why trust img

Why Trust Web3Bet

Our team of experts has independently reviewed and evaluated all the products and services featured on this page to ensure you receive accurate and reliable information

Stake co-founder Ed Craven has backed Melbourne-based MainCode, a bid to train a large language model entirely on Australian soil and operate it under local rules. The team is led by Dave Lemphers, former EasyGo CTO (the group behind Stake and Kick), who stepped down in November 2024 to focus on the venture.

The roadmap targets a late-2025 release and positions the model as a domestic alternative to leading US and Chinese systems. The emphasis is on legal certainty, freedom from third-party APIs, and reduced exposure when handling sensitive information.

Owning the stack: infrastructure and team

MainCode is building out in-country compute: GPU clusters, storage, and training pipelines. By avoiding rented platforms, the company seeks end-to-end control over data paths and training parameters. Upfront costs are higher, but the trade-off is transparent processes and predictable unit economics.

The team brings PhD talent from the University of Melbourne and RMIT, plus experience from Google and the Allen Institute for AI. Rather than relying on light fine-tuning, MainCode prioritizes from-scratch training, enabling stronger auditability, reproducibility, and long-term maintainability.

Why it matters for the market

Enterprises and public agencies in Australia need data residency and governance within a single legal framework. A local LLM streamlines privacy compliance, clarifies dataset provenance, and reduces reliance on opaque overseas services.

Following the initial release, MainCode plans pilot deployments with corporate and government clients, then iterative feature expansion. If milestones hold, Australia gains a self-reliant AI provider with full lifecycle control — from data preparation through to production use.

Source: https://www.forbes.com.au/news/innovation/ed-cravens-multi-million-dollar-bet-on-australias-answer-to-openai/