Commentary · AI · Machine Learning · Data

Where the demos end
and the architecture
begins.

Unfiltered technical commentary on AI, machine learning, and data systems. The industry has enough cheerleaders. This is not that.

About Jason Bell

Practitioner and commentator at the intersection of AI, machine learning, and data infrastructure. Writing about the gap between what the industry promises and what the architecture actually delivers.

Interested in market structure, yield systems, agentic workflows, and the incentive problems that make good technology fail in the real world.

AI Systems Quant Finance ML Data Architecture Market Structure
Featured
Latest piece
Agentic AI Aviation Market Structure March 2026

Agentic AI Won't Beat Aviation Yield Management.
It Will Feed It.

The prevailing assumption that AI agents will find consumers cheaper flights is structurally wrong. The architecture of aviation pricing guarantees the opposite — and deploying agents at scale will inflate prices for everyone.

Read the full analysis 12 min read · Technical
Agents search 8–15× per intent Demand signal inflates Yield engine raises price Agents re-search — more signals Routes switch — prices follow feedback loop
Price Inflation Feedback Loop · Fig. 4
All Articles
More commentary incoming
Agentic AI Aviation
Agentic AI Won't Beat Aviation Yield. It Will Feed It.
Why deploying AI booking agents at scale is a market-wide price inflation mechanism, not a consumer benefit.
Read Mar 2026
Coming Soon
Quant Finance
The 24-Hour Market Problem: Why Retail Traders Are the Product
When equity markets go 24h, the institutional extraction playbook from crypto will replay at scale. Agentic trading bots won't save retail — they'll accelerate the harvest.
Forthcoming Q2 2026
Coming Soon
Private Markets
Venture Valuations and the Hidden Markov Problem
The model works. The customer doesn't want it. Why the people who most need better valuation methodology have the least incentive to adopt it.
Forthcoming Q2 2026
The
Position
The technology industry has no shortage of optimists. What it lacks is people willing to look at the distribution architecture, the incentive structure, and the incumbent moat — and tell you honestly why the demo you just watched will never work at scale. That is what this is for.
Topics
Agentic AI
1 article
Where autonomous workflows meet the reality of closed systems and adversarial infrastructure.
Machine Learning
Coming
Applied ML, model decay, and why the training set is always the wrong one for production.
Market Structure
Coming
Yield management, PFOF, GDS gatekeeping, and the incentive problems that keep bad systems alive.
Data Systems
Coming
Why data is the moat, who holds it, and what that means for anyone trying to build on top.