專門設計給 AI 寫程式的程式語言 -- 由 AI (gemini 3 pro) 指導 ccc 設計
對話連結如下:
先準備好 rust 環境 (包含 cargo)
測試: run_sort.rs
(py310) cccimac@cccimacdeiMac asym-kernel % cargo run --example run_sort
Compiling asym-kernel v0.1.0 (/Users/cccimac/Desktop/ccc/project/asym-kernel)
Finished `dev` profile [unoptimized + debuginfo] target(s) in 1.56s
Running `target/debug/examples/run_sort`
--- A-Sym Matrix Sort Test ---
Input (Unsorted): [5.0, 1.0, 4.0, 2.0, 3.0]
Executing: MatrixSort
Output (Sorted): [1.0, 2.0, 3.0, 4.0, 5.0]
Verification: PASSED
測試: run_branch.rs
(py310) cccimac@cccimacdeiMac asym-kernel % cargo run --example run_branch
Finished `dev` profile [unoptimized + debuginfo] target(s) in 0.22s
Running `target/debug/examples/run_branch`
--- Scenario 1: High Mean Data (100, 200, 300) ---
--> [Decision] Current Mean: 200.0000, Threshold: 50
Path: TRUE_PATH (Mean > Threshold)
Executing: Normalize
Final High Result: [-1.2247449, 0.0, 1.2247449]
--- Scenario 2: Low Mean Data (1, 3, 2) ---
--> [Decision] Current Mean: 2.0000, Threshold: 50
Path: FALSE_PATH (Mean <= Threshold)
Executing: Square
Executing: MatrixSort
Final Low Result: [1.0, 4.0, 9.0]
測試: run_memory.rs
(py310) cccimac@cccimacdeiMac asym-kernel % cargo run --example run_memory
Finished `dev` profile [unoptimized + debuginfo] target(s) in 0.25s
Running `target/debug/examples/run_memory`
--- Task A: Process and Store ---
Executing: Square
--> [Memory] Stored tensor to slot: 'feature_01'
--- Task B: Load and Finalize ---
--> [Memory] Loaded tensor from slot: 'feature_01'
Executing: Normalize
Final Result from Memory: [-1.1111674, -0.20203052, 1.3131976]
我想,或許很快,AI 會設計出
全世界最多『人+AI』使用的『程式語言』
到那個時候,身為人類程序員
就差不多要退出世界舞台了 ...
但是,你可以叫 AI 用那個 AI 發明的程式語言,寫出你要的程式
於是,我叫 Gemini Pro 設計一個這樣的語言 (稱為 A-Sym)
他設計出來了,我看不太懂 ....
我照著做,第一版他說用 rust 寫
AI 除錯幾次後,程式還真的能跑,目前最大的瓶頸是我的理解力 ....