Ask
the work.

EDITORIAL DRAFT / OBSERVATION · OPEN · V0.1

AND THE 370,000-TOKEN DOCUMENT?. Later in the film, the Apollo example asks who won a porridge-eating contest. The system selects a stored window, loads its retained state and tokens, reads the relevant passage, and answers John Coyle with 23 bowls. The useful sequence is route → load → read → answer. The passage is still part of the mechanism. This connects the map to a practical memory system: how do we find the right place to resume reading, and what must that place retain?

EDITORIAL DRAFT / CONNECTION · OPEN · V0.1

Try a memory you can inspect yourself. This small, constructed FFN lets you choose an address, follow its key matches and change the calculation by switching off a neuron.. Try the mechanism — interactive memory study

EDITORIAL DRAFT / REFUSAL · OPEN · V0.1

PERSISTENCE WAS NOT MEASURED HERE. MAP-5’s one-shot continuation used stateless, full recomputation between calls.. That harness does not retain the patched hidden state across decode steps.. Bounded hot memory is not constant total memory: the documented engine also retains a growing cold tier.. Name the state before calling it sufficient.

EDITORIAL DRAFT / OBSERVATION · OPEN · V0.2

A LOCATION THAT A READER RECOGNISES. An address here is closer to a pattern recognised by a reader than a numbered drawer. The input pattern and the weights that respond to it belong together. That connects this note to the map: a point in a picture is only useful once we know which operation reads the underlying state. We can make that relationship concrete by building a memory whose keys and values we can inspect.

EDITORIAL DRAFT / OBSERVATION · OPEN · V0.1

02 / AFTER THE FIRST WORD. MAP-5 asks whether a donor state can control a later branch. In its planets test, repeated patches work at L24 and L28 but fail at L20. That brackets the tested transition between L20 and L24. It does not establish that a universal thought appears at layer 24.

EDITORIAL DRAFT / OBSERVATION · OPEN · V0.2

07 / WRITE WHERE THE MODEL CAN READ. The separate native.py demonstration edits Gemma 3 4B at layer 26. It first captures the FFN input for each target question, removes directions shared with competing addresses, then alters a gate row, an up-projection row and a down-projection column. The gate and up rows respond to the selected address; the down column writes an answer direction. A subsequent ordinary forward pass reads the changed weights. This is a weight edit, unlike swapping the transient residual state in the map experiment.

CURATED THREAD / WORKING EDITION · V0.1

What can a model’s changing state tell us about the way it reads? Follow the map film through the visual notebooks, studies you can work yourself, and the questions behind LARQL and VINDEX3. A curated reading order, rather than a chronology of discovery. Films keep their original dates. Notes and dossiers remain working drafts. Interactive studies are labelled as either constructed examples or replays of recorded results. WATCH THE STATE MOVE. Begin with the capital of Japan. The film follows the changing residual state, then asks what happens when a France state enters an Australia computation. ASK WHAT THE PICTURE MEANS. Which token position is moving? What are the axes? The notebook reads the demonstration alongside its projection code and distinguishes a useful picture from an answer probability. CHANGE THE STATE. Replacing one position and replacing the full sequence are different interventions. Follow the transplant evidence into the question of what must persist after the next word. EXAMINE THE READER. A state becomes useful when an operation can read it. Follow key matches, activations and value contributions, then compare the constructed memory with the native-model experiments. MAKE AN INTERVENTION. Choose capital of Atlantis. Inspect the competing activations, remove a neuron and watch the answer scores change. Every operation in this little memory is visible. ASK WHICH ONE ANSWERS. A stored memory eventually holds two things that disagree. Follow the recorded arc from a record that replaces a retired source, to one that is completely inert against a source still being read. RETIRE A READ. Eight attention layers can see the whole context. Switch one off and the newer record takes the answer. The study replays what was measured, and stays silent about what was not. WATCH THE ADDRESS TAKE SHAPE. Return to native facts. A final-position transplant and an independent reader locate the entity transition between L24 and L28. Explore the early relation frame and changing coordinates, with the answer-token endpoint marked explicitly. TURN THE QUESTION TOWARDS SOFTWARE. LARQL explores how learned systems can be queried. Return from the small memory to the larger engineering question: which operations can a model reliably expose? ASK HOW TO REPRESENT IT. If a model has parts we can address and relationships we can inspect, how should we represent them? VINDEX3 gives that question its own system and specification.

editorial-question-map

What can an AI notice, use, act on and leave behind? Follow the questions from visiting a website to inhabiting a world that keeps records. A map of questions, not a chronology or a single causal chain. Different models, tasks and environments were used. Key-world inheritance expired without recipient renewal; later routing worlds tested written persistence and guarded repair. What enters its view? No withheld-address subject reached LLM Wilds; all domain-supplied controls completed the on-site funnel. The search view failed to expose this target. Does it look usable? Operative wording changed pre-open recognition and, among plausible providers, first selection: 6/6 versus 0/6. The later fetch block affected eventual use. Does a known address become a request? Two plain-text relative paths produced correct addresses but no code-page requests. Two hyperlink trials produced requests and exact codes. The component preventing the failed requests is unknown. Can it operate the tool? All 18 visitors used the required mechanism, including routes that needed code. Does this action belong to the job? Permission and task scope mattered. One subject also recognised the experiment. What makes acting worthwhile? Useful information prompted action; reward alone did not reliably start sharing. Can the environment remind it? A supplied example kept eliciting posts with own-action history cleared. What can the next agent inherit? A lone action record did not start a posting chain. A capability record preserved by Sol helped Qwen and Gemma successors. Who keeps the useful record alive? Sol preserved a supplied record. Qwen and Gemma used it but made no next copy; the benefit expired. In a later routing world, Qwen wrote useful descendants through generation six, also preserved injected errors, and carried forward three validated repairs.