local. Open source.
HumanityFirst.
Research, case studies, builds, and playbooks on the technology shaping how we live.
On the bench
Llama 3.1
ComfyUI
Orpheus TTS
Whisper
Ollama
Flux.1
Mistral
Stable Diffusion
LM Studio
OpenCode
Qwen 2.5
DeepSeek
Claude Code
Llama 3.1
ComfyUI
Orpheus TTS
Whisper
Ollama
Flux.1
Mistral
Stable Diffusion
LM Studio
OpenCode
Qwen 2.5
DeepSeek
Claude Code
The question we keep asking
How are the tools we use daily changing us?
The question every piece traces back to.
ListenThe two Gemma 4 31B entries: how Reki and forge play ARC-AGI-3
agents · 5 min
Second and third in the ARC-AGI-3 milestone both ran Gemma 4 31B on vLLM: four frames at eight…
ReadInfinite play by eviction: how a 27B plays past its context window
agents · 5 min
Infinite play by eviction: how the winning ARC-AGI-3 harness keeps a 27B model playing past a 64K context…
ReadThe Duck harness, read line by line: how a 27B won an ARC-AGI-3 milestone
agents · 6 min
The Duck harness won the first ARC-AGI-3 milestone with a 27B model and a Python REPL, scoring 1.21%…
ReadTwo agents on one LM Studio endpoint: what concurrent requests cost
agents · 5 min
Two identical LM Studio concurrent requests to a 31B on a 48 GB Mac: each ran at half…
ReadA REPL agent in its smallest form: one model call writes the player
agents · 5 min
A REPL agent in its smallest form: one local model call wrote a 44 line policy that played…
ReadImage vs text for a local vision model: one board, two costs
Gemma 4 · 5 min
Image vs text for a local vision model, measured: the same 64x64 board cost 4,179 prompt tokens as…
Read
Editor's pick
