Models as Insider Risks: What Satya Nadella's Principles Mean for AI-Native Work
Satya Nadella wants frontier models treated like insider risks. Why CNEXT agrees, and how we build identity, logging and an emergency stop into agents.
Read articleGuide OpenAI · Reference
The reading order per chapter leads through the library on ChatGPT, Codex and the Agents API, searchable by product and task. Check the date before you adopt it.
As of: October 2026 · Editor: Marcel Haas
Articles
For every chapter the articles that go deeper: the core pieces first, then the deeper reads, the dated notes last. Every entry says why it is here.
Chapter 4
How does AI reach our data at OpenAI, and what does it see?
Open the chapterChapter 5
ChatGPT, Codex, Agents SDK or Agents API: which fits this task?
Open the chapterChapter 6
Where do ChatGPT, Codex and the API process our data, and what may we send there?
Open the chapterChapter 7
How does the meter run at OpenAI, and what moves the bill?
Open the chapterChapter 8
How do agents and models at OpenAI stay under our control?
Open the chapterChapter 9
Who does what, so that AI with OpenAI stays in operation?
Open the chapterGuide OpenAI
Every article on this guide, newest first. Search by title or keyword and filter by topic and year; the date sits on every card.
19 articles in this selection
Satya Nadella wants frontier models treated like insider risks. Why CNEXT agrees, and how we build identity, logging and an emergency stop into agents.
Read articleDavid George (a16z) argues that OpenAI will not win because of the best models but because it awakens new behaviour and owns distribution. What that means for Swiss organisations on Microsoft 365.
Read articleEmbed Codex as a service in your own applications: use the new OpenAI Agents API with Python and connect Microsoft Foundry through MCP.
Read articleCNEXT supports Dario Amodei’s call for a more deliberate AI pace and focuses on better harnesses, skills, tools and evaluations.
Read articleThe investigated OpenAI evaluation incident shows why AI agents need technical boundaries, isolated systems, and rehearsed incident response.
Read articleWhich Codex resets OpenAI documents, why free capacity distorts cost calculations, and the three figures teams should keep separate.
Read articleCompare OpenAI's current GPT-6 Astra, Sol, and Luna models with GPT-5.6 by workload, cost per task, and a practical model-selection test.
Read articleThe five layers of an AI Company Brain, from sources to interface, plus access control, data protection and data residency in Switzerland.
Read articleWhich AI agent suits which team? A comparison by data access, actions, approvals and operating model, with a fair 30-minute test.
Read articleHow we connected our Misty robot to Claude with Python: three prompt layers, confidence thresholds and why hard limits belong in code.
Read articleOur advice: start with Copilot Studio or Microsoft Foundry. If it has to be OpenClaw, run it only in isolation, with Entra, Defender and App Governance.
Read articleGenerator, checker, fallback: what we learned about LLM chains on our Misty robot, and how to measure whether the second step pays off.
Read articleHow I connected the Misty II robot to Azure Speech, GPT-5 and Claude Opus: architecture, noise, latency, states and data protection.
Read articleA practical roadmap for AI agents in your organisation: map the affected work, clarify responsibility, measure a pilot and involve teams.
Read articleWhat Claude's Microsoft 365 connector can do, which admin setup and Entra consent it needs, and how to test it in a controlled way.
Read articleA practical path for AI initiatives: value hypothesis, data and risk review, evaluation, human oversight and operational ownership.
Read articleMFA, Conditional Access and Purview can help protect a tenant. Available controls depend on licensing, configuration and supported services.
Read articleA clear task, bounded context, examples and criteria: eight practices for better prompts, and how to compare prompt versions with test cases.
Read articleChatGPT or Microsoft 365 Copilot: data access, permissions, data location and cost compared, plus a fair pilot with real tasks.
Read articleReference
Tell us about your current workflow, the systems involved and the result you need. In our first conversation we map your needs to the layers of the guide.