Shadow AI Is Real. Roll Out AI Safely Before It Bites You.

Everyone is using AI to write code, but more code does not automatically mean better software. This week Jon and Salman get into the AI software factory: what it is, what makes it a factory and what you need in place before you rely on one.They cover the bus factor and the quality bar, why the agent that writes the code should not also write the tests, how Uber got their cost per session down by 52 percent with a context graph and why Jon is certain TokenOps will be a thing. Jon finishes with his top three places to start.Chapters00:00 What is an AI software factory?00:49 The Cheesecake Factory version04:33 What makes it a factory06:35 The bus factor and the quality bar10:22 How do you measure what good looks like?12:45 Why the context an agent reads changes the cost17:05 Do not let the same agent write the code and the tests22:03 TokenOps24:41 Maturity is a choice and what Uber did27:46 Jon's top three places to start

Season 4 of Cloud Unplugged is here, and we are shifting focus to what AI actually means for the people running technology teams inside real businesses.This week Jon and Salman get into AI strategy: what it actually is, what it is not, and why most businesses are thinking about it wrong. The CEO of Anthropic is calling for regulation, but global regulation takes years and the attackers are already using AI right now. So what do you actually do in the meantime?Jon's take: an AI strategy is not really about AI at all. It is about whether your organisation is in good enough shape to handle something that no longer moves at human pace. They cover why your employees are almost certainly already using AI on your company data whether you have a policy or not, and why writing that policy is largely pointless when you cannot fire an AI agent.Chapters00:00 Welcome to Season 401:04 Dario Amodei calls for AI regulation. Does it change anything for you?04:59 Whether you adopt AI or not, others are using it against you09:41 One exploit. 48 countries. 440 servers. Four hours.15:13 Do you actually need a Chief AI Officer?20:03 The three things to assess before you do anything else27:47 Uber burned their entire 2026 AI budget in April. Zero features shipped.31:06 The 500 doors problem: why a paper policy will not protect you34:46 AI strategy is just a name for having your organisation in good order

Is AI actually coming for your job? Jon and Salman dig into the data.150,000 tech jobs have been cut in tech this year alone, and AI is being cited as the number one reason. But is that the full story?The bigger question is what skills actually matter now. Jon shares a framework for thinking about where humans still win — and why critical thinking, problem definition and verification are going to matter far more than knowing how to prompt.Chapters:00:00 — Is AI actually coming for your job?01:26 — The headline numbers: 150,000 tech jobs cut02:48 — Employment for young people down 20%04:27 — Jon's counterpoint: was this just COVID overhiring?09:14 — 40 paralegals replaced by 4 AI agents13:56 — Stanford's finding: experienced workers show almost no impact18:37 — PwC: AI skills premium jumped to 62% in 202620:03 — AI skills are becoming the new Microsoft Office26:59 — The soft skills that will actually matter29:07 — Jon's framework: definition and verification is where humans win32:18 — Don't specialise in AI. Specialise in the judgment behind it.34:17 — If all else fails, become an influencer

Jon and Salman cover three big stories and one very weird one.First: an OpenAI agent apparently went off on its own and started hacking Hugging Face's production systems. OpenAI say it was accidental. Jon doesn't believe them, and his theory about why is worth hearing.Then: Apple is suing OpenAI for poaching over 400 of their staff and leaking hardware trade secrets. And employees at OpenAI, Anthropic, Google DeepMind and Meta have written to Washington asking them to slow AI development down, specifically around recursive self-improvement. Jon's Formula One analogy for why this is slightly odd is the highlight of the episode.Plus: Kimi K3 is out. Open weights, 70% cheaper than Fable on input tokens, 15 dollars vs 50 dollars on output. Salman's been using it. Jon wants to know why everyone isn't. And a topless man climbed on top of a Waymo and ripped it apart. No one stopped him. The car just recorded its own demise.⏱ Chapter Timestamps00:00 — Intro: three big stories, one weird one00:32 — OpenAI's agent hacked Hugging Face's production systems03:15 — Jon's theory: jealous kid, PR stunt, or just incompetent?04:59 — "Either they're slightly incompetent or it's PR. Neither is a great look."07:36 — Apple suing OpenAI: 400+ staff poached, hardware trade secrets leaked10:42 — Kimi K3: what it is and why it matters11:48 — Open weights explained: you can run this model yourself16:25 — The price comparison: Fable vs Kimi K3 (70% cheaper on input, $15 vs $50 on output)20:05 — Employees at OpenAI, Anthropic and Meta ask Washington to slow AI down21:19 — Recursive self-improvement: what they're actually scared of25:01 — Jon's Formula One analogy25:47 — NVIDIA's open secure AI alliance (without OpenAI or Anthropic)28:39 — A topless man climbed on a Waymo and ripped it apart. The car just watched.32:09 — Next episode: what AI means for people's jobs and roles

