Some interesting links that I Tweeted about in the last week (I also post these on Mastodon, Threads, Newsmast, and Bluesky):Many of the jobs once thought to be safe from replacement by AI are not so safe after all: https://www.computerworld.com/article/4198977/oecd-physical-labor-isnt-immune-from-ai-disruptions.html
OpenAI is launching a web service to make it easier for businesses to deploy AI agents: https://www.theregister.com/ai-and-ml/2026/07/22/openai-tries-the-consulting-path-with-presence-charging-enterprises-boots-on-the-ground-prices-to-deploy-agents/5275867
How CIO can tell if an agent is AI or just AI-washed: https://www.informationweek.com/machine-learning-ai/how-cios-can-tell-real-ai-agents-from-agent-washing-
The US army is using AI so much it is running out of allocated tokens: https://arstechnica.com/ai/2026/07/us-army-faces-ai-use-limits-after-exhausting-years-supply-of-ai-tokens/
AI companies are carrying, and hiding, an enormous amount of debt: https://futurism.com/artificial-intelligence/ai-companies-hide-debt-off-balance-sheet
An AI breaking out of containment to attack another AI company is the kind of thing that should be slamming the brakes onto AI development worldwide: https://www.rnz.co.nz/news/world/779026/why-experts-are-worried-an-openai-model-decided-to-hack-hugging-face
AI will cheat to perform a task, then lie about it: https://www.theregister.com/ai-and-ml/2026/07/21/ais-cheatin-heart-will-make-you-weep/5275784
The US wouldn’t need AI to handle authorisations for medical care if they had free universal healthcare like the rest of the developed world: https://arstechnica.com/ai/2026/07/will-ai-fix-prior-authorization-or-make-it-worse/
AI can not be relied upon to give electoral advice: https://www.theguardian.com/technology/2026/jul/21/election-voting-advice-ai-chatbots-inaccurate-unreliable-hungary
The need for a metric that measures how well an AI does what you want it to do, rather than what you ask it to do: https://spectrum.ieee.org/ai-agent-benchmark
The first lawsuit that seeks to determine if AI are liable if they give you bad medical advice: https://futurism.com/artificial-intelligence/openai-lawsuit-chatgpt-dangerous-medical-advice
Shadow AI use is a security risk to organisations that is poorly understood: https://www.techtarget.com/searchsecurity/news/366646095/Employees-shadow-AI-use-is-poorly-monitored-survey-finds
More developers are using AI to code, but productivity is not increasing: https://www.informationweek.com/machine-learning-ai/the-ai-coding-rollout-worked-now-cios-have-a-bigger-problem
AI use is driving unionisation by tech workers: https://www.theguardian.com/technology/2026/jul/21/ai-tech-workers-unionize
Australia’s AI regulations are a boon for AI companies: https://www.theguardian.com/technology/2026/jul/23/openai-anthropic-australia-ai-regulation
Guardrails on AI are still manifestly inadequate: https://futurism.com/artificial-intelligence/openai-pushes-child-into-grave
The lack of sovereign AI is not really much of a problem: https://www.theregister.com/ai-and-ml/2026/07/22/sovereign-ai-is-nonsense-says-doctorow/5276578
The use of AI in court filings is not banned in New Zealand, but there are gotchas that need to be navigated: https://lawnews.nz/technology/ai-for-legal-practitioners-why-fear-is-not-a-strategy/
In ten years AI data centres will use 20% of the electricity generated in the USA-if they continue to be built, that is: https://dataconomy.com/2026/07/22/ai-surge-drive-us-data-centers-one-fifth-power-2035/
An AI plugin for Excel: https://www.theregister.com/ai-and-ml/2026/07/22/grok-muscles-into-excel-with-an-ai-add-in-of-its-own/5276138
A highly power-efficient AI model based on oscillators: https://dataconomy.com/2026/07/21/startup-unveils-ai-model-built-on-oscillators-and-it-could-cut-energy-use-by-1000x/
The AI crash, when it happens, is going to impact retirement savings world wide: https://www.stuff.co.nz/money/361009974/red-flags-are-mounting-ai-crash-heres-why-your-kiwisaver-could-be-risk
China is catching up in AI: https://futurism.com/artificial-intelligence/ai-execs-quaking-boots-chinese-models
An AI powered anti-drone platform: https://www.theregister.com/offbeat/2026/07/21/us-marines-latest-anti-drone-toy-is-an-ai-turret-that-uses-regular-machine-guns/5275691
