04 — Writing
Research & Insights
Technical articles and research notes on AI systems, agentic architectures, real estate data infrastructure, and fintech intelligence.
Featured
MCP’s Stateless Turn: What the 2026 Protocol Reset Changes
MCP’s 2026 move toward a stateless core changes how agent infrastructure scales, secures requests, and supports long-running work. Here is what architects need to reconsider.
Multi-Agent Systems Need a Control Plane, Not Just Better Prompts
As specialized agents collaborate across tools and organizations, the hard problem becomes coordination: identity, authority, state, evidence, and recovery.
Why Most AI Deployments Fail Before They Ship
The gap between a working prototype and a production-grade AI system is wider than most teams expect. Here's what actually breaks — and how to build systems that survive contact with the real world.
All Articles
Confidential Computing Could Unlock the Next Wave of Collaborative AI
AI often needs sensitive data that organizations cannot safely pool. Confidential computing changes the trust boundary by protecting data while it is being processed.
Physical AI Is Moving Intelligence Out of the Screen
The next AI interface is not a chat window. Physical AI combines perception, world models, control, and robotics to act in environments where errors have material consequences.
Post-Quantum Migration Is an Agent Infrastructure Problem
Agentic systems are expanding the number of identities, delegations, and signed events just as cryptographic infrastructure begins a post-quantum transition.
AI Supercomputing Is Becoming an Orchestration Challenge
The next AI infrastructure advantage will come from coordinating heterogeneous compute, energy, data, and policy—not simply accumulating more accelerators.
Digital Twins Are Becoming Decision Systems for Real Estate
The useful digital twin is not a static 3D model. It is a continuously updated decision layer connecting property data, visual intelligence, operations, and simulation.
Agentic Finance Requires Consent That Machines Can Enforce
AI agents can analyze accounts, negotiate, and initiate transactions. Financial infrastructure now needs consent, authority, and limits that software can interpret and prove.
Edge Multimodal AI Will Change Property Inspection
Running visual and language intelligence near the point of inspection can improve privacy, responsiveness, and evidence quality—but only with careful model and workflow design.
AI-Native Development Will Make Software More Machine-Readable
When AI agents become active software participants, codebases, APIs, tests, and operational systems must communicate intent in forms machines can reliably interpret.
MCP: The Protocol That Will Define Agentic AI Infrastructure
Model Context Protocol is emerging as the TCP/IP of agentic AI — a standard that separates model capability from context delivery. Here's why it matters and what it changes about how we build AI systems.
RAG Is Not a Feature — It's an Architecture Decision
Retrieval-augmented generation is being sold as a plugin. It should be treated as a foundational architectural commitment — with all the tradeoffs that implies.
The MLOps Gap: Why Model Quality and System Quality Are Different Problems
High model accuracy on a benchmark tells you almost nothing about system reliability in production. The disciplines required to solve these two problems are almost entirely different.
The Hidden Cost of Retrieval: Rethinking Search Infrastructure for AI-Native Products
Vector search is not a silver bullet. Most retrieval pipelines are optimized for recall at the expense of precision and latency. A closer look at hybrid retrieval architectures that actually scale.
Fine-Tuning vs. RAG vs. Prompting: A Decision Framework
The three primary techniques for adapting foundation models to specific tasks each have distinct tradeoffs. Here's a framework for choosing the right one — and when to combine them.
Real Estate Data is Broken — And AI Can't Fix It Alone
MLS fragmentation, inconsistent schemas, and siloed portals create a data problem that no model can solve independently. What the industry needs isn't better AI — it's better infrastructure standards.
Multimodal AI in Commerce: Beyond Image Classification
Image generation, visual search, and product placement are converging into a new layer of commerce infrastructure. Exploring what's working in production and what's still marketing material.
Agentic AI in the Enterprise: Where It Works and Where It Doesn't
Autonomous agents are moving from research demos to enterprise pilots. The gap between the two is larger than the hype suggests — but the genuine use cases are more valuable than the demos imply.
Affordability as a System: Financial AI That Goes Beyond Scores
Credit scores are a proxy built for a different era. Modern financial AI can model behavioral patterns, income irregularities, and spending trajectories to surface a richer picture of readiness.
Open Banking and the Future of Financial Identity
Open banking infrastructure is quietly becoming one of the most powerful data primitives in the financial system. The question isn't whether it will reshape lending and identity — it's who builds on it first.
Building for Trust: Why AI Transparency Is Now a Product Requirement
Users don't just want accurate predictions — they want to understand why a system made a decision. Explainability is moving from a research concept to a baseline expectation in production software.