Your biggest AI risk is probably a vendor you did not vet. Due diligence is the receipts you will wish you had. Nothing leaves your browser. Run the questionnaire: https://www.thedataexperts.us/tools/ai-vendor-due-diligence-questionnaire.html #AIGovernance #DataStrategy

The AI Trust Crisis Is Here: Frontier AI Models, AI-Governance, Why It's All Anthropic's Fault. The generator is becoming a commodity. The refinery is becoming the asset. The full receipts are in this week's piece: https://ernesttheaiguy.substack.com

Check out all the git for data tools and their relations at the usual place: the data engineering vault.

Simon Späti 🏔️@ssp.sh · 9h ago

Data Engineering Whitepapers: New GitLake Paper by Bauplan added. > A system lifts single-table Iceberg snapshots into lakehousewide commits, branches, and merges, letting agents work on isolated branches while humans review and publish changes.

Most databases have walls. Oracle Deep Data Security puts the guard inside the data itself. Row, column, cell level authorization in the engine. That is governance where it belongs. Not at the edge. Inside the core. #CyberSecurity #DataLeadership

Do not take the pitch. Read the deliverable. A real readiness snapshot, anonymized. See the sample: https://www.thedataexperts.us/ai-governance-readiness-sample/ #AIGovernance

If you're teaching a class on post-training (or part of a course) and my book, slides, code or videos don't help you, please lmk how I can improve it! Has been a ton of work getting everything done and part of the ROI is hope that it helps more education work grows around LLMs.

Big new release of my LLM CLI tool and Python library for talking to hundreds of different LLMs - reasoning traces, OpenAI Responses support, server-side tools, smarter logging and a whole lot more simonwillison.net/2026/Aug/4/n...

New release of LLM adds support for reasoning traces, OpenAI Responses, server-side tools, and smarter logging

I released LLM 0.32 this morning, the most significant new version of LLM since the initial launch of the project. The new version includes support for visible reasoning traces, server-side …

simonwillison.net

NIST AI RMF is not a PDF you skim. It is a checklist you operationalize. Nothing leaves your browser. Build your checklist free: https://www.thedataexperts.us/tools/nist-ai-rmf-checklist-generator.html #AIGovernance

The AI Trust Crisis Is Here: Frontier AI Models, AI-Governance, Why It's All Anthropic's Fault. New piece. Free tier covers the argument. Paid covers what to do with it: https://ernesttheaiguy.substack.com #GenAI #FractionalCDO

Dropbox proves AI code review is not about catching bugs. It is about surfacing security design context. MCP + Dash closes the gap between threat models and pull requests. The model is the engine. The context is the road. No road, no review. #AISecurity #DataStrategy

Most security reports are theater. This is what a real assessment looks like. See the sample deliverable: https://www.thedataexperts.us/downloads/dse-ai-security-assessment-sample.pdf

Maintaining context across multi-month deployments requires robust memory architecture. A breakdown of five design patterns shows how to manage short-term context, episodic memory, and state persistence to keep AI agents reliably on track. #AIArchitecture #LLM #MLOps

5 Architectural Patterns for Persistent Memory and State in AI Agents - MachineLearningMastery.com

In this article, you will learn the five core architectural patterns for managing persistent memory and state in AI agents, and why treating them as deliberate design decisions is essential for produc...

machinelearningmastery.com

Holly Cummins maps tech history: every hype cycle is a reissue of an old tradeoff with new branding. The free lunch is a myth. The tradeoff is the meal. Ask what debt you are accumulating, and who pays the interest. #Cloud #DataStrategy

The deadline does not move because you were busy. Track it before it tracks you. Free. No email. Free deadline tracker: https://www.thedataexperts.us/downloads/dse-finserv-regulatory-deadline-tracker-2024-2026.pdf

Deploying Model Context Protocol (MCP) in production requires security controls beyond simple gateways. A defense-in-depth model outlines four layers—from safe execution to management infrastructure—for agentic systems. #MLOps #AISecurity #LLM

