Research updates, company news, and perspectives from the Eternis team.
There are two fundamental mechanisms for aggregating human knowledge into collective action: markets and democracy. Both are about to change radically. Here's what that means for forecasting, coordination, and how we allocate our time.
Everyone means something different by "world model." We think the useful framing is simpler: a cheap, continuously updating projection of the world that makes search and decision-making faster. Here's what that means and why we're building one.
Chat interfaces are optimized for the wrong thing. The goal isn't to generate text — it's to drive an outcome. Axion is an agentic decision-making platform built for investors, allocators, and policy makers who need certainty over their next step.
We asked three frontier models the same forecasting question five times each. One model's answers ranged from 45% to 78%. Then we tried whispering "75%" beforehand — and watched one model shift by 40 percentage points.
An 8-dimensional evaluation framework for LLM-based forecasting systems, grounded in 56 papers and validated against 675 experimental API calls across three frontier models.
We trained an 8B-parameter model that surpasses all published baselines on open-ended forecasting — including models 10-15x larger. Here's how context building, calibrated reward design, and modified GRPO got us there.
Our thesis on why the next generation of AI must be able to continuously refine its own architecture — and what that means for forecasting, sovereignty, and the future of intelligence.
