# Voidmill mills documents into knowledge through a curated model stack After installation, its specialists build document knowledge on your machine; Voidmill answers with the passages that match. ## Voidmill’s specialist passes build one enriched knowledge database Each pass adds detail to the same database, from named things to summaries and links between passages. A model alone does not know your documents. Voidmill keeps what works in document search and adds a curated specialist model for each pass: reading, naming, classifying, representing meaning, summarizing, and linking. Instead of one model over one meaning index, the passes feed one another to build one SQLite database. With its models installed, Voidmill builds and answers without your documents leaving your machine. Today those models are tuned for life-science and scientific literature, with no promise for other fields. ### Voidmill stores figures beside passages in one database and searches them During a build, Voidmill also turns images into vectors: figures inside documents, standalone image files, and, when chosen, full page renders. After installation, it searches those images on your machine with a question in words or an example image. It finds figures on the right subject but does not yet reliably put the exact figure first. A search by example image carries no verdict yet. For images, an “answer” verdict means the question is on-topic, not that the figure exists. Voidmill’s passes in order, left to right, then its acceptance checks. A diagram of how Voidmill builds, not a screen it shows. Documents in, one knowledge database out, with each pass feeding the next. The stages as text - **Reading**: Reads many formats, keeps tables and figures, and names declined files with reasons. - **Naming**: Picks out the named things in the documents and stores them in the database. - **Classifying**: Classifies the documents, keeping those classifications in the database. - **Representing meaning**: Turns text, and images too, into vectors for the meaning indexes. - **Summarizing**: Writes a summary of each document, which the database holds. - **Linking**: Stores the links between passages in the database. - **Checking**: A database is not called finished until it passes its own acceptance checks. ## Voidmill separates failed searches from searches that found nothing Every document answer includes sources and a verdict: “answer,” “weak answer,” or “no answer.” Because a failed search reports the failure, a “no answer” verdict never hides one. Voidmill’s showcase is the scientific knowledge library, which holds 8,297 scientific and technical documents and 216,789 passages. Most of the library is primary scientific literature, gathered as full text that is freely available online. The showcase holds no images. On its largest database, the scientific knowledge library, some questions outside its subject still receive an “answer” verdict when “no answer” would be right. --- ### The scientific library Voidmill built This library brings together published papers and web sources across the life sciences, organized by Voidmill into a searchable database. 8,297 sources 4,577 published papers and preprints 3,033 web pages 99 research topics 8 fields of research 216,789 searchable passages Papers and web pages do not cover every source: 687 more documents are stored without a web address. Topic stars cluster by field inside slowly moving clouds; the founding collections sit at the center, and labels explain selected stars. Arrow keys move; plus and minus zoom; zero resets; Tab moves between stars; Enter pins labels; Esc clears. Explore full screen All research topics by field #### The founding collections The library began with these broad collections on biological manufacturing, brain research, preserved tissue, and the genetics of individual cells. - **Genetic studies of individual cells**: Methods for reading genetic information from individual cells are compared and combined with other measurements to study differences between cells. - **Developing biological manufacturing**: Research examines how cells are developed and grown to produce biological medicines and vaccines, from early processes to manufacturing. - **Neuroscience and brain disease**: Studies connect nerve cell biology, brain inflammation, and molecular changes with research into neurological disease and possible treatments. - **Molecular studies of preserved tissue**: Methods prepare tissue preserved with chemicals and embedded in wax for molecular studies, accounting for changes caused by preservation. #### Proteins and laboratory methods Research covers preparing samples and measuring proteins, genes, and living cells, alongside studies of brewing yeast. - **Preparing samples with sound waves**: Sound waves break open cells and shear DNA and its associated proteins, with methods compared for sample preparation. - **Antibody validation**: Antibody checks test whether a lab measurement detects its intended target and gives repeatable results. - **Protein interactions**: Mapping which proteins gather or work together helps researchers study the networks inside cells. - **Enriching modified proteins**: Methods concentrate proteins or fragments