Ask
the work.

EDITORIAL DRAFT / OBSERVATION · OPEN · V0.2

01 / WHAT IS AN ADDRESS?. Ask for the capital of France. France identifies the entity; capital identifies the relation you want to read. Currency would ask something different about the same place. In the opening animation, that question arrives at an FFN and a Paris direction is written into the residual stream—the model’s changing working state. The conveyor is an explanatory drawing. The address in a model is a vector of numbers, and matching it need not isolate a single neuron or a single fact.

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 / 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 / OBSERVATION · OPEN · V0.1

08 / FROM A PICTURE TO A MECHANISM. The Mechanism continues the investigation with something we can build and inspect. Its small FFN plants six address–value pairs by hand: a query activates key directions, and their activations combine value directions into an output. The saved example recovers all six answers. This constructed memory makes the operation tangible. The repository also contains separate native-model experiments; the toy’s clean addresses are not themselves evidence that a trained model organises every fact this way.

EDITORIAL DRAFT / OBSERVATION · OPEN · V0.1

THE WALK WAS CORRECT. THE ADDRESS WAS NOT.. South to Corvin is a correct traversal — of the wrong chain. The relational operator works; it starts from the wrong place. At long range the model loses the key, not the ability to follow an edge. That points somewhere specific: resolve the path deterministically against an index outside the model, materialise a compact typed closure, promote it once, and let the model perform only the local operation. The run’s own depth-one result is consistent with that division — stepwise promote-then-retire was indistinguishable from replaying the resolved closure in one go, so the stepwise machinery bought nothing.

EDITORIAL DRAFT / CONNECTION · OPEN · V0.1

Retirement turned out to be the mechanism rather than the optimisation, and addressing rather than traversal turned out to be the bottleneck. Both point at the same engineering question the software has been circling: which operations should a model be asked to perform, and which belong to an index outside it.. LARQL / querying learned systems. VINDEX3 / the representation. What has to survive?. Context should be abundant

EDITORIAL DRAFT / OBSERVATION · ONGOING · V0.1

THE WORK. LARQL is a software project for exploring learned systems. The question behind this record is how the model’s internal representations become something we can inspect, address and work with.

EDITORIAL DRAFT / OBSERVATION · ONGOING · V0.1

THE QUESTION. What changes when a model is approached as structured, addressable material? The proposition directs attention to representation, the relationships between its parts, and execution.

EDITORIAL DRAFT / COMPARISON · ONGOING · V0.1

A model — a conceptual comparison. As a file: An object to load; Addressed from the outside. As a database: Parts to address; Relationships to inspect.

EDITORIAL DRAFT / QUESTION · OPEN · V0.1

What if the FFN is a graph? The question is about whether the representation makes something useful possible: addressing, traversing or explaining the computation.

EDITORIAL DRAFT / OBSERVATION · OPEN · V0.1

THE THREAD. Start with the representation. Ask what can be addressed. Specify a traversal, then compare it with the original computation. Keep the distinction between an explanatory picture and an operational result.

EDITORIAL DRAFT · ONGOING · V0.1

LARQL explores querying learned systems. This dossier connects the software to questions about model state, representations and what it means to make a model addressable.

INTERACTIVE STUDY / CONSTRUCTED EXAMPLE

Choose a planted address, inspect its key matches and suppress a neuron in a constructed six-fact FFN. The browser study makes the calculation inspectable; it does not run a trained language model.

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. 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.

THE PRACTICE / AUTHOR DESCRIPTION

Chris Hay. A house for ideas, systems and objects. Building things to find out how they work. Ideas: Questions, research and the notebook. A place for thinking before the answer. Systems: Models, software, design languages and infrastructure. Giving an idea a structure that can be explored. Objects: What is made: software, interfaces, films and publications. An idea takes a form that can be used, watched or held. Chris Hay explores how intelligent systems are represented, executed, understood and experienced. Research, Engineering, Design, Film form one practice. LARQL: Querying and running learned systems. An investigation into what becomes possible when a model can be addressed. VINDEX3: Exploring how a model is represented, and how that representation relates to its execution. HAUSE: A language for how intelligent systems express themselves: claims, evidence, questions, comparisons and refusals. MCP-CLI / CHUK: Tools for connecting agents, models and the systems they use. Research: The questions, the evidence and the boundary of what has been shown. Notebook: The record before the answer. Thoughts, sketches and questions while they are still taking shape. Film: Another medium for thinking. Experiments, explanations and conversations in public. The record is the catalogue of the practice: stable identities, authorship, dates, source credits and citations. Publish the uncertainty as well as the result. IBM Mixture of Experts is an external production in which Chris is a regular panelist; IBM remains its producer.