Securing MCP in Production: Defense-in-Depth Beyond the Gateway

This article presents a defense-in-depth approach for securing Model Context Protocol (MCP) deployments in production. It outlines four architectural control layers: safe execution, management infrast...

infoq.com

A new ICML paper proves LLMs cannot be made secure. The flaw is in the architecture, not the alignment. For data leaders: stop asking 'is the model safe?' Start asking 'what happens when it is not?' The model is the engine. The contract is the car. #LLM #DataStrategy

Monolithic databases are the old contract. Disaggregated systems are the new engine. Compute scales up, storage stays put. That is cost efficiency, fault isolation, and the future. Ask: what is your idle compute costing you today? #Cloud #DataLeadership

Moving to a multi-agent architecture can silently triple your LLM token costs through repeated context passing and redundant tool calls. Here is a breakdown of why multi-agent setups bloat API bills and the practical architectural changes needed to fix token overhead. #MLOps #LLM #AIArchitecture

The 3× Token Bill We Didn’t See Coming | Towards Data Science

How a seemingly harmless move to a multi-agent architecture quietly tripled our LLM costs and what actually fixed it.

towardsdatascience.com

The Hidden Jobs Inside AI Work. The generator is becoming a commodity. The oracle is becoming strategic. New piece. Free tier covers the argument. Paid covers what to do with it: https://ernesttheaiguy.substack.com #GenAI #DataLeadership

A failing geothermal plant is not a dead asset. It is a data problem. Zanskar bought it, re-ran the model, found a hotter zone. Two years later: full capacity. The common belief is geology is fixed. The flip is our measurement of it is not. #AIGovernance #DataStrategy

LeetCode interviews test puzzle skills. AI solves them in seconds. Daniel Doubrovkine bombed them despite decades of leadership. The interview loop is broken. Show me a portfolio of decisions. Not a whiteboard of algorithms. That is the signal. #FutureOfWork #DataStrategy

AppSec scanners are a supply chain attack vector. Researchers show how attackers can poison scanner rules to greenlight malicious code. Treat your scanners as untrusted code. Verify their integrity before you trust their output. #AIGovernance #DataStrategy

OpenAI announces 10 discoveries from their next model. Observations:: 1) AI is getting very good at math 2) Two years ago LLMs failed at basic math 3) This cost less than $2000 in current API fees 4) OpenAI is focusing on announcing benefits, not just risks, of new models openai.com/index/ten-ad...

Data quality is not the real bottleneck. Reuse is. Qlik now offers data quality agents and reusable data products. That is how you get reliable AI. Stop moving data. Start refining it. #AIGovernance #DataStrategy

AWS Context gives agents the rules at runtime. That is governance as a pipeline, not a PDF. The model is the engine. The context is the car. Do you know which rules your agents are using right now? #GenAI #DataStrategy

The pace of progress on models from so many organizations at once is genuinely incredible. Building LLMs isn't driven by rare secrets, but consistent effort, mass capital, and effective organization design. It is great that know-how of such a powerful technology is diffused.

No LLM can be fully secured. ICML paper proves it. The model is a probability engine. Attackers just find the right context. Your defense is the governance envelope. Mitigation beats perfection. #LLM #DataLeadership

Disaggregated systems are not a feature. They are a necessity. Murat Demirbas shows how Paxos roles predicted the split. The cloud bill is the receipt. Decouple compute from storage, or pay for idle. The model is the engine. The contract is the car. #DataEngineering #DataStrategy

Everyone thinks the model is the bottleneck. It is not. The routing layer is. Microsoft just showed a three-layer architecture for AKS. Model selection. Request orchestration. GPU replica routing. That is the governance envelope. That is where cost and quality live. Build it #LLM #DataStrategy

DNS is not just plumbing. It is your governance envelope. Cloudflare now unifies public and private DNS. That collapses two silos into one control plane. Ask your network team: who audits internal DNS queries? #AIGovernance #DataLeadership