Insights

Analysis & commentary

Perspectives on healthcare strategy, policy, and industry from the field.

Technical Jul 2026 8 min

We Re-Ran Our Own Benchmark. Here's What Broke, and What It Cost Us.

We found four confounds in our own MuSiQue benchmark program — input budgets, output provisioning, reasoning-policy semantics, reader-mode asymmetry. The corrected record includes the strongest result on our own harness belonging to the baseline. Full accounting, pre-registered predictions, and the public repo.

Analysis Jun 2026 15 min

The Quality Wall: Why RAG Retrieval Degrades Silently and What the Research Actually Shows

A synthesis of 31 peer-reviewed papers documenting twelve failure modes in RAG quality — why retrieval degrades at scale, why multi-hop breaks vector search, why hallucinations persist despite retrieval, and what the research says about fixing it.

Technical Jun 2026 14 min

The Deployment Mess: How RAG Became the Broken Link in Enterprise AI

Five databases. Five query languages. Five points of failure. The RAG infrastructure mess is why 72-80% of enterprise RAG implementations never reach production. Here's the architecture analysis, what stack collapse looks like, and benchmark data on a single-binary alternative.

Analysis May 2026 16 min

The Multi-Source Synthesis Problem in Healthcare AI: Is Multi-Hop Retrieval the Missing Key Piece?

AI saves lives when it looks at one thing. It struggles when it needs to connect multiple things. We trace the pattern, examine the successes, and identify the infrastructure gap.

Technical May 2026 18 min

RAG Needs to Start Adding Real Value

A February 2026 NBER study of 6,000 executives found over 80% of firms report no measurable productivity gains from AI. The problem isn't the models — it's what we're feeding them. We benchmarked our patent-pending retrieval engine on one of the hardest multi-hop QA benchmarks using the exact protocol and corpus of the leading academic system. Here are the results, the methodology, and the honest limitations.

Press Release Feb 2025 2 min

Analysis & commentary

We Re-Ran Our Own Benchmark. Here's What Broke, and What It Cost Us.

The Quality Wall: Why RAG Retrieval Degrades Silently and What the Research Actually Shows

The Deployment Mess: How RAG Became the Broken Link in Enterprise AI

The Multi-Source Synthesis Problem in Healthcare AI: Is Multi-Hop Retrieval the Missing Key Piece?

RAG Needs to Start Adding Real Value

[Press Release] EluciDx Canada Ltd. and UBC Finalize Licensing Agreement for Oral Amphotericin B

Regarding monoclonal antibody access initiatives for LMICs: A pragmatic and straightforward opinion

A Case For The Perfect Healthcare System (for now)

Employee Incentivization: Getting it so wrong is so easy

We have a definition problem in healthcare

The Hubris of Equal Access in Healthcare

The Messed-Up Value of Healthcare

Why aren't we tackling the asymptomatic spread of COVID-19 more aggressively?

When the dust settles… (after COVID-19)