Open Model Strategy & Economics
Model Strategy & Training
Compare open, local, and hosted models by workload quality, token efficiency, latency, hardware requirements, and the cost of completing useful work.
View Research →Trilogy AI Center of Excellence
Practical AI research, implementation notes, and field-tested resources for leaders and builders working AI-first.
Substack
Get the AI COE signal as it ships
New research briefs, technical walkthroughs, and practical AI implementation notes from the Trilogy AI Center of Excellence.
Open SubstackFeatured Research
Highlights from our ongoing and completed research initiatives.
Autonomous Coding Agents
Agentic Engineering
Evaluate coding agents on real repositories: code understanding, edits, tests, review, context management, and the cost of reaching an accepted result.
View Research →
AI Security & Governance
Agent Security & AI Cybersecurity
Investigate credential mediation, execution isolation, repository exposure, and AI-assisted security audits, with evidence capture and verification of fixes.
View Research →Latest Publications
Stay updated with our newest findings, analyses, and insights.
OpenAI dot: Why Would I Start a New Chat?
Agent Infrastructure | 2026-10-01
A hands-on look at OpenAI's always-on dot agent and what changes when an assistant keeps context, continues work between conversations, and reports its own limits.
Read Publication →
Jev's Open Rivals Test the Decision Model Idea
Evaluation & Reliability | 2026-09-28
Five routes to typed, probability-bearing decisions — TypeSafe's hosted Jev and open rivals Laya, Kev, AnyJev, and DiffusionGemma — are judged by held-out coverage and accepted-case error rather than interface claims.
Read Publication →
A Charter Is Not a Control
AI Security & Governance | 2026-09-14
Applying a name-the-loser, name-the-mechanism test to frontier AI governance proposals shows charters and voluntary audits are descriptions of cooperation, while the June 12 export-control directive was the only enforceable control on record.
Read Publication →