Writing and daily intelligence

Insights

Two things, published on a schedule. A daily brief on what actually shipped, and longer articles written for boards and the people who report to them.

Articles

Written for South African boards and the people who report to them. Specific, dated, and signed by the person who wrote them.

Article
Published 26 August 2026

People Will Take the Bot. They Cannot Find the Door.

A company that sells AI customer service agents published the research showing people yell and swear to escape one. The complaint is not the machine. It is the missing exit.

Read the article →

Article
Published 26 August 2026

The 45 Jobs That Were Not Redundant

A bank cut 45 customer service roles because a voice bot had reduced call volumes. The union said volumes were rising. Nobody has established the technology underperformed at all.

Read the article →

Article
Published 26 August 2026

Nobody Published the Denominator

A deflection benchmark quoted across the industry traces to no primary source. It looks entirely reasonable, which is exactly what makes it dangerous.

Read the article →

All articles →

The Daily AI Brief

Every morning: what shipped, what it changes, and what it means for an organisation running AI in production. Compiled by an AI system from the sources each edition lists, dated, and labelled as machine-written on every edition.

It is compiled for us first, which is why it is specific. Model releases, pricing moves, regulatory changes and the security disclosures that turn into controls. Follow the source links before relying on any item.

Every edition, dated and archived →

Latest edition
Published 3 November 2026

Anthropic launched Claude Opus 5.5, establishing a new frontier benchmark for autonomous agentic coding, native computer use, and enterprise knowledge execution.

  • Anthropic detailed Claude Sonnet 4.6 featuring a beta 1M-token context window and upgraded planning capabilities, alongside a significantly tightened 2026 Usage Policy.
  • OpenAI deployed new developer frameworks for GPT-6 Astra, prioritising skill engineering, long-horizon Codex execution, and resilient agent SDK lifecycle maintenance.
  • Google DeepMind unveiled Gemini 4 Argon for complex multi-step reasoning, supported by targeted rollouts of Gemini 3.8 Flash, Flash Cyber, and EmbeddingGemma 2.

Read the brief →

The writing shows the method

The same discipline that produces these articles produces the assessments. Read a few and decide whether it is the kind you want applied to your estate.

Book the two-hour diagnostic