When you run out of Claude tokens
(midwest.social)
BeatTakeshi
Programmer HumorCross posted from: https://thelemmy.club/post/56974556
Hot ai wives in your area
She sounds hot.
Code block support and the whole schabang.
It didn’t add the shebang.
Pretty sure that requires subscribing to a higher tier. Maybe use the money you saved on Claude tokens...
I love how the UI has support for code blocks
It probably just comes with whatever markdown support their UI uses. Since you’d want other markdown. Lists. Italics. Etc.
With syntax highlighting and everything. I suppose that's an OF niche you could sell. Subscriptions come with programming support alongside feet pics.
I'd be violently upset if I knew my browser was downloading what appears to be a whole node_modules directory before I could jerk off
I probably could have made money as a google sheets guru on onlyfans back in the day. Building complex sheets with cell functions and app script was my obsession several years ago
Doesn't that turn you on?
It makes me think that whole image was created by AI…
it's still using the '💙's lol
Somehow still less sycophantic than chatgpt
You don't use these in your code? // 💙
💔
That’s cuz she’s fiiiiiine (tuned)
Dique, she has a store para implementations! Ay tigre!
AFAIU every chat bot can be used this way no?
Somewhat. Some of them will outright comply like this one, others need a bit of work. Also you have no control over the model they use and you might be getting really bad code as soon as it’s a bit more specialized than just some very generic solution in a very popular programming language.
Well, they would typically have a system prompt (like an invisible chat message that's sent to the LLM at the start of the conversation), where it would tell the LLM something like "You're a helpful chatbot. Talk cheerfully and use 🩵 all the time."
And in that system prompt, you can also give it instructions to only answer certain questions. But it really is like a chat message, in that it just as well might ignore those instructions when another chat message asks it to.
