conversation focuses on his AI-native workflow, how he used long-context projects as a second brain, the safety/physics considerations of high-voltage hardware, and what his project suggests about the future of AI-assisted making and learning.

Main Topics: Viral fusor project and public reaction (Priority: 5/5): The hosts discuss the tweet/video that blew up online: a teenager in Waterloo building a nuclear fusor with AI assistance. They frame it as a striking example of hacker spirit and a glimpse of the future. Hadza’s background and motivation (Priority: 4/5): Hadza explains he is primarily a software person and math student at Waterloo, originally from Sri Lanka, who wanted to tackle a harder, more physical project after feeling less satisfied with software-only work. AI-native workflow with Claude, ChatGPT, Cursor, and projects (Priority: 5/5): He details how he uses AI for everyday tasks and especially for deep project organization: dumping notes, emails, PDFs, and conversations into Claude Projects to create a searchable, recursive planning system. How he built the demo fusor (Priority: 5/5): Hadza walks through the learning process: reading forums and videos, using Claude to break down parts/materials, checking McMaster/CAD-style components, and assembling a step-one demo rather than jumping straight to fusion. Safety, risk, and verification (Priority: 5/5): The hosts press him on safety. He describes concerns about high voltage, radiation, and x-rays, says the demo version is intended to avoid dangerous neutron output, and emphasizes consulting experts and double-checking calculations. What Claude does well and product gaps (Priority: 4/5): He praises Claude Sonnet for long-context memory, math/physics explanations, diagramming, and project-based reasoning, while criticizing missing features like URL ingestion and faster voice-driven workflows. Future plans and funding goals (Priority: 4/5): He says the demo fusor is only step one, with goals of joining the Plasma Club and eventually the Neutron Club, and that he posted the project partly to strengthen an Emergent Ventures funding application.

Key Arguments: AI projects are most powerful when treated as a long-context “second brain” that can store conversations, documents, and research, not just generate code. Claude Projects helped him turn chaotic research into a recursive breakdown of tasks, rapidly producing boilerplate and implementation plans. The demo fusor was chosen as a safer first step toward fusion because jumping directly to a full fusor would have been too dangerous. Safety in hardware projects requires both AI assistance and human verification; he used experts, datasheets, and calculation checks to reduce risk. Claude is especially useful for explaining physics/math and generating diagrams, making it valuable for non-software domains. The future of AI use may look like children’s natural prompting habits: quick, direct, and task-oriented, which he sees as a model adults should learn from. His public post was not only for attention but also a practical artifact to show funders concrete progress and justify further support.

Data Points: Views on the viral post: almost 3 million - Hadza’s tweet/post about the fusor gained nearly 3 million views Location: Waterloo - The project was built while he was studying at Waterloo Age group: teenager - The hosts repeatedly describe him as a teenager/very young builder High voltage concern: 12,000 volts / 5,000 volts - Hadza cites the high-voltage risk level while discussing the build Electron energy in demo: 3 kiloelectron volts - He says the demo fusor’s electrons are only at about 3 keV, which he argues is not enough to cause harm Timeline target: 21 days - He says he aims to advance toward the full fusor within roughly 21 days Project cost: $2,000–$3,000 - Estimated cost of the demo fusor build Full fusor cost estimate: around $10,000 - He estimates the more advanced build could cost about ten thousand dollars Demo step status: Step one - He describes the current build as a first step toward a full fusor Existing AI usage duration: 1.5–2 years - He says he has used Cursor for about two years and has been on GPT Pro for a very long time University move date: 2021 - He moved to Canada in 2021 to start university at Waterloo Current personal status: gap term / no job - He says having no job and being on a gap term made it easier to pursue the build

Pivotal Quotes: "I think every once in a while, you get something online where it makes people go, I'm seeing a glimpse of the future." — Host: Opening reaction to the viral fusor post "I just want to give a glimpse into the future, like you said." — Hadza: Hadza responding to the host and framing the project "My motivation right now is in like, oh, fix energy or whatever it might be. But right now, it's like, I only just want to get into the neutron club." — Hadza: He explains his current goal and what drives the next step of the project

Implications: The episode suggests AI is evolving from a coding helper into a general-purpose research and maker companion. For builders, long-context AI plus expert verification can dramatically lower the barrier to complex hardware projects—while also raising new safety and reliability questions.