Builder demos from the weekly show-and-tell

Members shared projects from an MRT visualisation to a research reproduction and a brain-computer interface experiment.

Editorial attribution: Melverick Ng, Nexius Labs. Published 31 October 2026. Method and evidence limits.

Community-led discussion article. Observed 24–30 October 2026 (Asia/Singapore); anonymous paraphrase, ABC group (single-group depth). Privacy-safe paraphrases; no member quotation or identity is published.

What the signal means

The ABC group's weekly show-and-tell produced a varied set of projects, and together they show how wide the range of things individual builders now ship has become.

Shared projects included a 3D visualisation of Singapore's MRT network built with a frontier coding model; a service and MCP server that gives AI-made web pages a proper public address; an attempt to reproduce a research project on flapping-wing fluid dynamics with a coding agent, with solvers in two languages that the member said largely matched the original results on a set of test cases; a brain-computer interface experiment built with open hardware and Python, with an AI coding assistant as the main development tool; a small finder for children's playgrounds built inside a chat assistant; an update to a member's identity-photo phone app; and a bot that reports the usage-limit status of several AI services every few hours.

Later in the week, members also shared an open-source project that turns a cheap voice remote into a controller for coding by voice, and an MCP server for looking up card merchant category codes.

Practical implications

- Solo builders are shipping across domains that used to need specialist teams, from scientific computing to hardware. - Hosting and publishing are still friction points. Tools that make AI-made pages easy to share are filling a real gap. - MCP servers are becoming a normal way to package small, useful data services for agents. - Monitoring AI service limits has become a tool category of its own, which says a lot about how much limits matter to daily work.

Content and community opportunities

- Feature community demos as case studies of what one person can build in a weekend, with the process rather than only the result. - Use the research reproduction as a teaching example: how to check an agent's scientific output against known results. - Run a short session on packaging a small dataset as an MCP server.

Risks and open questions

- Demos were not independently tested for this edition, and a reported agreement rate on test cases is the member's own claim. - Hobby projects touching biosignals or identity photos raise privacy and safety questions if they move beyond personal use. - A demo is not a product. Repeat usage and support needs are still unknown.

Watch next

- Which of these projects find users beyond the group. - More scientific-computing reproductions with agents, and how they are validated. - Growth in small, single-purpose MCP servers.