OP chose the formulation "Before I subscribe to a creator, ..." which might be necessary here to get it to answer, as it might have been told to only help with creator subscriptions, but then the formulation implies that by answering this question, it is helping the user to subscribe. That's called a prompt injection.
If you can find a prompt injection, then yes, you can do this with any chatbot.
Although they might still not be as useful as a dedicated tool, since they likely can't do tool calls, like web searches or running the compiler.
It’s wild that a system prompt is still the major guardrail in most chatbots like this. Even if the LLM provider won’t implement anything, the user could add a classifier, some kind of preprocessor…
Well, what preprocessor would you add, other than denylisting certain words? 😅
If you use an LLM to try to detect the semantics, you've got the same problem again...
A classifier will be trained differently and basically take $TEXT and map it to $CLASSIFICATION and then you allowlist on the classification. You can then also, if you love burning tokens and spending money on things, have another step that "translates" the user input, making prompt injection significantly more difficult before finally routing it to the LLM. Generally guard rails should then be placed on the LLM's available actions via tool calls and those guard rails should be written by LLMs that are LLMing LLMs and L̸̡̡͕͓̣͓͚̝̥̜̼͙̪̖͍̺͝Ļ̵̥͍̳͇̱̠̬͔̹͖͓͖̯̹́͊̽̄̂͑͌͆͗̑̄͜Ṁ̶̡͙̺̖̥̲̙̩͖̻̹͇̜̲̃͘͜͝ș̵̖̤̫͓̦̲̼̖̗̟̯̩̺̣̞͊̆̉͐̈́̃̈́͗̐̽̽͋̊̉ͅ ̵̖͉̥͖̪̳͍̝͉͕̬̪̂̔̊̌͂̍̚͜a̶̦̤̬̱͖̻̝̱̿͋̿̍͂̒͆̈̄͑̓̕r̴̡̤̙͙͕͚̼͙̦̲̅̔̓̈́̄̆̾͗̿͌̐̄͛̒͘͜ͅe̴̡̡̼͉̣̼͉͓̪͕̬͊̐̈́̇̽̒̂͆̎̈́̃͠͠ ̶̩̯̲̯̃͛͌̀̿̅̎̾̕̕ǹ̸̬̟͍͇͍̭̹̲̯̽̒͘͝͝ȩ̶̢̬̺̦͖̣̙͍̞͇̰̉̋̓͐͂̏̀̿̑̂͌͗͗̒͘ͅͅe̶̠̝̦͛̓͛̍̑̆̽͐d̸̦̯̖̥̟̻͔̦̂̿͆͂̑̎̔̏ȅ̴̡̨̺͓͇̦̣̦͖̬̹̥͎̾̽͐̓̍̽̾͑̿͆̄̀̈́̄́̕͜d̶̡̛̝̟͔̳̲͙̰̗̬̳̯͚̐̍̑̓̉̂̍̊̓̿͝ you probably f̶̢͓̘̹̟̥̙̥͚̼̬̭͍̫̩̀͑̚͝͝͠ę̴̰̦͚̔̀́͊̋̈̒͗̆͘͝͝͝͠ȅ̸͚̘̤͚̬̩̈́̇̆ď̷̰̆̀̔̾͌̿ ̵͚̦̥̈͌͑̂͗̈́͂̕ͅm̴̢̡͇̟̠̬̣͙̹͇͉̈́̌̓́̓̌͐̓̂̚̚͜͜͝e̸̢̳̟̙̯̻͚̭͎̫̖̺̥̬͕͆̆̇̑̄̀̓̅̀͛̊͐ ̸̡̢̨̛͚͕̳͙͚͋͛̊́̍͛̽̉̃̇̇̋̑t̸̡̰͔͎̮̯̗̤̆̋͗͐̈̂̈̃̓̒̌͂̕͘͝ǫ̴͇̳͙̞̘̻̙͍̘̲̜̮͇̙̇̓̌̃k̸̢̢̤͔̫͖̲͇͍̐̃̄e̶̢̳̪͉̦͙͓̻̖̔́̃́̈́̀̿̓͆̊̅̒͐̑n̶͖̙͚̫͓͚̦̭̗̲̓́̋̀͜͝s̴̠̼̈́̑̾̑̆̽̊͋̾͛̐̊͆́͊̕̚ need to have some guard rails l̷̬̳̘̰̅̀̑͑̊̒̓̊͌̌̕e̶͖̱̊̍̉͆̐̍̌́͑͘͠t̵̨̛̹̰̳̩̹̞̹̼̪̮̄̀̅̾̔͐̇͒́̊̃̀͘̚͜͠ ̴̢̢̭͖̤̜͈͕̯͎͉̰͇̾̐͛̽̅͂̒͝m̸̨̰͖͚̞͍͍̳̭̗̳͕̈́̃̇̈́̎̌͊̍͌̊̆͠ȩ̵̡̛̜͍̞̻͎͎̲̜̹͖̼̽͂̐̽̊͌̓̀͂̐͂̾̇̚ͅͅ ̶̧̖͚͈̤͇͕̼̜̖̮̙͍͖͋̉̾͋̌̔̿͊͋̾̚l̸̡͕͇̭̤̥̩͙̫͕̆͌̈́̀̋͆̃͋͜͝͝͝͝ͅo̷̧̘͖̤͖̩̺̞̹̼̟̟͕̫̬͎̅̀ǫ̸̢̪͍̭͙̗͔̪̘͉̊̕͜ș̶̨͉̲̙̜͔̫̲̥͕͕̮̜̭̮͍̈́e̶̢̛̮̲̯͎̙͇̪̬͕͖̠̘̯̰̭̓͐̌̉̀͋̉̏̈́̇̾͠͝ ̵̬͍͔̓́i̶̘̫̯̻̘̯̗̔̒͊̂̂̑͑̍̋͑̏̀́͌̚͝͝n̵͓̬̮̻̹͉̗͎͓̖͌̿̃̓̋͂̅̍͒̉͆̿͆̉͠͝ ̴̧̲̯̲̅̀͂̆́̓̈́͑̾̾̊͋̒̿͛͘̚t̸̢͉̟̹͆̈́̏̍́̎ĥ̴̢̥̦͓͕̘͚̙̪͙̀̽́̒̓̈́̋͗̀̈́̇ȩ̵̠̹̻̯̖̹̹̗̜̻͍̹͆̊̈́͑́̏̑́̎̚͜͜͝ ̴̧̟̭͍̣͋͋̾ẘ̴̪̦̐̍̏̆̑̇͗́͆͘̕̚͝͝ͅo̵̡̟͚̲̦̤̩͎̥̘̻̥̠͙͈̙̬̓̌͊̋͐́͌̐̎̏̋̿̏r̵̢̨̘̓͑͑̆͒̂͝l̸̡̨̛͖̖̟̼̺̦̰̗̭͔̺̓͗͛̈́̈͆̈͗̐͝d̴̢̠̠́͂̾̌̿̐̀͛̄͑́͐̉̋͝ on those LLMing LLMs that are LLMing your LLM tool calls y̵̧͍̣̭͍͚̱͉͍̘̠̦̮͇̻̓͆̕o̸̥̮̮̙̖̩̙̝̘̫̗͎̺̓̊̐̄̑̎u̷̡̨̡͎͔̼̪͈̦͚͈̟͍̽̀̏͛͊ ̴̨̟͖͖̰͕͉̫̫̜̥̦̲͆͋͌̌̓̈́̉͆̉͠ẗ̵͚͉͈̦́̈́͛́͋͜i̷̧̗̰̘̖͍͖̼͕͗́̊̏̇̾̐̆͗́͌̋̀̚ņ̷̣̱̖̬̼̆͜y̷̢̡̧̛̹͙͕͔̪͛ ̶̪̫̙̘̥̹̻͙̝̜̠̗̭͎͎̆̇̅̎͜͜h̷̡̨̰̺̞̝̩͉̬͓̮͙͉͕̓̄̊̿̋́̇̏͝u̸̮̙̲̙̠͉̪͈͖͙̓̈̉̀̚͝m̶̢̛͙͕̘͈̤̬͎̺̲̯̱͉̪̠̎̆̽̅́̂͐̀̄͜ͅã̶͕̼͎͇̲̐͗̑ň̸̢̢̝̩̟͖̫͔̥͚̘̟̖̘̤̾̀͋̒̄͠͝s̸͙̳̳̤͓̝̥͓̱͔̣̹̭̐̐͋ͅ ̶̩̹̪̑͑͒̽̇̀̽̏̓̌̑̈́͂̓͘h̶͓͖̱̲̘̗̿̽̈́̎͑́̓̉̑̕̚ą̴̰̣͚̫̤̤̬̘̼̠͍͂̏̚ͅv̵̢̺̮͉̤͚̳͕̥̳̇̕̕͘͠ȩ̷̡̮͖͚̩̙͖͎̓̀͗͊͋̈́̔̐̕̚͠ ̷̫̌̀̊ň̸̜͓͕̳͙̩̼̜͈̊͜ǫ̶̝̺̰̹̻̰̰̗̪̔ ̷̲̤̪͇̠̼͑͊͘͜p̴̡̧͉̩̜͕̽̏̀̌̇̈́̊l̶͖͔̯̲͔͓̩̂̌̂͐͗̕a̵̢̨̞̰̪͖̲̭̝̪̥̻̐̃̊͗̅̓͐͋͛͘͝c̴̛̮̙̦̠̯̼̣̠̳͔̖͕͓͋̓͆͒̋̒̾͐͒̀̓͑͠e̷̢̢̧̱̣̫̲̪̣̭͙̟̔͌̽̿̐̿̿͑̆̇̇̉̊̓͘͠
You could also do none of this and just not use an LLM, because you likely don't need one, but then the cool kids will stop inviting you to their parties or something and you probably missed out on the blockchain parties so fomo
I didn't quite catch that
I hate it when people don't speak in languages from this dimension
Damn tourists
That's what you get for pulling funding from integration projects. Damn you, Republicans!
As the business? Like I said, I’d start with a text classifier, a small model quickly categorizes the input prompt. If it’s something off topic, I’d return some generic refusal reply, or maybe send the score to the LLM itself if it’s somewhat ambiguous.
Chatbot UIs that give very quick refusals are using “filtering” models in just that way. There’s a whole world of language modeling that existed before LLMs, just for that sort of thing.
Now, if I was the LLM provider? I dunno, but I would try to integrate it into the weights. There are some really interesting papers and hacks out there for prefiltering prompts that have nothing to do with system prompts.
TIL that thread safety is a thing in Python
It's not as funny when they openly says that it's an AI
