Notes from the shadow.
System Zero, what we measured, and how to run a reflex in front of your LLM. Losses included. RSS
2026-10-22 · 4 min read
shad0w at work: five jobs real teams can give itSupport routing, a prompt guard, product photos, a live data stream and a zero-code proxy: what shad0w does in each, the code, and what it measured.Read →2026-10-21 · 4 min read
shad0w 0.4: one reflex for text, images and guardsshad0w 0.4 takes its learned table and certificate from text to pixels, prompt-injection guards and video frames. What works, what doesn't, and what to try.Read →2026-10-09 · 5 min read
Zero-code System Zero for the OpenAI Decisions APIPoint the OpenAI Decisions API client at the shad0w proxy: certified questions come back from a table in microseconds, the rest go upstream. No code change.Read →2026-10-09 · 3 min read
Snake at 120 moves a second: System 0 vs System Oneshad0w, three System One decision models and a hosted LLM play the same snake games at game speed. shad0w keeps scoring; the LLM falls behind its own moves.Read →2026-10-09 · 5 min read
94.3%, 51.3%, 0%: honest numbers against a hosted LLMWhat shad0w answered, and missed, when it shadowed a hosted LLM through the official OpenAI client: wins, losses and the traffic where it should defer.Read →2026-10-08 · 2 min read
Adding a category without breaking productionA trained shortcut and a changing list of options are a dangerous mix. The rule shad0w follows, and what happens when you add, remove or rename an option.Read →2026-10-08 · 4 min read
System Zero: the reflex layer below your LLMBelow your LLM (System 2) and decision models (System 1) there is room for a reflex: System 0, a table that answers repeat decisions in microseconds.Read →2026-10-07 · 5 min read
How shad0w proves it agrees with your LLMThe certificate behind shad0w, in plain words: a held-out test, a confidence line, an allowance for bad luck, and what the promise does and does not cover.Read →