Exploring a white-box AGI architecture: metacognition (self-awareness loops), continual learning (knowledge flywheel), world models (conditional space + semantic spatiotemporal graphs), self-improvement (bootstrapping discipline), zero-LLM white-box pipelines, and auditable trust guardrails.
Skill Claude Code
A compiler check for removing array or range-bound checks when the allowed range can be proved safe. A compiler translates source code into a form a computer can run.
Skill Claude Code
A control-flow graph builder for finding the jump relationships between basic blocks. A control-flow graph is a map of the possible execution paths through a program.
Skill Claude Code
A bytecode serializer that encodes an operation name and its argument type into bytes. Bytecode is a compact form of program instructions used by a virtual machine or other runtime.
Skill Claude Code
A function-signature parser that turns a comma-separated parameter list into parameter names, removing default values.
Skill Claude Code
A stack-depth analyzer that simulates an instruction stream to find its greatest stack depth. A virtual-machine stack is temporary storage used while instructions run.
Skill Claude Code
A compiler analysis that calculates a trust value for expressions from known environment values. It combines logical operations using minimum for AND, maximum for OR, and a complement for NOT.
Skill Claude Code
A compiler operation for chained comparisons such as `a < b < c`, treating them as `a < b` and `b < c` with short-circuit evaluation. Short-circuit evaluation stops once the result is known.
Skill Claude Code
A tracker for changes in information across recorded events, with operations for getting the latest value, finding a maximum, or recording data.
At most 3 mods per repository are shown here, and a mod shipped inside a plugin is left to that plugin's page — the rest are on their repository pages: