Something tells me you didn't try to deploy the switches that turn off people's appliances for them, nor did you install the equipment that transmits data from meters These are systems that require hardware, not just software
Something tells me you didn't try to deploy the switches that turn off people's appliances for them, nor did you install the equipment that transmits data from meters These are systems that require hardware, not just software
Actually this was deployed in a utility.
And how does the data get from the meters to the utility?
We used the power monitoring on the device as a sub meter
OpenADR
So the meters already had the automated reading system installed, you just wrote a script that calls the API and puts the data in the utility's DB?
No the utility meters were not involved. We stood up an OpenADR system that broadcast the prices and collected power consumption from home batteries and EVs. We also had a bidirectional V2G charger on the program. These were billed separately from the primary meter.
I think we're talking about different systems... how does the data get from the meters to you local water company?
Ah ok. Yes for the water company each meter has a Raspberry pi pico that reads the switch closure on the meter and counts gallons. It wakes up periodically and communicates with another pico that is on s meshtastic lora network. We are installing the devices and setting up the network.
Wait, you budgeted for an in-house hardware solution but wouldn't have bothered writing the data collection software without an LLM? Does this water co. serve more than, like, 25 homes?
Less. Maybe 5-10. But yes. This is me and another person doing the development and we want it to be a cheap as possible. I used the LLM to get the two devices talking to each other and to get the power management sorted. I don't have the coding chops to write the code from scratch. But i can debug
So you didn't actually need the LLM you just didn't personally know how to code it? Every time I want to code something and I don't know how I learn something which I can take into other projects. Because you used LLMs to do it for you you're no better off than you were before in terms of knowledge.
The LLM is often wrong or writes buggy code. But stepping through it with a debugger gives me a good sense of what it is supposed to do. It still takes time and effort. However I still think it is a "hit the ground running" kind of start rather than self bootstrap. It's an assistant.
I'm using the LLM to write V1 of the code which i then debug. It's in C which i know. What i don't know is the sdk for this system (raspberry pi pico). Having to make it run is how i learn. Your still have to learn it you can just learn by doing. That's the value for me.