Unnecessarily clever. Occasionally useful.
Nerd Lab is an open, experimental playground for people who like to explore technology in unconventional ways — a place for knowledge creation, awareness building, and showing colleagues what these tools actually do on a normal working day. It is not a company, not a course, and not a corporate AI initiative; it is a group of true nerds in the making, and the entrance requirement is curiosity.
Learn by building. Nothing here starts with a curriculum. We take a problem that is mildly infuriating, overengineer a solution to it, and work out what we learned on the way back out.
Make new tech approachable for the curious. AI agents, RAG, automation, data, web and app building — most of it stops being intimidating the moment you watch someone else do it badly first. We do that part in public, on purpose.
Open source everything. Every experiment, including the ones that did not work. Especially those — the failure is usually the interesting half.
| Experiment | What it does | Status |
|---|---|---|
| UN Data Commons Field Guide · source | A practical field guide and live query lab for the UN System Data Commons. It teaches you to ask well, because the search box will not tell you when you asked badly. | live |
| AI at UNGA81 · source | The week AI took the floor: 133 UN-linked events on AI around the 81st General Assembly, plus every General Debate statement that mentioned it. | live |
| Country Classification Commons · source | How many countries are in Africa? 54, 60 or 48, depending on which official list you ask. Every UN, World Bank and OECD grouping in one auto-refreshing table, with a column that settles the argument. | live |
| UNAIVERSE · source | An interactive galaxy-style explorer of the UN's journey with AI. Every link in it is official. The orbital mechanics are editorial. | live |
| undata · source | UN data in Python, without the guessing. pip install undata, type a country's name, get a table with UN names and a source on every row, then save it to Excel. |
live · PyPI |
| Four Years Left · source | Eleven years into the 2030 Agenda, with four to go. A scrollytelling briefing on AI and the SDGs: a deck at the front of the room, an article afterwards, and a live quiz that keeps the room honest. | live |
| The Future Is Not Artificial · source | AI feels like software. At scale it is also land, electricity, fibre, cooling and water. An interactive documentary where the website is the talk. | live |
| The Ledger of Trust · source | Thirty years of Bangladesh's bad-loan ratio, followed one number at a time. Bangla-first, with a full English toggle. | live |
| 3I/ATLAS · source | A data-driven comedy special on the interstellar comet that is definitely not a spaceship. The comet is real. The alien hypothesis is satire. | live |
You do not need an AI background. You do not need a data background. You do not need a technical background at all. If you are the kind of person who opens the settings menu just to see what is in there, you are already qualified.
Everything we build is open source. People who keep showing up become co-maintainers — not as a reward, just as an accurate description of what has already happened.
Two doors, both unlocked:
- Discussions — ask the question you think is too basic. It is not. This is the best place to start if you have never used GitHub before.
- Issues — pick something up, or report that something is broken. Reporting counts as contributing.
Neither door asks what you do for a living.
Built in public, broken in public, documented either way.