GhostTown: a city of synthetic residents
GhostTown simulates a city of synthetic residents with randomized profiles and emergent social patterns. The next version opens more of that world to interaction.
Lynn Cole / Laboratory
Laboratory
Experiments, research notes, mad science, and things in progress.
What a tiny desktop agent is good at
A tiny desktop model is not equally good at everything. My tests separate useful automation, tagging and summaries from the reasoning tasks where it struggles.
A small local agent: part bookkeeper, part oracle
Blossom12 gives an agent associative memory behind a ticket-system interface, with a tiny local model doing the bookkeeping and recall.
Sporkle: putting memory and device orchestration first
An uncomfortable truth: our CUDA based AI stacks were built by people who optimize equations, not hardware.
The GPU worked—and then I discovered the CPU regression
The first GPU implementation worked. Then a CPU benchmark exposed a regression: a novel memory-management system had been replaced by seven nested loops.
Sliding context windows and the case against lossy compaction
Sliding context windows, agent-directed retrieval and clearly labeled tool results are my alternative to lossy context compaction.
Toward Sporkle: a compute-framework progress note
A Fortran compute-framework experiment reaches strong CPU benchmarks while GPU and mesh networking remain explicitly mocked.
Planning a CPU implementation of CUDA-style operations
A plan to implement CUDA-style operations on CPU: start with kernels, then optimize around memory and concurrency.
Cleaning competing implementations out of an inference stack
Competing forward passes and cache implementations exposed drift in an inference project. Aggressive cleanup comes before the next experiment.
What is Horizon Beta?
A dated investigation of Horizon Beta through its coding style, creative behavior, constraints and failure modes. What kind of model was hiding behind the name?
Building a transformer in Go: a night of increasingly interesting nonsense
A Go-native transformer almost works. Tokenizers, vocabulary, norms, cache initialization and attention all take their turn in a night of increasingly interesting nonsense.
Porting Memvid to Go: token-aware chunking
Porting Memvid to Go exposed a chunking problem: six messages can mean wildly different token budgets. A token-aware strategy makes the boundary more predictable.