# Writing Essays and notes on governed AI in financial services by Muthukrishnan (Muthu) Bommiah. ## Context engineering - [Self-describing knowledge: why the Open Knowledge Format matters, and what enterprises must add](https://muthub.org/writing/self-describing-knowledge-open-knowledge-format.md) (Sep 15, 2026): In June, Google Cloud published the Open Knowledge Format: knowledge as plain Markdown with a little structured metadata. It is the right foundation. Regulated enterprises need the next step: knowledge that states its own owner, provenance, freshness and trust. - [Passive capture: knowledge that writes itself, with humans who approve it](https://muthub.org/writing/passive-capture.md) (Aug 18, 2026): Documentation fails because it is a second job. The knowledge we need is already produced as a by-product of work: commits, tickets, incidents, reviews and decisions. Capture it passively, distill it, and keep a human approval gate before anything becomes trusted. - [Not another data lake: context needs curation, not collection](https://muthub.org/writing/not-another-data-lake.md) (Jul 21, 2026): The tempting answer to fragmented context is to pour everything into one place. We tried that with data, and many lakes became swamps. A context layer has to be a curated overlay: retrieval-first, owned, and pointing back to the systems that hold the truth. - [Context fragmentation: why your agents keep starting from zero](https://muthub.org/writing/context-fragmentation.md) (Jun 30, 2026): The knowledge an agent needs is scattered across code, tickets, chat, wikis, incident tools and people's heads. Each fragment is accurate and useless alone. Fragmentation, more than model quality, is what makes enterprise agents brittle. - [The cost of unmanaged context](https://muthub.org/writing/the-cost-of-unmanaged-context.md) (Jun 2, 2026): Enterprises budget for compute, models and licenses, but almost none budget for the context those models depend on. Unmanaged context is the largest cost nobody tracks, and every new AI agent adds to it. ## Governed AI - [Inside an Architecture-as-Code platform: a working example](https://muthub.org/writing/inside-an-architecture-as-code-platform.md) (Jun 9, 2026): A tour of aac.muthub.org, the open prototype I built to show architecture as a control plane, and the design choices that mattered most: catalogs, executable guardrails, waivers as data, a computed scorecard and an MCP server for coding agents. - [Model is a tool, harness is the strategy](https://muthub.org/writing/model-is-a-tool-harness-is-the-strategy.md) (May 26, 2026): Handing an AI agent a production codebase without a harness is like putting a teenager in a sports car with no dual brakes. Deterministic guardrails and sensors, owned as a platform, are what make agents safe to scale. - [Architecture that runs: why AI-era governance must be code](https://muthub.org/writing/architecture-that-runs.md) (May 12, 2026): AI now writes code faster than any review board can read it, while architecture still moves at meeting speed. Governance has to become something that executes in the pipeline and in the agent's context. - [If you can't prove why an AI wrote it, you can't ship it](https://muthub.org/writing/if-you-cant-prove-why-an-ai-wrote-it.md) (Feb 24, 2026): Shipping AI-generated code without recording where it came from is unmanaged risk. Regulated teams need deterministic provenance for non-deterministic models: line-level attribution, Git-native session history and AI bills of materials. - [Designing an enterprise MCP registry: the control point agents need](https://muthub.org/writing/designing-an-enterprise-mcp-registry.md) (Sep 18, 2025): As agents multiply, every enterprise needs an authoritative directory of the tools they can call. What a governed MCP registry must do, what the official registry leaves to you, and the mistakes to avoid. - [Securing AI agents: identity, not secrets](https://muthub.org/writing/securing-ai-agents-identity-not-secrets.md) (May 29, 2025): Agents act on their own, at machine speed, across many systems. Static credentials and perimeter thinking don't hold up. Attested identity, explicit delegation and policy enforced outside the model do. - [Adopting MCP in a bank: the four risks to solve first](https://muthub.org/writing/adopting-mcp-in-a-bank.md) (Apr 22, 2025): The Model Context Protocol standardizes how agents connect to tools and data. It is silent on whether they should. Four risk areas a regulated institution should resolve before agents get tools. ## AI-native SDLC - [The AI-native SDLC: when typing code stops being the bottleneck](https://muthub.org/writing/the-ai-native-sdlc.md) (Jul 28, 2026): AI has made writing code cheap. Verification, trusted