As much as I dislike the buzz changes, I think calling blue buzz totally useless is an overstatement. There's so many more fun things you can do with text to image that are still possible with blue buzz. Here are some of my favorites.
Generate characters in the style of a different work
When these generations work, I've always found them good fun. Tonal opposite, thematic similarities, rivals (e.g., Marvel characters in DC styles, Final Fantasy in the style of Akira Toriyama), and "could have been"s (Curious what your favorite character would look like if they did get in the big crossover? Want to see what a new character would look like in a work's classic style?) are all great fun, though "just because" is still fun too. If you can't find styles useful for this because the normal "style" category is so broad, I have a collection sorting them (actually, I've got two, but barely anyone uses Pony anymore).
Memes (Jack)
Make characters do funny things. The "meme" tag is probably the only model tag on Civitai that consistently gets applied, so its good to browse.
Experiment with tags
Danbooru tagging used by SD1.5/PDXL/Illustrious is rigid, yet deeper than many assume. Listing underused tags that do fun stuff was the focus of my Tag of the Day series and it only cracks the surface. Look through Danbooru's tag list by count (either through a dump closer to the model's creation or Danbooru itself, which has the benefit of often having articles explaining what a tag does), and find stuff that pops out to you. Concepts with 200+ uses (at least when Illustrious was trained) will often work surprisingly well without any kind of LoRA, and even many popular characters can be spawned without issue. Give some of these a try or two with blue buzz and add them to your techniques, and can help you stretch limited yellow buzz by getting higher quality out of the gens you make.
Make a LoRA
Making LoRAs is fun, not that hard (just follow my guide), and can potentially earn you yellow buzz if someone uses it for Not SFW generation (don't be too optimistic: Even before the changes I would get less than 1k a week from everything total, and most of that was from a single series of models). As far as I can tell, making Not SFW focused LoRAs with the on-site trainer using blue buzz is still possible (for now), though you'll likely want to get some good sample images to make people actually want to use your result (play Knights of New to scrape up enough to get started? Or generate externally.)