carrying chemical changes so those changes can be studied in complex samples. - **Measuring multiple DNA targets**: Chemistry and optical detection methods let researchers measure several DNA targets together in a single test. - **DNA fragmentation for sequencing**: Breaking DNA into fragments is part of preparing it for sequencing, where sample preparation can influence the results. - **Measuring proteins in cells**: Antibody tests measure proteins inside cells or on their surfaces to screen how cells respond to drugs. - **Fluorescence imaging in living animals**: Near-infrared light reveals fluorescent signals inside small animals, allowing researchers to study biological processes before clinical testing. - **Reproducible protein blot measurements**: Tests that detect several proteins on a blot are checked for consistent results and reliable measurements of protein amounts. - **Validating gene edits in cultured cells**: DNA tests and protein measurements check whether an intended gene edit worked, including whether a targeted protein is absent. - **Detecting protein modifications**: Different detection methods reveal chemical changes added to proteins after they are made, helping researchers compare ways to measure them. - **Preparing samples to map DNA binding**: Sample preparation methods preserve and isolate contacts between DNA and proteins so researchers can map where those proteins bind. - **Brewing yeast biology**: Studies of brewing yeast examine its genetic history, metabolism, and breeding to understand the cells behind fermentation. - **Large-scale protein analysis**: Research compares ways to prepare, measure, and analyze proteins across blood samples and individual cells to study biological differences. - **Preparing proteins for analysis**: Extracting proteins and breaking them into smaller pieces prepares samples for analysis, including when tissue is preserved or scarce. - **Measuring protein phosphorylation**: Protein blots measure phosphate additions on several targets at once to study signaling pathways inside cells. - **Kinase drug activity inside cells**: Tests measure how strongly drugs inhibit kinases, enzymes that add phosphate groups, and whether the drugs engage their targets inside cells. - **Imaging radioactive labels**: Imaging plates record signals from radioactive labels, allowing molecular biology experiments to be read and compared with photographic methods. - **Finding disease markers in tissue**: Protein measurements in broken-down tissue samples help researchers find and check possible markers of cancer. - **Separating tissue into cells or nuclei**: Methods separate tissue into cells or the nuclei that hold their DNA, preparing these parts for individual study. #### Blood cancers and cell genetics Studies trace blood cancers, treatment resistance, and genetic differences between individual cells, including investigations of hidden viral infection. - **DNA marks and cancer cell families**: Chemical marks on DNA and other controls of gene activity help researchers examine how related cancer cells develop distinct identities. - **Cancer cell diversity and evolution**: Genetically distinct cell families can coexist within a cancer, and tracking their changes helps researchers study disease progression and resistance. - **Overlapping bone marrow disorders**: Some bone marrow disorders combine abnormal blood cell development with excess cell production, and researchers study how these conditions progress. - **Acute myeloid leukemia biology**: Research examines gene changes and disrupted cell controls that drive acute myeloid leukemia, a cancer of blood-forming cells. - **Repeated gene segments in leukemia**: Repeated stretches inside a gene often changed in leukemia are traced in individual cells to study variation and drug resistance. - **Hidden HIV in individual cells**: Studies examine cells where HIV remains inactive to understand the hidden reservoirs in which the virus persists. - **Drug resistance in blood cancers**: Research investigates how blood cancer cells survive treatment and the biological changes that make therapies lose their effect. - **Expansion of altered blood cell families**: Blood cells descended from a genetically altered ancestor can expand without a diagnosed blood cancer, prompting studies of their clinical significance. - **Blocking cancer cell survival**: Studies follow individual cell families during treatment that targets a protein helping cancer cells avoid death. - **Genetic diversity within solid tumors**: Reading genetic information from separate tumor cells reveals differences that pooled samples can conceal, with possible applications in cancer care. - **Combining measurements from single cells**: Methods connect DNA variants with other measurements from each cell to study genetic differences alongside cell behavior. - **Choosing genes for clinical sequencing**: Designing a focused gene test means selecting DNA regions that can answer a particular clinical question. - **Linking cell genetics and protein markers**: Measuring DNA changes and protein markers in the same cell connects its genetic makeup with features used to identify it. - **Detecting cancer remaining after treatment**: Sensitive