Journal · field notes

The Brief.

The Brief publishes expert contributors writing from real deployments, real decisions, and real failures. For operators. Not for the timeline.

Filter ›
Recent · 06 posts

Lately.

Showing 1–7 of 15
Why most AI pilots fail at month three — and the four habits that save them.

Why most AI pilots fail at month three — and the four habits that save them.

A pilot that "works in the demo" is the easy part. Pilots die in the boring middle: when the data drifts, the champion gets pulled, and nobody owns the eval. Here is what separates pilots that survive month three from those that don't.

RAG isn't search: a primer for operators.

RAG isn't search: a primer for operators.

Most teams confuse retrieval with search. The difference is why your first agent felt clever and your second one feels confidently wrong.

The data lake question every CFO should ask first.

The data lake question every CFO should ask first.

Before you buy another SaaS dashboard, answer this one question. If you can't, you don't have a data problem — you have a procurement problem.

Five workflows that pay for themselves in 90 days.

Five workflows that pay for themselves in 90 days.

We've stopped pitching transformation. These five wedges fit on a Post-it, ship in a quarter, and convince finance to fund the next thing.

On choosing your first AI partner.

On choosing your first AI partner.

Six questions to ask before you sign anything. Half of them have nothing to do with models, frameworks, or benchmarks.

Anatomy of a quoting automation rollout: what eleven weeks actually looks like.

Anatomy of a quoting automation rollout: what eleven weeks actually looks like.

A composite walkthrough of how quoting automation unfolds in a mid-sized firm — the eval that saves you twice, the change management nobody budgets, and the numbers that matter more than the headline.

Archive

Earlier.

8 posts
The Brief · weekly

AI Weekly
Digest.

Key features and updates for Anthropic, OpenAI, Gemini, and CoPilot.

Latest issue

Codex CLI 0.147.0, Opus 4.1 retired, Atlas shuts down

31 July – 7 August 2026

No new model this week, but three substantial enterprise and infrastructure stories. Anthropic launched inference hooks — server-side inline DLP for Claude Enterprise — and opened the public beta of self-hosted environments for Claude Code, letting enterprise teams run agent sessions on their own compute. Anthropic confirmed it is building an in-house silicon team. The Fable 5 subscription structure permanently settled on July 20; the $100 credit claim window closed August 2. Opus 4.1 retired August 5. Workbench and experimental prompt APIs retire August 17. Claude Code shipped archive plugins for air-gapped environments and a range of security and reliability fixes. OpenAI released Codex CLI 0.147.0 with portable Agent Plugins, MCP 2026-07-28 support, and a breaking removal of --full-auto. ChatGPT for PowerPoint moved to token-based credit pricing. Atlas shuts down August 9. o3 retires from ChatGPT on August 26. Google shipped inline context-filtering hooks for Vertex AI and self-hosted CLI runners for AI Studio. Microsoft Copilot mirrors both with inference hooks and self-hosted environments in public beta.

Read latest issue →
Aug 07, 2026
Past issues

Archive.

20 issues