context and human judgment are now the scarce resources, and teams that redesign their lifecycle around them will pull away from teams that only add assistants. - [Three architectures for AI coding agents, and how enterprises should choose](https://muthub.org/writing/three-architectures-for-ai-coding-agents.md) (Dec 16, 2025): Platform-embedded, sandboxed and evolutionary coding agents each earn trust in a different way. Then the editor became a fleet manager. Choose and govern agents by how their work will be verified. - [Spec-driven development for regulated industries: audit the contract, not every line](https://muthub.org/writing/spec-driven-development-for-regulated-industries.md) (Oct 21, 2025): For complex work in regulated sectors, prompt-to-code cannot pass an audit. Making the specification the contract an AI must fulfil can, provided the spec stays anchored to the code and machines check conformance. ## Data & AI platforms - [Capture the why: the knowledge layer enterprise agents are missing](https://muthub.org/writing/capture-the-why.md) (Jul 14, 2026): Most enterprise AI pilots hit the same wall: systems record what happened, but the reasoning behind decisions evaporates. Durable, governed context about the why is the real foundation for useful agents. - [Agent-first data architecture: five principles](https://muthub.org/writing/agent-first-data-architecture.md) (Sep 30, 2025): Data platforms built for people asking questions will strain when the main consumer is software taking actions. Five principles for designing data platforms where agents are first-class users, and why an open lakehouse helps. - [Python or the JVM for production enterprise AI? Put the agent where the transaction lives](https://muthub.org/writing/python-or-jvm-for-enterprise-ai.md) (Jun 24, 2025): Python is unrivaled for research and prototyping. Production enterprise AI is mostly integration, security and scale, and with JVM agent frameworks now maturing, the runtime that owns the transaction is often the right home for the agent. - [From MLOps to AI engineering: what industrializing ML still teaches us](https://muthub.org/writing/from-mlops-to-ai-engineering.md) (Nov 19, 2024): A trained model is the start of the work, not the end. The lessons of industrializing machine learning carry straight into generative and agentic AI, which is why AI engineering has to unify DataOps, MLOps and DevOps. ## Leading AI change - [Leading architecture change without direct authority](https://muthub.org/writing/leading-without-authority.md) (Jul 8, 2025): In a large organization you rarely own the teams you need to move. Architects lead through clarity, credibility and standards people want to follow, and by treating the North Star and today's fires as one job. ## Notes - [From pair programming to fleet management](https://muthub.org/writing/from-pair-programming-to-fleet-management.md): Editors now run local, background and cloud agents side by side. The leadership job shifts from writing code to designing the system that writes it. - [Your prompt history is a more honest review than your calendar](https://muthub.org/writing/audit-your-year-with-ai.md): My 2025 prompt history split 60% enterprise, 40% personal growth. Here is the prompt I used to audit my own year. - [Agent design is still hard](https://muthub.org/writing/agent-design-is-still-hard.md): High-level SDKs break under real tool use. Caching, reinforcement, failure isolation and shared state decide whether an agent is reliable. - [Living documentation and the human trust gap](https://muthub.org/writing/living-documentation-trust-gap.md): Docs that regenerate on every commit describe what code does. The harder, more valuable problem is capturing why it exists. - [AI is an amplifier, not a silver bullet](https://muthub.org/writing/ai-is-an-amplifier.md): AI magnifies whatever is already in a team. Platform quality, clear policy and updated workflows decide whether that means strength or dysfunction. - [Build for today's task or tomorrow's platform?](https://muthub.org/writing/build-for-today-or-tomorrow.md): The classic architect's dilemma, upgraded for agents: how much to invest in reusable agents versus fast, single-use ones. - [Rethinking operational databases for the agent era](https://muthub.org/writing/rethinking-oltp-for-agents.md): The lakehouse unified analytics and AI. Bringing operational Postgres into it points toward one governed source of truth for agents. - [Beyond vibe coding: specs as executable artifacts](https://muthub.org/writing/spec-kit-beyond-vibe-coding.md): Ad-hoc prompting yields code that looks right but misses intent. Making the spec the gated source of truth is how AI-assisted delivery becomes repeatable. - [Closing the