tests look for small traces of blood cancers after treatment to monitor disease that may remain. - **Single-cell genetics in clinical trials**: Studies assess sample handling and test reliability when individual cells are genetically analyzed as part of clinical trials. - **Genetic drivers of blood overproduction**: Gene changes that drive excess blood cell production are studied in individual cells to understand related bone marrow cancers. #### Stem cells and tissue growth Research follows how stem cells become specialized cells, how tissues are grown, and how cell therapies are prepared. - **Laminin and cell attachment**: Different forms of laminin, a protein beneath cells, bind specific cell surface receptors and help shape how cells attach to their surroundings. - **Stem cell models of disease**: Reprogrammed stem cells and gene editing let researchers build laboratory models for studying how genetic changes relate to disease. - **Growing heart muscle from stem cells**: Researchers guide stem cells toward becoming heart muscle and study how the surfaces beneath them support that development. - **Developing nerve cells from stem cells**: Researchers guide stem cells into nerve cells that release the chemical messenger dopamine or control muscles, studying how these specialized cells develop. - **Gene editing and cell families**: Researchers edit DNA and grow cell families from individual cells to study genetic changes among cells with a shared origin. - **Stem cell therapies and immune rejection**: Researchers develop donor stem cell therapies and ways to reduce immune recognition so the resulting cells could be used in different people. - **Scaling up cell therapy production**: Growing larger batches of therapeutic cells brings manufacturing challenges in maintaining consistent quality as production expands. - **Growing retinal cells**: Studies guide stem cells toward the eye’s light-sensing cells and their supporting cells, with potential applications in retinal therapy. - **Checking edits in individual cells**: Examining edited cells separately helps researchers check intended DNA changes and look for unwanted alterations, including rearranged chromosomes. - **Building tissue with laminin supports**: Tissue engineering and biological printing use supports containing laminin, a cell attachment protein, to help organize growing cells. - **Growing organ models**: Researchers grow small tissue models called organoids from different cell lineages to study how organ-like structures develop in culture. - **Maintaining versatile stem cells**: Methods keep stem cells growing in the laboratory while preserving their capacity to develop into many different specialized cell types. - **Developing insulin-producing cells**: Guiding stem cells into pancreatic cells that make insulin is studied as a route toward therapy for diabetes. - **Growing blood and vessel cells**: Stem cells are directed toward blood vessel, blood-forming, and immune cell identities for vascular research and cell therapy development. - **Developing liver cells and organ models**: Researchers guide stem cells into liver cells and grow small liver tissue models to study liver development in culture. #### Blood tests and cancer treatment Studies cover cancer genetics, blood tests for tumors, treatment selection, and prenatal screening. - **DNA repair and cancer immune treatment**: Faulty DNA repair in colorectal cancer is studied for its relationship to treatment that helps immune cells attack tumors. - **Genetic drivers of colorectal cancer**: Research examines the gene changes and molecular subtypes of colorectal cancer to understand how these tumors develop and differ. - **Blood tests for prenatal screening**: Comparisons examine placental cells and loose DNA fragments in blood as approaches to genetic screening during pregnancy. - **Liquid biopsies in cancer trials**: Cancer trials study how tumor DNA fragments and tumor cells found in blood can be measured together under clinical testing standards. - **Preparing DNA for targeted sequencing**: Methods prepare selected DNA regions for sequencing tests, including tests designed to detect cancer mutations present in small proportions. - **Tests that guide cancer drug selection**: Diagnostic tests measure tumor features to help determine which cancer treatments may be suitable for a patient. - **Colorectal cancer treatment markers**: Tumor location, protein markers, and cancer DNA in blood are studied together to inform colorectal cancer treatment and monitoring. - **Direct capture of tumor cells in blood**: Methods seek rare tumor cells directly in blood samples without a preliminary step to increase their proportion for detection. - **Changing tumor cell identity**: Markers in circulating tumor cells help researchers study shifts toward nerve-like and hormone-producing features associated with treatment resistance. - **Colorectal cancer cell evolution**: Following related cells within colorectal tumors helps researchers examine how cancer changes during treatment and develops resistance. - **Bloodborne tumor cells during treatment**: Examining the features of tumor cells circulating in blood helps researchers monitor treatment response beyond simply