Matrix Particle Programming — Author: Wang Jiaocheng
Artificial intelligence programming is mainly limited by the context being too small to remember the complete code of software projects within the output rounds. Once it exceeds the memory rounds, obtaining subsequent code becomes chaotic.
The solution is to break down a software project into smaller parts, to a granularity that allows artificial intelligence to provide the complete code for that granularity within the limited context and memory rounds. Then, the complete code of the split parts can be integrated into the complete software project code by either humans or artificial intelligence.
The one-dimensional linear hierarchical granularity is still too large for artificial intelligence, so a two-dimensional matrix block granularity is used, with a 10-row by 12-column matrix block, totaling 120 particles:
Rows:
Server Database
Server Entity
Server Repository
Server Service
Server Controller
Client Data Source (Android)
Client Repository (Android)
Client UseCase (Android)
Client ViewModel (Android)
Client UI Layer (Android)
Columns:
L0 Business Ecosystem Layer
L1 System Layer
L2 Subsystem Layer
L3 Security Architecture Layer
L4 Module Layer
L5 Package Structure Layer
L6 Class and Interface Layer
L7 Method Layer
L8 Code Block Layer
L9 Statement Layer
L10 Expression Layer
L11 Atomic Operation Layer
All AI IDEs attempt to place all the code of the entire software project within a single context. Various agent creations and calls actually disrupt the context, making the implementation very complex and the results still unsatisfactory, requiring users to pay for the complex implementation work.
In fact, it is entirely possible to abandon the attempt to place all the code of the entire software project within a single context, as well as the context disruption method of agent creation and calls, and directly adopt the matrix particle method for programming, which can be achieved by everyone using a free artificial intelligence main dialogue box.