A genuinely new class of AI arrives, the way the best people use it shifts, and the bill keeps coming due

The big event was the launch of a model that is a real step up, not a small one, and that changes what you can hand off to AI. Underneath it, the way the most effective people work with these tools is shifting fast, and the cost of using them is becoming a real budget line.

• A new top-tier AI model is a real step change, not a small upgrade
  My take: Anthropic released Fable 5, and for the first time in a while the jump is big enough to matter for how you work, not just for benchmark charts. The people pushing it hardest describe handing it a large, messy goal and walking away while it works for hours, then coming back to something close to finished. One company reported it did a code change across a 50 million line system in a day that would have taken a team two months by hand. The practical shift is that you can now delegate bigger, vaguer pieces of work and do less babysitting. The new skill is what one commentator called 'task imagination': getting better at picking ambitious things worth handing off, because the limit is now your ideas, not the tool.
• The most capable model comes with serious strings for business use
  My take: Before you put company data into this new model, read the fine print. For now, prompts and outputs are kept for 30 days and can be reviewed by people, on every platform that offers it. If you have the memory feature on, it can pull your past chats into new ones. Together that is a real problem for anything under an NDA or covering sensitive client work. There are also heavy filters: questions touching biology, chemistry, or cutting-edge AI research get quietly routed to a different, older model. I expect the data retention rule to loosen over time because enterprises will not accept it, but right now treat the newest model as fine for general work and off-limits for confidential material until that changes.
• AI pricing is now firmly pay-for-what-you-use, and the new model proves it
  My take: The flagship model is being pulled from flat subscription plans within weeks, after which using it means paying per use. This is the clearest sign yet that the all-you-can-eat era is over and the price reflects how much you actually consume. The more capable models also tend to chew through far more behind the scenes, so the bill is driven by the task, not just the sticker price per use. The move for your business is to stop defaulting every request to the most powerful model. Match the model to the job: use the cheap, fast option for routine work and save the expensive one for the genuinely hard problems where it pays for itself.
• The way the best people use AI is changing, and the gap is widening
  My take: The frontier has moved from giving AI single tasks to giving it standing responsibilities. Instead of asking it to fix one problem, advanced users now set up 'loops' that keep an AI running against an ongoing goal, like watching every error report and fixing issues as they appear. The catch is that almost no one has been taught how to do this, so the gap between the handful of people getting huge value and everyone still using AI like a search box is growing fast. The takeaway for leaders: your own skill with these tools is the single biggest factor in whether your team adopts them well. If you do not have a real, hands-on training effort for your people yet, you are now behind, and the cost of staying behind is rising.
• Stop sending files, start building small websites
  My take: A pattern worth copying: more knowledge work is moving from documents you send to simple websites you share by link. The reason is practical. A link is always the latest version, works on any device with no download, can be organized so different readers jump to what they need, and can show you what people actually read. New tools now let you turn a document or analysis into a shareable web page or small app with little effort. Think a sales proposal a client can interact with, a live project dashboard, or a handbook that is always current instead of a stale PDF. Pick one thing you regularly send as an attachment and try building it as a site instead. It is becoming a basic work skill, like making a slide deck.
• OpenAI declares a 'third phase' as both big labs march toward the stock market
  My take: OpenAI published a plan saying AI has entered a new phase where the economy reshapes around it, released on the same day it filed to go public. Anthropic has filed too. Strip away the grand language and the useful signal is that these are becoming large, durable, public companies you can plan a multi-year strategy around, not fragile startups. There is also a quieter split worth watching: the casual chatbot most people use and the heavy 'do real work for me' tools are starting to look like two different products with very different value. For your business, the money and the real leverage are firmly on the work side, so that is where to focus your attention and your budget.