Readit NewsBut I don't know that I accept the core assertion. If the engineer is screening the output and using the LLM to generate tests, chances are pretty good it's not going to be worse than human-generated tech debt. If there's more accumulated, it's because there's more output in general.
I assume the code that checks for tap vs start-of-swipe is to blame. I have no idea why that would cause word recognition and/or autocorrect to work so differently, but it seemingly does.
I assume it's something to do with distinguishing swipes from taps with both active, but it really is a marked difference.
It looks the same but behaves differently enough that I have a hard time believing it shares code. When I turn off swipe, my tap accuracy goes MASSIVELY up, and a lot of the autocorrect screwiness seems to abate considerably. I can go back to blind thumb typing.
That said, swipe is so useful, I’ve left it on, and I deal with the degraded tap behavior. But maybe that’s a trade-off for you to consider.
If the experiment had been based on the idea that that option isolated the question, it may have been flawed. I found my ChatGPT’s o3’s accuracy went way down when I cleared personalization and deleted all past chats (turning off extended memory would’ve been equivalent, I think).
Importantly, only once did the o3 reasoning mention it was fishing from my past chats—that’s what clued me in I messed up the isolation—but the guess rate was still radically different from all the times before once I cleaned house. That suggests to me that it was quietly looking before, and it just didn’t make the cut for explicitly saying so.
When I cleared personalization data and turned off extended memory it quit being nearly so accurate.
Trust me, if you met me you would not be suggesting I am PC - but this really jumped off the page at me. Such a strange example for this context!
Think about it as a decomposition where "dancer" means "dancing person."
In the simple case, both "blonde" and "dancing" separately modify "person." If you diagram that it’s a Y: either modifier could be removed without changing the meaning of the other, and their order isn’t important.
In the complex case, "bad" modifies "dancing," which together modify "person." That’s an ordered chain, which is more complex to build and comprehend. Your clarification illustrates the chaining and why it’d be a fundamentally different meaning if that wasn’t understood.
I’m not even touching whatever you were going for with the blonde/brunette thing. It's plain they used the example because there's no possible way hair color could be a modifier for "dancing," and they wanted something unambiguous.
iOS, iPadOS, and macOS have a pretty great built-in background-noise generator these days. While lots of actual beaches can go dead silent and then have a loud wave crash in, the waves that
It’s available in Settings -> Accessibility -> Audio & Visual -> Background Sounds. You’ll have to download the sounds each once, but after that they stay on your device.
Digging this deeply in Settings isn’t pleasant if you just want some white noise, so you may want to add a control to Control Center like “Background Sounds” (way down in the Hearing Accessibility section) to turn the ocean noise on and off.
I turn this on my iPad when going to bed if I want to take extra steps to ensure that I don’t wake up in the middle of the night.
I do use a standalone Lectrofan for sleep as I prefer my noise machine to be across the room and Alexa-controlled (via a smart switch), plus it’s louder and the brown noise is “browner.”
But I keep iOS BG sound mapped to the triple-click shortcut for when noise-cancelling just isn’t enough in loud restaurants etc. It works great with AirPods for reducing my noise sensitivity issues.
You could take the exact same documents, prompts, and whatever other bullshit, run it on the exact same agent backed by the exact same model, and get different results every single time. Just like you can roll dice the exact same way on the exact same table and you'll get two totally different results. People are doing their best to constrain that behavior by layering stuff on top, but the foundational tech is flawed (or at least ill suited for this use case).
That's not to say that AI isn't helpful. It certainly is. But when you are basically begging your tools to please do what you want with magic incantations, we've lost the fucking plot somewhere.
And even a human engineer might not solve a problem the same way twice in a row, based on changes in recent inspirations or tech obsessions. What's the difference, as long as it passes review and does the job?