Thoughtlines
An interactive atlas of philosophy, built to follow how ideas take shape, change, and connect across thinkers and time.
v0.10.123 · Alpha · Reviewed 14 October 2026This documentation describes the product, its research method and its current implementation. It distinguishes available features, working prototypes and planned capabilities.
Purpose and research goals
Thoughtlines represents the history of ideas through concise, sourced treatments: accounts of what a thinker contributes to a particular idea. General connections locate a topic; treatments make its distinctive claims, context and development available for investigation.
The project also evaluates a collaborative research method. People set direction and editorial standards, repeatable software maintains knowledge and workflow state, and Codex contributes source interpretation, application development and writing. Each investigation should leave reusable records that improve subsequent work.
Hegelian and Deleuzian perspectives inform optional investigations of transformation, difference and branching across the wider history of ideas. They do not limit the atlas to those philosophers or establish historical influence.
Purpose, scope and related approaches explains this mixed-initiative research environment in more detail.
Explore the atlas
Start anywhere
Search for a person, idea, work, or treatment. Follow nearby records, inspect a source link, and move between local detail and the wider graph.
Change the view
Use a free network, type lanes, or a timeline. Filter which connections appear and open compact groups as you need more detail.
Browse collections
Explore a question such as the Trinity within the wider atlas, or read its research history and planned source batches.
Read a story
Follow a comparative or chronological article with an interactive companion graph. Inspect the accounts and sources behind each section.
What is in the graph?
People, ideas, works, and treatments are all navigable records. A treatment is an account of a thinker’s particular approach to an idea—for example, “Hume on causation.” It connects the thinker and idea while leaving room for the thinker’s distinctive claim, its date, and its sources.
General ideas remain useful bridges between treatments. Some large treatments also have smaller component accounts. Connections have different meanings: authorship, topic membership, a source association, and a documented response should not be read as the same kind of relationship.
Research collections
A collection gathers work around a question without removing those records from the larger atlas. The current pilot follows ideas around the Trinity, including earlier philosophical backgrounds, Christian accounts, and later interpretations. Its membership is a research aid, not a claim that every member forms one continuous historical lineage.
Collection work is organized into bounded runs. A run starts with chosen seeds, selects nearby research targets, fixes a source worklist, and records what was found, added, or saved as a draft for another topic. New runs can broaden an initial collection, deepen a small set of records, or resolve missing dates and sources.
The collection dashboard shows plans, saved results, and reuse opportunities. Sourced records that pass publication checks can enter the atlas before individual human review; reviewers can mark uncertain records and reversibly remove them from active views.
Research with language models
A second goal of Thoughtlines is to test whether language models can help build this atlas responsibly and efficiently. The research task is bounded: read a selected set of relevant source pages, identify several useful thinker–idea accounts, draft short summaries with dates and precise source links, and save adjacent findings for later collections.
- Choose and acquire sources. A reproducible plan sets the pages and limits. Scripts keep copies and record failures and versions.
- Draft candidates. A language model may help interpret the text and propose treatments, components, or nearby ideas. Each proposal keeps its source and uncertainty.
- Check, publish, and revisit. Structural, identity, source and date checks gate publication. Research judgment routes valid records into the atlas or reusable draft library. Human review can follow afterward, with reasons and reversible pruning for questionable records.
Evaluation considers source support, date accuracy, duplicate work, useful records per page, time, and cost. Research efficiency is assessed alongside source support and the ability to inspect and correct an account. Ordinary atlas browsing does not require a live model call.
Planning, saved research decisions, draft reuse, validation and review tools are implemented. Interpretation currently runs through Codex collaboration; a single unattended source-to-graph controller remains planned.
Stories and interpretation
Stories provide authored routes through saved research. The selected-anchor prototype compares Augustine, Aquinas and Hegel; the temporal prototype follows a broader development of Trinitarian questions. Both retain source-linked accounts in an interactive companion.
Research accumulation and narrative synthesis are separate stages. Story writing selects and interprets existing material; it does not silently initiate new research or turn editorial transitions into graph relationships.
Development priorities
Populate a collection
Complete bounded Trinity research runs, improve source coverage, and reuse drafts found while reading adjacent topics.
Explain routes
Help a reader move between accounts, compare alternatives, and inspect why a suggested path through the graph is worth following.
Evaluate research assistance
Measure where language models save effort, where they need stronger checks, and which tasks are best kept deterministic or editorial.
Share a public edition
Explore a periodically published, read-only snapshot of the atlas. Research editing and source acquisition remain in the local workspace.
How to read the evidence
Thoughtlines combines imported structure, reference works, and locally researched accounts. The InPhO graph helps locate promising neighborhoods; reference articles such as the Stanford Encyclopedia of Philosophy (SEP) and Internet Encyclopedia of Philosophy (IEP) help explain them. Source associations are starting points. A treatment should provide its own support, date or qualified dating note, and a link back to a discussion readers can inspect.
Historical relationships, editorial groupings, and computed navigation aids remain distinct. Missing records are gaps in this atlas, not evidence that a connection did not exist. Dates may refer to a work’s publication or an approximate period; the record should say which.
Use this guide for product concepts and operating principles. The graph model and workflow guides describe the current behavior in more detail; the data architecture page provides an optional technical reference. Repository contracts and change logs preserve implementation-level decisions.