loop: agents that verify their own work](https://muthub.org/writing/closing-the-loop.md): An agent that runs the tests, reads the failure and fixes its own work is the line between autonomy and a one-shot tool. - [A data engineering agent in practice](https://muthub.org/writing/bigquery-data-engineering-agent.md): An experimental agent built pipelines, changed schemas, reverted its own mistakes and generated quality checks from plain language. Self-correction was the part that mattered. - [Technical debt is an architectural liability](https://muthub.org/writing/technical-debt-is-an-architectural-liability.md): Technical debt is a strategic risk rather than messy code. Coding agents make continuous paydown affordable inside everyday work. - [MCP in three primitives: tools, resources, prompts](https://muthub.org/writing/mcp-tools-resources-prompts.md): MCP ended the copy-paste problem with three simple primitives. Being open is why it beat a thousand bespoke integrations. - [Remote MCP, straight from the API](https://muthub.org/writing/mcp-connector-messages-api.md): Direct remote MCP access through the Messages API removed a client layer and moved agent integration from clever hacks toward engineering. - [Code that validates itself: AI-generated assertions](https://muthub.org/writing/ai-generated-assertions.md): Only about 2.5% of open-source Java repositories use production assertions. LLMs can write them, adding the verification layer AI-written code needs. - [A2A and MCP: complementary, not competing](https://muthub.org/writing/a2a-and-mcp.md): MCP connects agents to tools; Agent2Agent connects agents to each other. Together they turn single agents into systems, and raise the governance stakes. - [Governance belongs in the catalog](https://muthub.org/writing/unified-governance-in-the-catalog.md): When business meaning, data products and relationships live in one open catalog, agents stop guessing at tables and governance travels with the data. - [The MCP spec grows up](https://muthub.org/writing/the-mcp-spec-grows-up.md): OAuth 2.1 authorization, streamable HTTP, batching and tool annotations: the March 2025 update that made enterprise MCP plausible. - [Simplicity is not simplistic](https://muthub.org/writing/simplicity-in-architecture.md): Designs that are easy to explain, build and run take the most effort up front, and they are the ones still working years later. ## From the archive (Medium) - [Introducing the AI Agent Canvas](https://medium.com/@muthukrishnanb/introducing-the-ai-agent-canvas-45418db4aa3c) (2025-05-27): A one-page blueprint for designing and documenting AI agents across the software lifecycle, with a worked example. - [Databricks, Delta Lake, and the Rise of the True Data Lakehouse](https://medium.com/@muthukrishnanb/databricks-delta-lake-and-the-rise-of-the-true-data-lakehouse-a-blueprint-for-modern-data-f7c4d56c231e) (2025-02-11): A blueprint for unifying the data lake and the warehouse for analytics and AI. - [A Deep Dive into BigQuery Managed Iceberg Tables](https://medium.com/@muthukrishnanb/unlocking-your-data-lakes-potential-a-deep-dive-into-bigquery-managed-iceberg-tables-73ddc585e277) (2025-02-11): Managing large data lakes on open table formats with BigQuery. - [Simplicity in Architecture](https://medium.com/@muthukrishnanb/simplicity-in-architecture-emphasizes-71f40e4870a6) (2025-02-11): Why designs that are easy to explain, build and maintain win over time. - [The Architect's View: Taming Wild Intelligence with Enterprise-Grade Generative Systems](https://medium.com/@muthukrishnanb/generative-ai-3e1fe4eb20b1) (2024-05-13): Building generative AI systems that meet enterprise standards. - [The Unrivaled Engine of Big Data: Its Architecture and Adoption](https://medium.com/@muthukrishnanb/new-article-3b2e1fc9b153) (2023-03-24): How a big data processing engine works and why organizations adopted it. - [Demystifying AWS Data Transfer Costs: A Guide for Cloud Architects](https://medium.com/@muthukrishnanb/demystifying-aws-data-transfer-costs-a-guide-for-cloud-architects-e1c753060c83) (2022-01-01): How AWS data transfer pricing is structured and how to design around it. - [The Rise of MLOps: Industrializing Machine Learning](https://medium.com/@muthukrishnanb/the-rise-of-mlops-industrializing-machine-learning-bd9b13d9c098) (2021-10-21): Why a trained model is the beginning, not the end, and how MLOps closes the gap to production. - [The Accidental Revolution: How Containers Quietly Conquered the Datacenter](https://medium.com/@muthukrishnanb/containers-3712d348a036) (2021-10-21): How containers spread through the data center, largely without fanfare. --- Muthukrishnan (Muthu) Bommiah · https://muthub.org · Views are my own and do not represent my employer.