Our first phase of embedded assessments
- Four independent assessors gain deep access to model checkpoints, source code, and training data
- Assessments focus on alignment and safeguards during training and internal use
Daily
10 selected stories for 2026-11-13, grouped by topic.
Dates use UTC+8. The latest edition uses a day with at least 5 completed reports when available; until then an earlier edition may appear. There is no fixed delivery time.
Estimated briefing time 4 min
You have reached the end of this edition.
Your reading listOpenAI launches ChatGPT Wallet, the first wallet that lets AI agents complete purchases without access to users' real card details, requiring a security check for every transaction and preserving existing card rewards and dispute protections. It debuts for US ChatGPT Pro subscribers through dots, built with American Express, Mastercard, and Visa.
Agent skills capture repeatable engineering work in files teams can review and reuse, while tools, project context, and permissions remain specific to the host that runs them.
Natera built SWARM, an agentic AI system on Amazon Bedrock AgentCore that captures institutional lab-operations knowledge and deploys it at scale, delivering about 130x faster first response and 96.2% automated coverage across support tickets.
Constellation, the largest nuclear fleet operator in the U.S., is building a governed Enterprise Data Platform on Databricks to create shared data definitions and enable AI use cases in nuclear engineering, outage planning, and sales.
Comparing per-token prices hides the real cost of running a model on Amazon Bedrock. This post shows how to calculate the true cost per task across four hidden cost drivers, reduce it with caching, routing, and output constraints, and prove agent ROI with human baselines, holdout experiments, and a five-dimension scorecard.
Databricks announces the general availability of Genie Migrate for Lakehouse, an AI coding agent that converts proprietary SQL from seven data warehouses—Microsoft SQL Server, Azure Synapse, Teradata, Oracle, Snowflake, Amazon Redshift, and Google BigQuery—into open ANSI SQL with over 80% accuracy (90%+ on benchmarks), helping customers like MIXI accelerate migrations.
Amazon Bedrock AgentCore runtime gets major upgrades: sessions start up to 15x faster, scale to 4x more compute, run up to 68% more efficiently, extend isolation to data storage, and introduce usage-based billing.
H6.ai explains on the Databricks blog how its accounts payable automation platform uses Lakebase Postgres database branching to give each customer an isolated database and replay real invoice cases against frozen snapshots, making AI-driven financial automation both fast and auditable.
Databricks introduces Lakehouse//RT, a real-time analytics engine built to handle AI-agent-driven query volumes on the lakehouse without proportional cost increases, with customers reporting 6x to 10x lower infrastructure costs.
A migration guide in the amazon-bedrock skill helps coding agents move customer-managed Bedrock knowledge bases to Managed Knowledge Base, using read-only discovery, side-by-side resource creation, and pre-cutover validation.
Shield's Forge platform uses OpenAI's GPT-6 Astra to let AI agents navigate legacy portals lacking APIs, freeing technicians to focus on issues that require judgment.
Amplitude's proactive product agent Wave, built primarily on OpenAI models, identifies what to build, helps teams ship changes with coding agents, and measures impact—cutting product improvement cycles from about a month to 1–2 days.
This post shows how to build an OpenAI-compatible, self-service LLM gateway using Amazon Bedrock, LiteLLM on AWS Fargate, and supporting AWS services, giving developers instant model access while platform teams gain per-user cost controls, quota isolation, and observability.