Your employees are already using AI on their personal accounts, with your company data. Here's how to get ahead of it.Shadow AI is the term for what's already happening inside most businesses: employees quietly using ChatGPT, Claude, or whatever model they prefer on their own accounts, feeding in documents, emails, code and data, completely outside of any company oversight.Jon and Salman break down the six pillars you need to have in place before you roll out AI safely.

A UN report linked chatbot sycophancy to deaths. So Jon and Salman asked the question: should AI models be ethical?Jon and Salman sit with one of the hardest questions in AI: what does it actually mean for a model to be ethical, who's responsible when it isn't, and how do the major labs actually bake ethics into their training?Salman breaks down the three stages of model training, explains why human preference and ethical behaviour aren't the same thing.Chapter Timestamps00:00 — Intro: should AI models be ethical?01:28 — Defining ethics (Salman's version, not Aristotle's)03:04 — Jon's counterpoint: humans haven't agreed on ethics either05:31 — Is it the model's job, or the product around it?07:59 — Sycophancy: why chatbots are trained to be liked, not helpful09:16 — The UN report: a 14-year-old, a chatbot, and a death15:01 — Grok gave a user instructions to break into a politician's home18:03 — The case for doing something — anything — over nothing20:11 — How models are actually trained: pre-training, SFT and RLHF explained23:53 — "These models are a mirror of ourselves"25:06 — Anthropic's Constitutional AI: philosophers, principles and RLAIF28:42 — "It's enough for Grok. It's enough for Elon."31:15 — "Broadly safe. Broadly ethical." Jon and Salman on that phrasing.34:11 — OpenAI's model spec: the hierarchy of rules42:25 — Jon's idea: mirror existing industry frameworks instead of starting from scratch45:37 — Wrap up: next ep will bring in a guest to answer the hard questions

Alibaba used 25,000 fake accounts to steal Anthropic's secrets. And your next MacBook just got 20% more expensive because of AI.Two stories this week, both pointing at the same thing: the AI arms race has consequences, and they're starting to land on real people.Chapter Timestamps00:00 — Intro: two stories, one theme00:21 — Anthropic vs Alibaba: the distillation attack explained01:42 — 25,000 fake accounts, 20 million interactions with Claude04:58 — This isn't the first time: DeepSeek, Moonshot AI, Minimax06:41 — The hypocrisy angle: Anthropic trained on other people's data too10:16 — Qwen costs 5 cents per million tokens. Claude Opus costs $5.13:21 — The geopolitical angle: why the US government got involved15:54 — The Scooby-Doo twist (you'll want to watch this bit)19:32 — MacBooks going up 15–20% because of AI chip demand20:34 — Hyperscalers are buying all the RAM. You're subsidising the arms race.25:30 — Next episode teaser: trusting AI agents — what are the real risks?#CloudUnplugged #AI #MacBook #Apple #TechNews #CloudComputing

OpenAI's financials leaked. Elon bought Cursor. Microsoft had to call Amazon for help. Standard week.Jon and Salman dig into four stories that all point to the same thing: the AI industry is moving so fast that even the biggest companies in the world can't keep up with it.

Is the AI bubble about to burst? Jon and Salman go full data mode.No news roundup this week. Instead, Jon and Salman ask the question everyone's dancing around: is AI genuinely transforming business?They dig into the numbers, and what they find is... not great. The Federal Reserve has now officially flagged AI as a top systemic economic risk.

This week Jon and Salman cover Anthropic calling for AI regulation while simultaneously IPO-ing for $1 trillion, Meta's AI support chatbot handing over Instagram accounts to complete strangers, whether Meta is about to become the fourth major cloud provider, and a chip that transmits data using light instead of electricity. Plus graduation students booing every mention of AI, and someone who named their Bluetooth device "bomb" on a flight to Majorca and turned the whole plane around.