Insights
Notes from the work.
Practical writing on building enterprise AI that reaches production.
Local or cloud is no longer the question. Windows and NVIDIA turn model placement into a routing policy
Microsoft is building Windows around hybrid intelligence and NVIDIA is putting large local models on desks. The decision that matters is who writes the rules for routing each task.
Read
An agent cannot be its own security authority. Microsoft Execution Containers move the boundary into Windows
Microsoft made Execution Containers generally available: a policy layer, enforced by the operating system, that decides which files, networks and processes an AI agent can touch.
Read
In regulated AI, the model never grades its own work. Proof has to come from outside it
Dual Run, verification tools and calibrated reviewers share one rule: evidence that an AI output is right must come from something independent of the model. A test for regulated teams.
Read
Your agent is slow, and it may not be the model. Agents put a new kind of load on the data path
Google now pitches storage and network instances for agents that query vector stores and operational databases. The point is broader: agent latency and reliability live in the data path.
Read
Reflection's Beam is a 501B open-weight model. With open weights, the operating duty moves to you
Reflection announced Beam, a 501B-parameter mixture-of-experts model for coding and agents under Apache 2.0. Weights arrive later this month, after final red-teaming. Plan for what you inherit.
Read
Agents are becoming authors of production changes. Your change process is now your agent governance
ServiceNow, Google and AWS now ship agents that write workflow, infrastructure and code changes. The control that governs them already exists: change management, with a few new fields.
Read
SAS AI Navigator starts governance with an inventory. Mapping a use case to a rule is not compliance
SAS's governance SaaS, now on Microsoft Marketplace with six months of free access, inventories AI use cases, models and agents and maps them to policy packs from Protiviti and Asenion.
Read
Google built an agent to move your Kubernetes off AWS. Migration agents make the exit plan real
Google's EKS-to-GKE agentic migration, in Public Preview, translates manifests, storage and networking behind approval gates. Cheaper switching changes negotiations, and who you trust.
Read
Cohere North 2 puts agent budgets next to agent permissions. Production needs both
North 2 adds memory, reusable skills and an admin layer with roles, model restrictions, token quotas and spend alerts, deployable down to air-gapped sites. PwC will help deliver it.
Read
Google lets Gemini read the mainframe. Dual Run, not the model, proves the replacement works
Google Cloud Modernize pairs Gemini-based code analysis with Dual Run, which replays live transactions on old and new systems. Intesa Sanpaolo uses it to build evidence for regulators.
Read
OpenAI starts watermarking text for the EU. A watermark is evidence, not a verdict
OpenAI's textGrain is opt-in for API customers worldwide and coming to ChatGPT and Codex in the EU. OpenAI's own tests show why detection cannot decide authorship on its own.
Read
ServiceNow wants AI to find, build and ship workflow fixes in one loop. Decide where people still sign off
AI Workflow Factory, generally available now, links Process Mining, Build Agent, App Engine and AI Control Tower. Autonomous Engineer, the unattended coder inside it, is Early Access only.
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