What's your most interesting automation?

I wrote a scraper for various dating websites that would automatically generate activity on my behalf according to several empirical criteria (very simple stuff, no ML!), so that I could minimize the amount of time spent manually browsing profiles, and dealing with being ignored 99% of the times.

It worked amazingly well, it got me hundreds (yes, 100+) of dates, incredible dating stories and a lifetime of memories. As a matter of fact, my wife (of 7 years!) was found through the software, and she is the most important person in my life. She knows about the scripts and laughs about it.

I would just let the script send the initial messages with some basic filtering and scraping, and I would take over the conversation only when I got an initial positive reaction (which, as you can imagine, was a tiny fraction of the outreach) and manually confirmed that the profile was potentially interesting to me. The script would present the relevant information (pictures, links, etc) in an ideal way to be consumed in bulk instead of having to click through a million profiles. It really saved a lot of time.

I have several Home Assistant Automations. The most useful one is blinks my indoor lights and sends a push notification to my phone when my garage door opens and closes.

I also have automations for sending a notification when my chest freezer goes above 10 degrees.

In the winter, I have an automation that turns my humidifier on/off to keep the indoor humidity within certain ranges depending on the outdoor temperature, to avoid excess condensation on my failing windows.

At an early job all of my offical job duties came down to data collection and transcription.

If a certain router started timing out, I had to go reboot it. I placed it on a networked power switch so if the router timed out 5 times in a row, the power switch bounced, rebooting it.

I had to go collect temperatures from the outside thermometers around the small 3 building campus. Replaced those with net thermometers with APIs that could be hit to get a reading.

Every 30 minutes we had to record information from some netsend ping and check some jobs. I had to first engineer a solution to get the net send messaging working on Win7. Then I had to extract the information from it, transform it into a usable data structure and use a perl script to load that into our mysql journal DB.

There were data management tasks that required clicking precise locations in an application. I used tools to make sure that the window opened in the same place then I learned how to accomplish precise cursor location and manipulation with either batch or wscript, depending on the needs at the time. Then I nailed down a precise timing for each action in the tasks and virtualized the entire setup into a VM.

Then I used the host machine to refactor our Java 5 application to Java 7 with considerations for Java 8.

But that wasn't "my job" and when the owner died and his son inherited the business he hired his buddy to manage me out. When his buddy found out that I'd "automated my entire job", I lost my job. (Never mind the fact that the job had become refactoring Java code, writing blog posts for the website, and maintaining said website.)

I was getting paid $9/hr.

You won't ever catch me automating anything anymore.

I do lots of tiny automations (ahk, js bookmarklets). E.g. when to scrape the web for images (training loras), I wrote a set of scripts that save an image bound to a single button. From rclick-s---enter and similar to bookmarklets that recover hires images from specific sites.

Another example is “1234 rename” script that does (F2 ctrl-v - enter right) 4 times where n is 1,2,3,4. This way I name 4 random generation examples.

Recently I automated some govt site with literally pages of inputs and combo boxes and non-existent API docs to input a couple hundreds of cards into it.

Also semi-automated lora dataset preparation which makes it a lot easier to handle. I can collect and prepare a new dataset while a previous lora is in training, very productive compared to default kohya_ss experience.

I had a bunch of Puppeteer scripts to pull my financial info into a spreadsheet.

Figuring out how to enter my pin automatically in my banking website was a surprisingly fun challenge: https://www.ing.com.au/securebanking/

My boss at Allied Gear Company had programs his dad wrote in 1980 using TK!Solver under MS-DOS. I thought I was going to have to redo them in something modern, but I found an easier path with VirtualBox and some port rerouting to get the printouts captured into a text file in notepad.

He was happy and it kept everything untouched, thus I wasn't going to induce bugs, a very important thing in a job shop with repeat products that only show up every few years, but represent the life blood of the business.

Interesting to see spreadsheet in some answers.

We used Pabbly automations on our Google Sheets to send WhatsApp messages to candidates, fetch candidates confirmation for scheduled interview in Sheets directly, etc.

Not sure if an automation, but a European marathon has an official resale platform which is first come first serve, which in practice means resale tickets are snatched up almost instantly.

I built a cron job which checked the page every minute, and sent a push notification to my phone when the page changed.

I probably could have optimised it further, but this was sufficient for me to secure a ticket.

Excel Spreadsheet containing Fibre Optic network node placements and connections (wavelengths) for entire country. Group of 8 of us working on project. Major piece of work, I'm brand new to industry and enjoy coding

--->>>

Wrtie py script to scrape the spreadsheet with Python (20 sheets, 1000s of rows and columns)

--->>>

Calculate optimal networking conditions and connections/allocations for 1 entire full country network using Djikstras algo and some other custom code. Solves a major allocation issue that can't really be solved manually, very painstaking.

--->>>

1 hour later, Report results to bosses, solve entire task, ask for 2 pay-rises, leave the next week for double my salary

lol

Won a hackathon where we wrote test cases into a spreadsheet in plain human language. The tool would make a ticket in Jira, write the test, run the test to make sure it fails (because the feature hasn't been implemented), link it with Buddy, makes a pull request in GitHub, and notifies project owners on Slack.

The backbone of this was Pipedream, Jira, and Buddy, and GPT-4 handled the test writing. This was a year ago when GPT-4 was the top of the line AI model. Yes, we needed all of these, it was a fragile thing built on external tools.

I scrap tens of job boards with Selenium, JavaScript and Python, and generate random links to the companies' job offers to contact them directly.