Just had a small eureka moment today and i was wondering if someone was already doing this. Putting a simple structure in place for ai to simply version what it changes and keeping the older data or pieece of work Does that sound like token wastage or overkill I have had experiences where ai assumed i gave it permission and made unapproved changes which took some dilly dallying to get sorted out. Whats been your experience

I’m working on an automation system for real estate offices, and I’d love honest feedback from anyone who has worked in this field. It’s built for the inside of the office, a shared brain for the team: - A clean, organized database for all units, with different permissions for different people, so everyone can work and communicate more easily - Lead matching: when a client asks for a unit the office doesn’t have, the lead is saved. Every time a new unit enters the system, it’s checked against those saved leads, so employees don’t forget them - A lead cycle: when a unit is rented, the contract details go into the system. As the contract nears its end, the system reminds the employee to call the tenant. If the tenant isn’t renewing, that’s a new lead. The owner is also contacted, which means new data and a new unit for the office. Then it starts again I’m still learning this field, so I’d really like to hear from people who have worked in real estate: - What problems do you face day to day that I may be missing? - Does this cycle match how it actually works, or is something off? Any honest criticism is welcome. Thanks!

So i did mention in the past how claude was going green on all its feauters and still i was experiencing disappointment with the features built as they still failed to perform as needed Well i got it to create a 'breakit' test for every new feature as minute as a button. Here is the prompt below - see if its any good Thank me later I want you to build and run a "break-it" check for the feature we just built. What a break-it check is: a script that deliberately breaks the code one rule at a time, runs the tests, and checks that at least one test fails. If a test fails, that break is CAUGHT (the tests protect that rule). If every test still passes, that break is MISSED (the tests have a gap). Steps: 1. List the rules. Read the feature's code and write a plain-English list of every rule it enforces. For example: who is allowed in, what is refused, what is checked before saving, limits, expiry times, what must never be shown. Show me the list before you continue. 2. Write one Python script, breakit_<feature>.py, saved outside the project folder (for example on my Desktop). For each rule it must: - make a copy of the original file first, - change that one rule in the code (remove a check, flip a condition, change a number, skip a permission test), - run only the tests for this feature (for example: php artisan test tests/Feature/<FeatureName>Test.php), - record CAUGHT if any test fails, or MISSED if all tests pass, - ALWAYS restore the original file afterwards, even if something crashes (use try/finally), - never leave a change behind. Print a clear table at the end: rule, CAUGHT or MISSED, plus a total count. 3. Before running it, run the normal tests once and confirm they all pass. If they don't, stop and tell me. 4. Run the script. Then confirm the code is back to normal by running git status and git diff on the changed files (they should show nothing). 5. For each MISSED result, tell me in plain English what it means. Then either write a new test that catches it, or explain why it is genuinely redundant (for example, the framework already checks it). Run the script again.

I have an opencode go subscription that i use to supplement my claude code. To be honest it does a better job often than claude when it comes to setting up my morning coffee drive. Today out of curiosity i used both claude and big pickle to set up my morning and big pickle picked more stuff for my itinery more than claude. Why, i did not know. Back to my moan, i have been using my freemium serve on opencode go for about almost a quarter of the day only to see this amazingly laughable notification asking me to come back in 10hours to contimue using the freely available agents.. "Free usage exceeded, subscribe to go [retrying in 9h 49m attempt 1] I guess that is my cue to hang my laptop and go out to skate. I think sincerely i have overused the available space today Exceeding a free usage is hillarious, off to skate and get my outdoor life back if the UK weather will let me

Like i said in my last post. Claude code is horrible at writing test cases and pretty much will pass its own tests In my case most of the features still mess up in production A resolution was me asking another subagent to write tests not only to validate the requirements but to also try to break the features. It took a darn loooong time for that to be executed but in the long run i had less broken features and the test wasnt outcomes driven but requirements driven. I wanted to share the prompt claude gave me but its missing in the mire. Its a game changer for me and i have mildly converted it to a global skill Testing must be done properly, Claude