Twelve months ago the most capable model you could rent was Claude Opus 4, at $15 in and $75 out per million tokens. Today the flagship Opus costs a third of that: $5 in, $25 out. So the price of frontier AI collapsed over the year, right?
Not even slightly. It went up. You just have to know where to look, and who is paying.
Microsoft’s annual developer conference, Build, kicked off at 3am Melbourne time on Wednesday. I didn’t stay up to watch - but I’ve absorbed the media releases and technical docs, and there’s a genuine shift happening here that’s worth unpacking.
TL;DR for the non-technical: AI assistants are about to get much more capable, but that creates a trust problem - how do you let a smart assistant do things on your computer without giving it the keys to everything? Microsoft just announced that Windows itself will act as the security guard. It will control exactly what an AI assistant can see and touch on your machine, track what it does separately from what you do and run smaller AI models directly on your computer so your data doesn’t have to leave your desk. Think of it as giving your AI assistant its own office with its own keycard, instead of letting it wander freely through yours. The catch: it needs newer, more powerful hardware to work properly, and most of it isn’t shipping yet.
Now, the details.
The Tifosi are furious, and they are aiming at the wrong target. The Ferrari Luce, Maranello’s first production electric car, was unveiled near Rome on 25 May to a wave of revulsion. “A Nissan Leaf with a prancing horse,” said the internet. Nissan, delighted, leaned in and thanked Ferrari for the compliment. The market was less amused. Ferrari shares fell more than eight per cent in Milan and over five per cent in New York the next morning. For a marque that has sold beauty for almost eighty years, that is a remarkable thing to do with a single reveal.
But the styling is not the story. The styling is a symptom.
A Chinese carmaker just out-engineered the people who invented the prestige limousine.
The Nio ET9 is the proof. Watch what its suspension does to a broken road, then watch a BMW or an Audi attempt the same. One glides. The others crash about like farm equipment.
Nio pitches the ET9 as an executive flagship and the cabin earns it: four seats, reclining rear thrones, a fridge and screens everywhere. Nio reckons the ride is “comparable to cruising in a business jet”. On this evidence that is not the usual launch-day hyperbole.
Last time it was the cables lying to you. You binned the mystery leads, bought the certified ones with the speed and watts printed on the side, labelled the survivors. Good. You fixed the drawer.
Now look at the laptop itself, because it is about to play the same trick on you. Two ports, same oval socket, same confident silver moulding. One is fast. One is not. And nothing on the outside tells you which is which.
Open the drawer where you keep your cables. Go on. Somewhere in that tangle are two USB-C cables that look identical, came in similar boxes and feel the same in your hand. One will charge your laptop at full speed and drive a 4K monitor. The other can barely run a mouse.
The connector tells you nothing. That is the whole problem.
The work I’ve been doing between Claude Cowork and Excel lately has been considerably more involved than what follows. But this example makes the pattern clear - and it captures exactly why the combination saves so much time and frustration.
I had a list of 20 ASX stocks in an Excel file. Two columns: ticker code and company name. I wanted two more: market cap and 12-month return, pulled from most recent data.
Singapore’s Foreign Minister assembled a personal AI agent on a Raspberry Pi 5 with 8GB of RAM. He hasn’t dared switch it off in three months. He is not an engineer.
That’s the story. But the interesting part is what he learned building it - and why he built it at all.
Dr Vivian Balakrishnan gave a 22-minute talk at AI Engineer Singapore on 16 May. He described himself as a practitioner with a day job - “a retired eye surgeon who took a detour into politics, perhaps for too long.” The talk is worth watching in full. His framing of what AI agents are actually useful for cuts through more noise than most conference keynotes manage.
I wanted to know if my AI subscription was earning its keep. So I repriced 30 days of real usage at pay-per-token rates and compared it to what I actually pay.
The answer: $96 USD equivalent in tokens consumed. My subscription costs $100 USD a month. That’s not a rounding error - that’s a subscription running at near-full utilisation.
Not a prototype. Not a weekend experiment. A system I actually depend on - morning briefings, task management, research, document work, health tracking, portfolio analysis. The Autonomi, as I call it, runs continuously and does real work.
Benedict Evans just dropped the Spring 2026 edition of his annual AI presentation and it is, as usual, the clearest thinking on the subject you’ll find in a single sitting. The deck is 79 slides. It is not a TED talk. It does not tell you AI will save humanity or that the robots are coming for your job by Thursday. What it does is something rarer: it maps what we actually know, what we can reasonably infer and where the honest answer is still “we don’t know yet.