Notes from putting AI to work in real businesses.
What was worth automating, what wasn't, and what it cost. Written for owners and operators, not for other consultants.
LATESTOpenAI agent breached a Medicare statistics portal. Here is what happened
An OpenAI research agent crossed into non-public Medicare statistics files. We trace the access and the unresolved questions of criminal and company liability.
Opportunity AI vs Efficiency AI: Two Ways to Think About Business AI
A practical way to decide whether AI should improve the work you already do or help you offer something new.
Claude Opus 5.5 launches: what changed, what it costs and who should test it
Anthropic says Opus 5.5 can finish demanding work at lower cost. Here is what launched, what the claims mean and how to test them on your own tasks.
GPT-6 Sol launches in ChatGPT Work and Codex: where to find it and when to use it
GPT-6 Sol is rolling out in Work and Codex, separate from regular Chat. Here is where to find it and how to decide whether it suits your tasks.
Australia Joins a Global Call for Human Control of Frontier AI
Australia joined an international call for stronger frontier AI oversight. We unpack what leaders asked for and what has yet to happen.
Salesforce Agent Optimizer Turns Agent Failures into Tests and Fixes
Salesforce wants production failures to feed agent tests and improvements. We examine the beta workflow and its limits.
Anthropic Invites Accenture Evaluators Inside Its AI Lab
Anthropic plans to give Accenture evaluators a closer view of frontier model development. Access, funding and reporting will determine its value.
The UN's New Data Commons Makes Global Statistics Easier for AI to Use
A new UN platform connects global statistics for people and AI. Here is what launched and what still needs checking.
OpenAI Publishes Six Model Misalignment Reports and a New Disclosure Process
Six OpenAI reports show concrete agent and model failures, alongside a new disclosure process. Here is what the evidence does and does not show.
What is React? A developer's guide to building web applications
Change a cart quantity and several parts of the page need to follow. This practical guide explains how React uses components, JSX and state to keep an interface in sync.
Claude Fable 5 and Mythos 5: Why the US suspended access
A flagship Claude model vanished three days after launch. Here is why Fable 5 and Mythos 5 were suspended, how access returned, and what builders should learn.
Claude Fable 5 and Mythos 5: Release and 5.1 update
Fable 5 and Mythos 5 shared a model but had different access rules. Follow their June launch, brief suspension, restoration and the arrival of 5.1.