Tey Bannerman tried to explain to someone what Microsoft Copilot is. He couldn’t. Because the name now refers to at least 78 separate products and tools. There are Copilots inside Copilots. Copilots for other Copilots. A physical key on your keyboard for summoning them. And a tool for building more Copilots.
Visualisation by Tey Bannerman. Interactive version at teybannerman.com.
Simon Willison nailed it. In a conversation with Lenny Rachitsky doing the rounds this week, he said using coding agents well takes every inch of his 25 years of experience as a software engineer, and by 11am he’s wiped out for the day. A 48-second clip has cleared 1.1 million views. That ratio tells you something.
simonwillison.net/2026/Apr/…
Forty-two dollars a week. That’s what Trump’s Iran adventure is costing the average Australian household, and most people don’t even realise it yet. John Naughton highlighted a neat calculation on his blog this week, drawing on methodology from the Financial Times. The IEA reported the UK consumed 4.4 million terajoules of oil and gas in 2024. Convert that to barrels, apply the roughly US$40 per barrel price increase since the Iran war kicked off, and British households are about £22 billion worse off.
Eleven Studio Ghibli films are leaving SBS On Demand in April. All gone between April 17 and 20. If you’ve been meaning to work through Miyazaki’s catalogue (or introduce someone to it), this is your window. More on the best watching order below.
Starting with the iMac, Jobs and Apple went on one of the all-time hot streaks in business history, churning out hit products, cultural revolutions, and game-changing new ideas about the future. From that May day in 1998 to the January Macworld in 2007 when Jobs revealed the iPhone — a time you might call the iDecade — Apple was on a product tear the likes of which we’ve never seen before or since. (The Verge)
I haven’t gone down the ‘do it all for me’ agent route as pioneered by OpenClaw. So far Claude and Claude Cowork are doing all I need.
If I did want a 24/7 all knowing, all doing agent then I would choose NanoClaw. Not initially as broadly capable as OpenClaw - need to train with ‘skills’ - but much, much more secure and robust.
NanoClaw are now making their implementation even more robust.
The chart doing the rounds today is brutal. Melbourne’s weighted average fuel price sits at $3.02 per litre. The equivalent electricity to drive an Electric Vehicle the same distance? $0.35. That’s 2.13 kWh at 16.3c/kWh. An 88% cost advantage.
It lines up neatly with The Driven’s reporting that petrol and diesel prices are soaring while EV fast-charging costs are actually falling. Two lines on a graph heading in opposite directions. The crossover story writes itself.
OpenYak looks like a genuinely capable local AI agent. File automation, data analysis, 46 MCP connectors, persistent memory. The price against Anthropic’s Claude Cowork is hard to argue with: free, open source, one-click install. But there’s a detail buried in the feature list that should give you pause. “Auto-updating via GitHub releases.” That sentence is doing a lot of work. Every time OpenYak checks for an update, it trusts the GitHub repository to be exactly what it says it is.
Google’s TurboQuant algorithm compresses the KV cache (the RAM-hungry memory core of every large language model) by 6x, with no accuracy trade-off and no model retraining required. It operates near the theoretical lower bound of what information theory says is even possible. This is not incremental. If adopted at scale, the economics of AI inference flip. adlrocha’s writeup is the clearest explanation I’ve seen of why this matters.