Blog
Writing about what we're building, what we're finding, and what we still don't know. Filter by tag.
May 10, 2026 2 min read
The Nicholas Foundation runs two research initiatives in parallel: Rule-Before-Train AI Safety and Privacy-First Mobile OS. Both are slow. Both are honest about being slow. Here's how they connect, and what the next 12 months look like.
May 4, 2026 3 min read
Can a small team ship a daily-driver privacy-first mobile OS in three years? Our feasibility study says yes — with caveats. Here's the short version of what we found, what we'd pivot on, and what the next 12 months look like.
April 22, 2026 2 min read
Modern AI safety is retrofitted. Train a model on broad corpora, then patch behaviour with RLHF and content filters. Nyx inverts that — start with explicit rules, extend with ML — and the engineering trade-offs are clearer than you'd expect.
Type to search.