It is hardly surprising, to me at least, that Android Things is being "refocused". It was unimaginative in the extreme, misunderstood the problems of the IoT and security and was unattractive to anyone knowing Google's tendency to drop projects.
Android Things could have been a great success because the idea is a good one. Take the Android operating system, with all its UI and other facilities, and give the programmer some extras that allow interaction with sensors and effectors. The problem is that the perceived needs of security made using it very difficult to use it to build anything innovative. The security crippled the hardware.
At first Things looked promising. It was available on ARM and x86 via the range of IoT boards that Intel was promoting. Then Intel scrapped the Curie and all of the systems that made use of it. This left the Things SDK running on the NXP and the Raspberry Pi - not a great range of choice. Even so the Pi is so flexible and popular that it still could have worked.
The big problem was that the Things team interpreted security to mean that user space software shouldn't have direct access to GPIO lines and memory. This effectively made Things very slow - even flashing a few LEDs directly was a tough challenge.
An IoT programmer expects to get full access to the machine's hardware and this was not possible using Things. Security in the IoT world means making sure nasty things can't get into the user space and is all about controlling access from the outside world not the inside world. When asked if it was possible to optionally drop the tight security so that the hardware could be used creatively, the answer was that it would be unthinkable - it seemed to matter more that security was seen to be done rather than it making any sense.
Add to this the long time it took to get Things to a release, the minimal hardware support as already mentioned and Google's past history of abandoning projects then it looked like a very poor bet. Especially as Google had other IoT projects that competed with some aspects of Things.
So has Google killed it?
More or less.
"Android Things continues to be a platform for experimenting with and building smart, connected devices using the Android Things SDK on top of popular hardware like the NXP i.MX7D and Raspberry Pi 3B. System images for these boards will remain available through the Android Things console where developers can create new builds and push app updates for up to 100 devices for non-commercial use."
So play with it as much as you like, but it isn't going anywhere in the future and it isn't a commercial proposition. In my opinion this makes it dead and I'm glad I didn't waste any time on it.
The final part of the blog post is also interesting:
"We remain dedicated to providing a managed platform for IoT devices, including turnkey hardware solutions. For developers looking to commercialize IoT products in 2019, check out Cloud IoT Core for secure device connectivity at scale and the upcoming Cloud IoT Edge runtime for a suite of managed edge computing services."
Now who is going to trust that any of thie above are around in a few years? IoT has a long development cycle and a long product lifetime - Google isn't really your trustworthy partner in such an endeavor. This is a very general problem when you are looking for cloud-supported services and the solution is to make sure that you have a standalone alternative.
According to Reuters, Amazon is working on a large language model (LLM), codenamed "Olympus". With, reportedly, 2 trillion parameters, it is twice the size of OpenAI's GPT-4 model.