counting those cells. - **Validating clinical marker tests**: Laboratory checks assess whether tests for disease markers give reliable measurements suitable for use in clinical research and care. #### Making cell and gene therapies Researchers study how to produce cell and gene therapies and test their quality, activity, and biological safety. - **Cell and gene therapy development**: Research surveys treatments based on living cells or delivered genes, including the manufacturing challenges of bringing them into clinical use. - **Quality checks for engineered immune cells**: Protein tests and imaging help assess the quality of immune cells engineered to recognize cancer during therapy manufacturing. - **Testing therapy quality and safety**: Before cell and gene therapies are released, analytical and biological safety tests assess their quality and look for unwanted contamination. - **Unwanted virus testing in gene therapy**: Tests look for viruses capable of reproducing and assess virus removal during production to support the safety of gene therapy materials. - **Growing immune cells taken from tumors**: Immune cells collected from tumors are grown in larger numbers and tested for activity during the development of cell therapies. - **Delivering genes into therapeutic cells**: Process studies examine how genes enter immune cells during manufacturing to give them receptors that recognize cancer. - **Making gene carriers for immune therapy**: Manufacturing research examines targeted carriers designed to deliver genes to immune cells inside the body, including how their quality is checked. - **Making adeno-associated viral vectors**: Manufacturing methods produce and purify viral gene carriers, separating filled shells from empty ones and reducing leftover cellular DNA. - **Cells that produce lentiviral vectors**: Cultured cell lines produce modified viruses called lentiviral vectors, which serve as gene carriers for gene therapy. - **Gene therapy in blood-forming stem cells**: Manufacturing processes add genes to blood-forming stem cells outside the body using modified viral carriers, with checks on treatment quality. - **Measuring cell therapy activity**: Functional tests measure whether therapeutic cells perform their intended biological jobs, helping establish criteria for releasing a batch. #### Mapping cells in tissue Imaging and analysis locate proteins and cells within tissue to study tumors, brain disease, transplants, and treatment responses. - **Mapping proteins in brain disease**: Images that locate many proteins together help researchers study brain tumors, inflammation, and diseases in which nerve cells deteriorate. - **Imaging many tissue proteins at once**: Fluorescent antibody labels reveal many proteins in the same tissue section during one staining round, preserving a view of their locations. - **Analyzing whole tissue slide images**: Computer methods align tissue images, identify individual cells, and analyze their surroundings to turn large scans into biological measurements. - **Mapping transplanted tissue**: Maps of cells and molecules help researchers study organ rejection, including in transplants using organs from other species. - **Preserving targets during tissue staining**: Studies compare repeated staining with staining in a single round to see how well each preserves the tissue targets being measured. - **Tracking immune cell therapies in tissue**: Locating engineered immune cells and surrounding tissue markers helps researchers assess cell therapies and monitor their effects. - **Artificial intelligence for tissue images**: Artificial intelligence models learn patterns from tissue protein images without manually labeled examples to support analysis of complex biological images. - **Choosing antibodies for tissue imaging**: Antibody combinations are designed and tested to distinguish many proteins together while retaining useful measurements of their positions in tissue. - **Immune markers in cancer tissue**: Researchers test tissue protein markers as clues to interactions between tumors and immune cells, aiming to inform cancer immunotherapy. - **Tumor cells and their surroundings**: Protein maps show how cancer cells and surrounding cells are arranged, helping researchers characterize the environment in which a tumor grows. - **Combining molecular maps of tissue**: Maps of proteins, RNA messages, and DNA are combined to relate different molecular measurements to their locations within tissue. #### Antibodies and immune responses Research examines what antibodies recognize, how immune responses vary, and how those patterns are measured in health and disease. - **Antibodies that target the body**: Researchers measure antibodies that target the body’s own molecules to study autoimmune diseases and assess possible tests for diagnosis and monitoring. - **Patented methods for antibody profiling**: Patents describe methods that read DNA sequences to profile antibodies and study patterns in the targets they recognize. - **Antibody signals in cancer blood tests**: Research tests whether antibodies against the body’s own proteins can serve as blood markers for cancer detection. - **Immune receptor diversity**: Reading immune receptor genes