About a year ago OpenVPN updated their app on IOS. I didn't realize it would cause me an issue but the update borked my VPN access through my phone to my Netgear router. I have searched many times trying to find a solution. You can't roll back apps in IOS, not easily anyway. So about a month ago I decided to ask Grok. It suggested some changes to my config that didn't work and asked me to post a screenshot of the errors I was getting. After posting the screenshot it told me exactly what I needed to change in my config and I was back and running in no time.
Today I was about to ask a question here about an issue that has been bugging me for awhile. I have a few IOS automations on my Iphone main screen. They were using cloud endpoints to execute HE rules. The problem was sometimes when running these automations previous automations would run. Somebody would send me something that opened in Safari and it would open my shop. My automations were using Safari to open the endpoint URL, and Safari would resend the URL if I opened safari again for something else. I had went in and added a close tab to each automation and that did seem to help but it would still do it from time to time.
Grok said it was a common problem and showed me a technique I was not aware of that eliminates opening a browser at all. Issue solved.
I wrote an app about 5 years ago that would keep a single switch on out a selected set of switches. It was my first app, and probably about 100 lines of code and took me a few hours to write and debug.
I needed a similar app today and used Gemini to create one for me. Only about 40 lines of code. And took all of 2 minutes.
I started programming in the mid-70s. It is safe to say that all my programming languages are dead. Well, except those poor souls tasked with keeping .gov services running. But I still retain a bit of program logic. Using Gemini (free) 'we' have written a couple of fairly complex Tampermonkey projects that are significantly above my pay grade.
I understand that LLMs are very much an existential threat for entry-level programmers, artists, photographers, etc but they are howling at the moon. It's here.
Microsoft AI CEO Mustafa Suleyman says AI will reach "human-level performance" in white-collar work.
He predicts most tasks in that field can be automated within the next 12 to 18 months.
Several leaders in the AI industry have warned of impending mass job replacement.
I'll admit I have not made use of AI in any meaningful way in my day-to-day job, or tbh, elsewhere.... at least not knowingly. And this response is largely from a perspective of delivering "data solutions" in a business context, rather than writing a driver or app on HE....
IMO, we (developers) can still offer a level of oversight and coordination in areas that may change significantly as a result of some of the advancements in AI. I feel that customers of the services we deliver, at least those internal to a business, would still appreciate the human interaction with those responsible for delivering the systems they rely on. If you start to move into external customers there can be a level of separation that can lend itself to more technology-driven solutions such as those offered by AI, but it is very much a case-by-case basis. I know I appreciated the personal touch of a person contacting me off the back of an automated process at my bank, allowing for a conversation and a chance to understand what lead to the call.
For all that I seen, AI can assist and make things quicker or easier, much like HE aims to do around the home, but AI is not a full replacement for the human interaction and involvement in significant pieces of work.
I'm beginning to find AI useful in day to day stuff. A few tasks I managed to accomplish:
1 - Managed to get ChatGPT to create a fully functional driver for the Sonoff Orb button. There was some back and forth to fix various issues but the resulting driver works well.
2 - It fixed dinner for me. I typed "I have chicken breast, peppers, onions, feta cheese (and other basics) - make me tray bake recipe that takes no more than 15 minutes prep time". The recipe looked good so I asked it to format it into a pdf for printing. Result - it was delicious and I'll be having it again.
3 - While at the finale of the World Suberbike championship in Jerez, Spain in November; we bumped in to Alex Lowe in the paddock and my wife wanted a photo with him - he blinked! So I dropped the image into the ChatGPT window, explained the issue - the image was edited perfectly.
I use ChatGPT and GitHub copilot and the combo gives me super human coding capabilities in updating my smart home apps, including fixing bugs in my Groovy code. If you do any amount of coding at all these tools will make you feel like a coding ninja.
I bought an early-bird, one-time, no subscription, lifetime (theirs) business license to an AI aggregator service - 1min.ai so I can choose the LLM that is better at images, or one better at coding, etc. You can also combine 2 or 3 LLMs and point the 'team' at a task. I was not using Gemini, ChatGPT4 & 5 were better, I thought. When Gemini updated in Nov 2025, that changed. It was now noticeably better at script coding than anything I'd used. Then the LLMs started expanding the sessions in their free tiers. Gemini made a competitive move and now has 3 free levels (but you can rotate through them to extend your session). So I quit spending fairy tokens in the paid service, letting them accumulate until the Free Tier Wars end. I did three 14-15 hr sessions (yeah, I binge code -old habit) on Gemini is a span of a week.
Real handy. Gemini helped me configure NUT on my raspberry pi to shut itself down gracefully when its ups is on battery and to send a shutdown command to hubitat through port 8080.
