Community Catchup #8

The eighth community catchup mixed community strategy with a deep founder demo. Wellington registrations were surging, the group debated how New Zealand should approach AI education and infrastructure, and Yadwinder walked through Nimble's AI-powered QA platform.

Highlights

The session kept returning to one practical question: how do we make AI useful and accessible beyond the people already living at the frontier? That meant token access for meetups, education for non-technical people, safer deployment models for businesses, and tools that package agentic capability into something a founder can actually buy and use.

Wellington Is Heating Up

Adam shared that the Wellington meetup was closing in on 300 registrations, with roughly 200 likely to be accepted based on show-up rates from Auckland. That raised the next challenge: if people are coming to learn hands-on, they need access to real tools and tokens.

The group discussed whether frontier labs, startup programmes, universities, or government could help fund practical AI education. Steve was exploring similar ideas for Christchurch, and Adam mentioned an upcoming AUT meetup as a possible start of a broader partnership.

TheColab, Claude Community, and the Education Gap

Yadwinder asked directly what TheColab is trying to achieve. Adam’s answer was simple: build a Kiwi community around AI, and help people who are not already in tech understand what is coming.

TheColab stays model-agnostic, but Adam also shared plans for Claude Community NZ as a Claude-focused home for ambassador meetups and resources. The group agreed that the real gap is not enthusiasm among builders — it is practical education for students, businesses, and everyday workers who are about to be affected by AI whether they follow the news or not.

Nimble Makes QA Self-Serve

Yadwinder’s Nimble demo was the centrepiece. The pitch: AI is letting people ship software faster than humans can test it, so QA needs to become cheaper, faster, and more accessible.

Nimble’s agents cover a wide range of testing and improvement workflows. A process designer turns natural-language workflows into editable swimlane diagrams and value stream maps. A browser extension records user journeys and converts them into testable English steps plus video evidence. Nimble Scan and Nimble Rank give business owners independent website quality and performance reports. Nimble Crawl tests authenticated areas. Nimble Flow takes OpenAPI or GraphQL specs, generates integration tests, runs them, and reports results.

The commercial positioning was just as interesting: $10 transactional tests for small users, Azure and AWS Marketplace deployment for enterprises, and fixed-price products instead of asking customers to understand token costs.

Frontier Models, Local Models, and Kiwi Infrastructure

The model discussion ranged widely. The group talked about Gemini Flash speed, diffusion-style text generation, Chinese open models, Cursor’s Composer pricing, OpenAI committed capacity, NVIDIA, and whether New Zealand should train or distil its own models.

The consensus was pragmatic: for many businesses, a few hundred dollars a month for frontier models is trivial compared with employee time. But smaller local models are increasingly useful for narrow tasks, sensitive workflows, and cost-controlled automation. Ollama, local GPUs, and right-sized models all have a place — especially when the task is structured enough that a mini model performs almost as well.

The Social Side Is Coming Fast

The final stretch moved from tooling into society. Schools are still telling children AI is cheating. Government agencies are starting to talk about AI but may not have the internal expertise to buy or implement it well. Public backlash is already visible, and the group discussed the risk of a divide between people who adopt AI and people who reject it.

Adam’s framing was blunt: AI is not going away. The useful work now is helping people understand it, use it safely, and avoid leaving the future entirely to the loudest vendors and the biggest consultancies.