|
|
|
no jealousy from ChatGPT
Wednesday, October 8 2025
At work today, my torture at the hands of Crystal Reports continued. Today I was having trouble figuring out where large grids of data were coming from in a report, and ChatGPT turned me on to the idea of subreports. It seems one Crystal Report can be embedded inside another to form a turducken of a report! Of course, to find this, one must poke around in Crystal Report's appalling interface, as the names of these embedded reports are not even searchable in the binary blob of its file format. The more I work with Crystal Reports, the more of a quagmire of technical debt it seems to be.
On the drive home from work, I stopped for much-needed provisions at MyTown Marketplace in Stone Ridge. Among the things I bought were a paper bag containing four sesame-seed bagels, crimini mushrooms, a cauliflower, tempeh, an overstuffed bag of very dark coffee, thick spaghetti, and lots of white beans (as Gretchen recently had to use pinto beans in place of white beans, as our stockpiles had been completely depleted).
It being Wednesday, after I took Charlotte for her late afternoon walk, I made spaghetti for dinner. I also cooked up some bok choy that had been languishing in the refrigerator as a weird Asian side dish. (Even though I boiled it in a shallow pan of bouillon, used no Asian spices and in fact used sauerkraut juice instead of lemon juice, it somehow still tasted Asian.)
Once that was done, I extracted the rotten plank from the east deck and substituted in the new one I'd bought yesterday. But I fastened it down with only two screws, since I wasn't happy with any of the fasteners I could find.
After dinner, I returned to the project of getting my Lilygo T_deck to join the Meshtastic channel that spans from the Catskills to the Adirondacks (admitted as two tiny islands separated by 100 miles). I was having ChatGPT help me, but ChatGPT is not a good assistant for a technology like Meshtastic that seems to have undergone wild changes from one version to the next. The newer version I was using couldn't make any sense of the musty old command-line instructions ChatGPT was suggesting. I'd then tell ChatGPT that the command it had just suggested does not work and that it should never give me that command again, but infuriatingly it would give me that command again and again (I checked, and it was identical to the one I had just complained about). When it was clear ChatGPT couldn't help me, I tried using Google's AI (Gemini, I think). I pasted in a huge dump of a good node's configuration and then a bad node's configuration and asked Gemini why the nodes with those configurations couldn't communicate. At that point Gemini evidently decided it was no longer a brilliant assistant but was instead an enshittified search engine and proceeded to dump an ad-rich collection of "also on the web" style links to my screen. "Oh," I said sarcastically, "you're just going to give me links!" That seemed to snap Gemini out of that mode and it immediately compared the two configurations and came back with a verdict. And the information it gave me was even actionable: the version of Meshtastic I was using with the bad node, it said, was older than the version I was using with the good node, and they were incompatible. (One was like 2.4 and the other was like 2.7, which doesn't seem like a big numeric difference, but evidently that is huge in the fast-changing Meshtastic universe.)
So I reflashed the Lilygo T_deck with the 2.7 firmware and damned if it didn't start communicating reliably. Because I had some lingering animosity at ChatGPT for its failure, I told it that I'd turned to another large language model to solve my problem, perhaps hoping it would fly into a jealous rage. But no, it remained ever cheerful and supportive.
For linking purposes this article's URL is:
http://asecular.com/blog.php?251008 feedback
previous | next |
