Insights
Notes from the work.
Practical writing on building enterprise AI that reaches production.
A typed answer can still be wrong. Snowflake and Cloudflare build fast models for decisions and extraction
Snowflake's Decision Fast and Cloudflare's Clairic return categories, scores and values, not prose. They remove parsing failures and cut latency. They do not make a confidence number trustworthy.
Read
Databricks wants agents to remember what they learned yesterday. The hard problem is what they should forget
Databricks launched Agent Memory in beta, built on its MemCurrent research. A curated memory beat the full history while using far fewer tokens. The open questions are staleness, scope and deletion.
Read
AI found two critical PHP flaws before attackers did. Your patch process now sets the defender's window
PHP fixed two critical remote code execution flaws, found with help from OpenAI and Wiz models, on October 21. The fix is in the PHP runtime, not your plugins, and the window is already closing.
Read
OpenAI's COBOL migration benchmark is close to saturated. That does not mean banks can retire the mainframe
OpenAI reports 97.9% on its own COBOL migration evaluation and big gains in COBOL, Fortran and Visual Basic. The useful part is how it defines success: byte-for-byte behavior, not code that compiles.
Read
Microsoft wants AI agents treated as insiders. Its answer is permission that changes while they work
Microsoft extended Project Perception to the agentic insider: Guardian Agents, task-scoped Dynamic Permissions, a live agent risk score and Conditional Access that can isolate a running agent.
Read
Your company has dozens of model deployments. Can anyone say which models, whose they are and what they cost?
Microsoft published a reference architecture for a tenant-wide view of Foundry and Azure OpenAI deployments: inventory, usage, cost and quota, with no prompts collected. Ownership tags come first.
Read
An agent that acts for your company needs two permissions: yours, and the other side's
Agent programs design the permission an employee grants and ignore the one the counterparty must grant. Research and recent news show both sides fail, and in different ways.
Read
An agent can reason perfectly over half a transaction. Databricks makes multi-table writes all or nothing
Databricks made multi-statement transactions GA across Unity Catalog managed Delta Lake and Iceberg tables. For AI agents that read and write the lakehouse, it is a consistency control.
Read
An Anthropic sales rep runs six agents instead of one. They share a memory, not their keys
An Anthropic business development rep in Japan described his fleet of Claude agents for lead routing, qualification, email and CRM. The design choices transfer better than the hours.
Read
Google changed the model behind "stable" in Agent Search. If you never chose one, your answers changed too
On October 16, Agent Search's stable answer-generation model moved from Gemini 2.5 Flash to Gemini 3.5 Flash. Stable is also the default. Why a moving alias belongs on the change calendar.
Read
Banks, airlines and retailers are turning AI agents away. Your company is on both sides of that door
The Verge reports JPMorgan Chase, Charles Schwab, Delta and others restricting AI agents. A warning for agents built on browser automation, and a policy question for your own site.
Read
AWS's agent now rewrites Azure .NET code for AWS. Swapping the SDK calls is the easy part of a cloud move
AWS Transform made Azure to AWS SDK conversion GA for .NET: Functions to Lambda, Blob Storage to S3, Queue Storage to SQS; Cosmos DB to DynamoDB in preview. What it does not convert.
Read
Have an AI use case stuck between prototype and production?
Tell us what you’re trying to ship. We’ll reply with honest next steps.
Discuss a use case