Rui CarmoThings have been rather slow, as befits both the time of year and the warm weather.
This year there has been less tinkering with heat pumps, but a couple of decade-old ones had to be replaced and I have been considering adding them to HomeKit as well.
But, overall, I have spent most of my free time unwinding and reading, with most of my copious amount of projects put on hold due to odd work hours, lack of inspiration and a creeping discomfort with the way the industry is “maturing”.
As an example, consider one of the books I finished recently: Losing The Signal, a pretty comprehensive account of the rise and downfall of Research In Motion and the Blackberry–a device I used myself and was even a product manager for, to the extent that I experienced first hand the effects of many of the events in the book, including persuading my then CMO at Vodafone Portugal to only order 200 units of the BlackBerry Storm (the least amount we could, since we were forced by Group to launch it) because it was obvious it was going to be a failure.
I really should write about those months in detail some day, but suffice it to say that the book is a pretty good account of the company’s history and the people behind it, and I found it a refreshing read given the overall circumstances.
It isn’t perfect, but harks back to a time where there was the same kind of excitement and optimism about technology that I felt when I started my career, and that is sorely lacking these days. But, more to the point, as I read it I couldn’t help but think that the same kind of hubris and lack of focus that led to RIM’s downfall is still rampant in the industry as a whole–you just don’t see it because it’s obscured by more layers of hype, and the scale of most companies is so much larger that it takes a lot longer for the effects to be felt.
Besides that, I’ve been puttering about with Large Language Models. I have been reticent to do so until now because of the hype, exceedingly strong distaste for chatbots and lack of “real world” applications that have a good effort/returns ratio, but trying to put together a simple workflow engine for my own use has been a good excuse to get my hands dirty with things like LangChain.
However, I’m frankly appalled at the overall quality of the code I’ve seen so far–in short, the current LLM tooling landscape is a mess, and I’m not sure if it’s because of the hype, the lack of a proper software engineering culture in the field or just a tendency of to overuse abstractions and frameworks until they’re nested six deep.
Putting that aside for the moment, I’m looking forward to taking an extended break and re-think a few things…