Eighteen months into deploying Rimu Fund I, it feels like the right time to write down what we've learned — about the sector, about the specific companies we've backed, and about the dimensions of the original thesis that have held up versus the ones that needed refinement. Fund I closed in June 2023 with $42M NZD committed, a portfolio of seven investments at that point, and a thesis built on three convictions: that AI was creating durable value in the assessment layer of education, that knowledge infrastructure broadly was the next-wave software category, and that Australasia had structural advantages as a building environment for EdTech companies targeting English-speaking institutional markets.
The assessment infrastructure conviction has strengthened considerably. What we've seen across our portfolio and in the broader Australasian market is that the companies generating the most defensible early traction are those that sit in the assessment and feedback loop — Learnosity's API-layer play, Amira Learning's literacy assessment infrastructure, Panorama Education's engagement analytics on top of student activity data. These companies share a structural characteristic: they produce data outputs that other parts of the educational technology stack depend on. That dependency creates a moat that is not purely technological — it's also organisational, because once a school's assessment workflow runs through your API, switching requires re-training staff and re-integrating data flows, not just choosing a competing product. We entered the fund with a hypothesis about assessment infrastructure and the evidence eighteen months in is that the hypothesis was right, though the specific mechanisms matter more than we initially appreciated. Not all assessment companies are infrastructure companies; the ones that are infrastructure have APIs and data models designed to be depended on, not just used.
The knowledge infrastructure thesis has proven harder to invest against than we expected, for a different reason. The category is real — the market for tools that help knowledge workers manage, curate, and certify their intellectual output is large and largely unserved. But the customer discovery for knowledge infrastructure products is complex because the need is latent: buyers don't yet know what they're missing. Authory's content archive and portfolio product is navigating this by finding a specific customer archetype — the professional writer or journalist who loses work to platform changes and wants an owned archive — where the need is acute rather than latent. The lesson we've taken from this is that knowledge infrastructure is a genuine long-term category, but the path to market for Fund I companies in this category is likely through specific high-pain customer archetypes, not through top-down category creation.
On the Australasian structural advantage thesis, the evidence is mixed in an interesting way. The advantage is real, but it's more contingent than we stated it in the original thesis. NZ and Australian institutional buyers are genuinely more willing to pilot than most markets, and the quality of founder-institution relationships we see in companies that have NZ school or university partnerships is very high. But the advantage only converts to a go-to-market asset if the founder can use the NZ pilot credibly as evidence for an Australian or broader English-speaking market expansion. That requires building the pilot with the right metrics from day one — not just adoption rates, which are easy to get from a willing NZ pilot school, but learning outcome data and teacher time-savings data that can pass the due-diligence bar of a larger institution. Founders who treated the NZ pilot as a proof-of-concept rather than a commercial template got stuck; founders who treated it as a case study they would need to replicate at ten times the scale designed it differently from the start.
What we still don't know is whether the AI capability inflection happening throughout 2024 will primarily benefit the infrastructure companies we've backed or will create a new cohort of application-layer companies that make some of our infrastructure investments less necessary. Our current view is that well-built infrastructure becomes more valuable when application-layer capabilities improve, because more applications need better infrastructure to support them. But that view depends on our portfolio companies having genuinely open and interoperable architectures — not building feature moats that inadvertently lock the infrastructure in, too. That's a portfolio architecture question we're actively working through with the founders as we enter the second half of the deployment period.