I thought it would be fun to explore what would happen if people using assembly in the past would have access to local models. It's like a giving a caveman a bazooka, it sounds so poetic, I think.
With the new decision models coming through, I realised that this is perfect for being about to bring the "analog world" to the "digital world", so imagine I want a small program to adjust the room temperature:
```
psh "It's too dark to read and I am sweating."
br temperature
end
temperature:
dup
psh "How does the room temperature feel?"
psh [ "Much too cold" "Cold" "Comfortable" "Warm" "Much too warm" ]
scl ; Stack = [confidence, score, feedback]
psh 0.5 ge ife adjust ask_person
ret
adjust:
psh 2 sub ; "Comfortable" is index 2
psh -3 mul ; 3 degrees celsius per step
prt
ret
ask_person:
pop
psh "Unsure about the temperature: passing to a person." prt
ret
```
Something that I have been exploring is where does the future of programming fit in? This project is my call for myself to go back to basics and throw away all the modern complexities and comforts. Make the programming language as limited as possible while giving access to a language model. Only then can we see where language models and agents truly fit. Is the future purely language model driven? Or not at all? Or somewhere in the middle?
We are typically mixing in deterministic instructions with ones that operate over the semantic or fuzzy domains. If we can limit the model to only handle the probabilistic portion of the task while keeping the deterministic part in code, we have a clean system where complex prompts can be decomposed into a series of verifiable steps.
vanilla-latte•1h ago
I thought it would be fun to explore what would happen if people using assembly in the past would have access to local models. It's like a giving a caveman a bazooka, it sounds so poetic, I think.
With the new decision models coming through, I realised that this is perfect for being about to bring the "analog world" to the "digital world", so imagine I want a small program to adjust the room temperature:
``` psh "It's too dark to read and I am sweating." br temperature end
temperature: dup psh "How does the room temperature feel?" psh [ "Much too cold" "Cold" "Comfortable" "Warm" "Much too warm" ] scl ; Stack = [confidence, score, feedback] psh 0.5 ge ife adjust ask_person ret
adjust: psh 2 sub ; "Comfortable" is index 2 psh -3 mul ; 3 degrees celsius per step prt ret
ask_person: pop psh "Unsure about the temperature: passing to a person." prt ret ```
Something that I have been exploring is where does the future of programming fit in? This project is my call for myself to go back to basics and throw away all the modern complexities and comforts. Make the programming language as limited as possible while giving access to a language model. Only then can we see where language models and agents truly fit. Is the future purely language model driven? Or not at all? Or somewhere in the middle?
We are typically mixing in deterministic instructions with ones that operate over the semantic or fuzzy domains. If we can limit the model to only handle the probabilistic portion of the task while keeping the deterministic part in code, we have a clean system where complex prompts can be decomposed into a series of verifiable steps.
What's your thoughts on this? Be happy to hear!