reveals the variety of receptors carried by immune cells, helping researchers study how that diversity arises. - **Mapping antibody binding sites**: Methods locate the precise parts of a target recognized by an antibody, informing studies of binding and immune recognition. - **Antibody responses to infection**: Blood antibody tests reveal patterns of immune responses to viruses, supporting studies of infection, immunity, and vaccine design. - **Screening displayed proteins**: Displaying many proteins on viruses, cells, or other supports helps researchers select molecules that bind a chosen target. - **Making and capturing proteins**: Methods produce proteins without living cells and add tags that allow those proteins to be captured for experiments. - **Protein arrays for antibody testing**: Many proteins are arranged on a surface so researchers can measure which ones antibodies bind to in a single experiment. - **Mapping a person’s antibody targets**: Broad protein screens map which proteins a person’s antibodies react with to describe the reach of their immune recognition. Reset view Close full screen ## Voidmill’s document stack can change as better models appear Replacing a specialist requires rebuilding and recalibrating the document database. Most specialists are fixed to an exact version. While the machine is online, some build passes can still fetch a newer version of their specialist. ## Voidmill names any document it cannot accept Each refusal includes the reason, so a declined file is never silently lost. It reads many document formats and keeps tables and figures. It builds from what an organization already has: primary scientific literature, technical documents and slide presentations with their images and figures, spreadsheets and text files, email exports, and web pages. ## Voidmill checks a knowledge database before calling it finished On the same models, the same documents rebuild into the same database. We use Voidmill in our own work. Voidmill built its showcase, the scientific knowledge library, and held that database to these same checks. --- ## Questions about Voidmill ### What is Voidmill? Voidmill is a document search and answers tool. After installation, it builds your documents into one knowledge database on your own machine and answers with the matching passages and their sources. A different specialist model handles each pass of the build. The database holds passages, an exact-word index, meaning indexes, named things, classifications, summaries, and links between passages. ### Must the documents go to a hosted service? No. Voidmill processes documents and answers locally after its models are installed. It never sends documents off your machine, even when a build checks online for model updates. ### Can a failed search look like absent evidence? No. Voidmill distinguishes a failed search from a completed search that found nothing, and announces degraded runs. ### Do model replacements make Voidmill suitable for any field? No. Voidmill’s models are tuned for life-science and scientific literature; retuning for other fields is not promised. ### Do the build checks certify Voidmill’s answers? No. Voidmill’s build checks determine when a database is finished. Passing them does not mean each verdict is right: on its largest database, the scientific knowledge library, some off-topic questions still receive an “answer” verdict. ### Why separate document knowledge from memory? Because documents and memories follow different rules. Voidmill keeps every source document as it was, while Voidloom, the memory we use ourselves, strengthens memories that get used and lets others recede. ### What does Voidmill need to build a knowledge database? Building a knowledge database needs an NVIDIA graphics card with at least 8 GB of its own memory, 16 GB of system memory and 30 GB of disk space to spare, on Linux or on Windows through Windows’ built-in Linux environment. Voidmill does not yet run on AMD graphics cards or on Macs. A mill changes its stones. Voidmill’s specialists can be replaced as better models appear, with a rebuild. ## In private use. Not yet offered to others. ### What it does not do · October 2, 2026 - Models are currently tuned for life-science and scientific literature. Other fields would require retuning, which is not promised. - Image search does not yet reliably rank the exact figure first. For images, “answer” means the question is on-topic, not that the figure exists. Voidmill gives no verdict yet on searches by example image. Its showcase, the scientific knowledge library, contains text and no images. - On its largest database, the scientific knowledge library, some questions beyond the subject still receive an “answer” verdict where “no answer” would be correct. - Replacing a model requires rebuilding and recalibrating the database. - Inclusion in Voidstead, a home for AI agents, being built for your own machine, is planned. A separate offer may follow. - Building a knowledge database requires an NVIDIA graphics card. AMD graphics cards and Macs are not yet supported. - A build can still check online for model updates. Not every model is pinned to an exact version yet. We will email you if Voidmill becomes available to others.