Insights
Notes from the work.
Practical writing on building enterprise AI that reaches production.
Uber gives 30,000 people a general AI agent. Microsoft says narrow agents are safer. Both are right
Uber began rolling out ChatGPT Work company-wide as Microsoft reported 2,000 responsible AI reviews and argued that narrow scope is a safety feature. The difference is the layer being scaled.
Read
Training a model now happens next to your data or behind an API. What stays yours is the evaluation
Databricks made AI Runtime generally available, AWS turned training loops into SageMaker API calls, and Snowflake opened ML Jobs to custom containers. GPU training is becoming a platform service.
Read
NVIDIA's Kumo models let agents predict from your tables. The ontology they depend on still needs an owner
NVIDIA's Agent Knowledge Foundation adds an Auto-Ontology library and Kumo models for relational, tabular and time-series prediction. Agents move past document search; meaning becomes the weak point.
Read
Cohere's Rerank 5 turns a relevance score into a probability. That lets an agent know when it found nothing
Rerank 5 Pro and Fast are generally available with calibrated scores and Context Focus, which returns only the relevant passages. The useful change for enterprise RAG is a reliable way to say no.
Read
ServiceNow's Arc agent is available today. The hardware that would revoke it in a second is not yet
ServiceNow's Arc, a personal agent for email, Teams and workflows, is GA. AI Passport enforcement in NVIDIA OpenShell and BlueField comes in a future release. Govern the gap.
Read
Databricks cut AI coding unit costs by up to 90% without capping its best developers. The gateway did the work
Databricks gave 3,500 engineers a frontier model and spend rose 60% without clear gains on routine work. The fix: cheaper defaults, routing, visibility and friction, not a hard cap.
Read
OpenAI now wants to approve the AI gateways that offer its models. Find out if yours counts before November 20
OpenAI's updated Usage Policies require model routers and AI gateways that provide direct access to its models to get its approval. The rule is contractual, and its edges are not yet clear.
Read
AI avatars are now a button in the office suite. Decide whose face and voice your company may synthesize
AI avatars for training, marketing and sales video moved into everyday tools like Google Vids. The enterprise questions are consent, disclosure and impersonation, by type of avatar.
Read
Four rival AI companies gave agent infrastructure to one foundation. Check which layers are portable
Anthropic, OpenAI, Hugging Face and Alibaba gave four agent projects to the Agentic AI Foundation. Part of the agent stack is becoming standard; the part that holds your lock-in is not.
Read
OpenAI models arrive on Google Cloud. The cloud no longer picks your model, but it still sets the boundary
GPT-6 Astra, GPT-6.1 Sol and GPT-6 Luna enter Google Cloud's Model Garden in private preview, bought against existing commitments. Procurement and the data boundary change, not the model.
Read
An agent skill is a procedure. In a regulated company, procedures live under document control
Vendors now ship agents with expert-written skills and agents that learn workflows. In pharma, banking or government those are procedures, which need an owner, a version and an effective date.
Read
AWS takes CloudWatch Omni's agent observability to eight more regions. Check where each step runs
Amazon CloudWatch Omni, AWS's AI-assisted observability experience with agent tracing across LangGraph, CrewAI and others, expanded to eight regions including Frankfurt, London and São Paulo.
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