This PET story most interesting detail is the cassette drive that had to be evicted to make room for a proper keyboard. Originally it was made a virtue of packing everything into one box, including a screen and storage. Then actual owners encountered those calculator-style keys. The machine had arrived ready to run, but its everyday interface needed changing. That's an unusually clear example of physical packaging becoming an architectural constraint. A programmer's most-used component, the keyboard, had been shaped by Commodore's history of making calculators and by the limited space inside the chassis. Once the keys changed, the storage design changed with them. It's hard to call a design decision merely cosmetic when fixing it requires moving an entire storage device. The PET made computing approachable partly by removing the assembly work its predecessors left to enthusiasts. Putting the entire experience in one enclosure also embedded a collection of assumptions about how people would use it. Sometimes the least glamorous component of a machine determines whether the rest of its engineering brilliance is worth anything.

GeoWorks could make a modest 286 feel like a respectable graphical computer, yet creating software for it initially required equipment vastly more expensive than its target hardware. That's a peculiar way to democratise computing: invite everyone into the shop, then make potential suppliers buy the building. The usual retelling of graphical desktop history gives Windows the advantage of momentum and Microsoft's established position. Smith includes both, but the development requirement points to something more structural. For a desktop platform, every awkward little user requirement is a potential application. An accessible development environment lets someone build it, test it and perhaps sell it. With GeoWorks, that experiment began with a substantial capital investment. Many ideas would never reach the stage of being rejected by actual users. AOL's involvement makes the contrast sharper. Its distribution could put GeoWorks technology in front of modem owners, but visibility alone wasn't going to produce a thriving software catalogue. Microsoft, meanwhile, was making Windows programming approachable through Visual Basic. GeoWorks had made running graphical applications remarkably inexpensive while making their creation unnecessarily difficult. GeoWorks was impressive engineering. Unfortunately, its development requirements meant that much of the ingenuity needed to make it successful would have had to come from people who couldn't afford to participate.

Right, this grocery example is where this stops feeling like a slightly cleverer search box... A household's regular shop is full of untidy constraints: favourite brands, acceptable substitutions, delivery slots, allergies, how ripe the bananas should be. McKinsey suggests an agent that knows household history could take much of that friction away. That's totally plausible and far more useful than having a chatbot describe ten varieties of washing-up liquid.... But - and it's a big one! - household shopping is also a running series of tiny decisions; meaning, that handing them to software and some decisions become invisible. A substitution that saves a pound, a familiar brand favoured again, a new product never surfaced: none looks dramatic in isolation. Across fifty weekly baskets, these defaults could become a significant influence on what a family buys. The claimed convenience might be real, while the customer gradually loses sight of the alternatives. That's a demanding definition of agency, and a better one than counting how many shopping conversations happened in ChatGPT.

Right, temperature surfing is an excellent name for an irritating problem! With the Rancilio Silvia, Steve had to guess when the boiler was at the right temperature to pull a shot. Make milk first, forget to purge the residual steam, and the next espresso could be ruined. By the time he bought the Rocket R58, he wasn't searching for a button that would make coffee for him. He wanted the machine to stop springing surprises. That's a curious sort of progress.... Plenty of appliances improve by hiding the process from the user. Steve went in the opposite direction: a more elaborate machine, a more serious grinder, fresher beans and greater attention to technique. Yet the extra engineering made the parts he enjoyed more repeatable. He could concentrate on learning rather than wondering whether a wandering boiler temperature had just sabotaged the experiment. An expensive way to protect a hobby from its equipment, perhaps, but a coherent one. Steve's journey suggest me that the best upgrade is the one that lets you taste the difference your own decisions made...

Sky Tracker's four Xenon beams rotate through a cloverleaf pattern, driven by a cam that would be absurdly unhelpful to an anti-aircraft crew... that detail captures the searchlight's long career better than any dramatic night-sky photograph. Sure, military engineers spent decades making the beam point at a particular object, with listening horns, synchros, binocular controls and eventually radar feeding the aim, and the advertising machine deliberately throws that precision away. Its purpose is to keep moving until someone looks up. What's striking is how little of the visible object had to change for its value to invert. A searchlight made its own position obvious, which was dangerous when bombers could attack the crew; remote operation became a survival requirement. Once military surplus became available at bargain prices, that conspicuousness became the selling point for a car dealership. The same brilliant beam that once announced a target could announce a sale. The transformation happened through economics as much as design: thousands of expensive military machines suddenly became cheap enough to hire for an evening. This history also explains why the visual cliché survived its technical purpose; radar displaced the searchlight because it could locate aircraft without the spectacle, while advertisers wanted the spectacle itself. Brilliantly inefficient, by design.