I prefer ITS DDT (aka HACTRN), with its built-in PDP-10 assembler and disassembler, that lets you do things like your login file customizing your prompt in assembly code to print the time by making system calls, without actually spawning any sub-jobs to merely print the time:
..PROMPT
holds the instruction which DDT uses to type out a "*".
You can replace it with any other instruction.
To use "%" instead of "*", deposit $1#%$> in that location
($> to avoid clobbering the opcode without having
to know what it is)
If you have to use arcane syntax and grawlix profanity, you should at least have direct efficient access to the full power of the CPU and operating system.
I keep submitting PR’s to get my assembler extensions in Fish and ZSH but so far no avail. Ideally all scripting should be done in single-clock-cycle assembly statements.
I mean it makes write-only languages like Perl look like beautiful prose but it’s hard to argue about efficiently setting the 20 environment variables used by my terraform jobs with a mere 20 clock cycles. It may seem silly but every clock cycle truly matters.
There's no need to tweak the default prompt with this approach. Just make sure that, at the point when the model starts generating, it already has "Yes sir!" as the first tokens of the response message.
It's very easy in the API, obviously, but most local chatbot apps can also do this. E.g. in text-generation-webui, there's literally a textbox in the chat labelled "Start reply with". In LM Studio, you can pre-create a response message with the desired prefix and then use the "Continue" action on it.
The human mind wonders and takes time to dream autonomously. Perhaps the llm.c we need for the next breakthrough addresses rounds of meditation in it's training in order to provoke more reason alike features to the NextGen LLM.
Interesting I have the same task, can you share your tools? My goal is to detect if documents contain GDPR sensitive parts or are copies of official documents like ID's and driving licenses etc - would be great to reuse your work!
reply