Guide

How Emesdos works. None of the vocabulary below is required to use it.

Just talk — journals catch everything

You speak in claims, stakes, beliefs, observations, decisions, and facts. The system types them. Nothing you say is ever lost for lack of typing: a sentence that matches nothing still lands as a journal entry with tags, and structure is always offered, never imposed — every typed write below is confirmed before it happens. The gradient runs:

journal/note — always succeeds
tagged — automatic
typed at capture — detection, always confirm-gated
typed retrospectively — proposals, propose-gated
linked, composed, watched — the living loops

Collections — storage

what it is: What a thing is. Journals, notes, sources, trackers, projects, artifacts, media — every row lives in exactly one collection.
use when: Always, without choosing: speech lands somewhere by construction. The journal entry is the universal, zero-ontology fallback — nothing you say is lost for lack of typing.
what the system does: A sentence that matches no detection case still lands as a journal entry with tags, and the promotion machinery carries untyped content upward later — into notes, trackers, and compositions, each step offered, never imposed.
what you might say: “Today was rough — the deploy slipped again and I skipped the run.” (a journal entry, no ceremony)

Tags — retrieval axes

what it is: Orthogonal lookup across everything, in a small set of namespaces — topics, entities, signals, operations, hub references.
use when: You never file anything: the capture skills tag what you say as they write it, and the lint keeps each axis canonical.
what the system does: Tags accrete automatically and become the pages you browse — topic pages, entity pages, signal queries — without you maintaining a taxonomy.
what you might say: “More gold chatter in the macro press today.” (tagged with its topics on the way in — no filing on your part)

Projects — lifecycle membranes

what it is: Bounded working contexts with open/close semantics: a project opens, holds work, and closes with a result.
use when: Never required on day one — projects emerge when work clusters into something with a beginning and an end.
what the system does: Rows can join a project for its lifetime; closing one sweeps the loose ends into a retrospective, and the result outlives the project.
what you might say: “This kitchen thing has become a whole project — let's treat it like one.”

Links — typed relations

what it is: The declared fourth layer: typed edges between rows — evidence, lineage, supersession, contestation.
use when: Skills write them as structure is detected; you almost never ask for a link by name.
what the system does: Invisible until they render — citations as marginalia, graded evidence under a belief's gauge, lineage on the worldview roll-up.
what you might say: “That filing supports the thesis.” (a graded evidence edge, written for you)

The kinds — what a tracker can be

A tracker is a time-evolving structured object: it opens, evolves, and may close. Seven kinds cover any domain where that shape fits — you never pick one; capture proposes the kind from what you said.

`hypothesis`

what it is: A predictive claim of yours under test — a falsifiable bet with a win condition, a loss condition, and a horizon.
use when: You hold a dated, testable belief: a financial thesis, a scientific prediction, a design bet, an expected market move. Your own claim — not someone else's (that's a stance) and not a standing generalization (that's a tenet).
what the system does: Evidence accrues against it over time — observations and cited sources, each graded for provenance and credibility — and the composition function rolls them into a confidence. Reviews resurface it as the horizon approaches; it closes as succeeded, failed, or aborted, and the outcome stays on the record.
what you might say: “I think AAPL outperforms SPY by year end.”“If the SPR draw continues, crude spikes within a month.”

`position`

what it is: A commitment with stakes — capital, time, or attention staked on an outcome, actively executed or passively held.
use when: You hold or run something under a thesis: a trade, a buy-and-hold, real estate, a collectible held with intent — or a non-financial execution like a candidate in a hiring loop or a negotiation. Perpetual holds with no planned exit are valid.
what the system does: Pairs with the hypothesis it derives from, carries entry/sizing/exit framing, and accrues valuation snapshots over time — the worldview page rolls open positions up under the beliefs they stake. Most close on exit; the result narrative records how it went.
what you might say: “Bought 200 shares of AAPL at 195 for the Q4 thesis.”“We're going ahead with the renovation — budget is committed.”

`goal`

what it is: A desired outcome with effort directed at it — a target with a definition of done and usually a deadline.
use when: You're working toward something measurable: a fitness target, a learning objective, a certification. Pick goal over position when the emphasis is the target, not the stake.
what the system does: The success criterion is lifted into the headline so progress is legible at a glance; the result field carries the running narrative, and reviews resurface it as the deadline nears. Closes as succeeded or failed against its own definition of done.
what you might say: “I want to run a sub-25 5K by August.”“I'm going to get the certification done before Q3.”

`monitor`

what it is: A perpetual observation with no attempt to influence — a quantity you watch, asserting nothing.
use when: You want a feed watched on a cadence: SPR levels, an exchange rate, a price, a repo metric, weather patterns. The monitor stays neutral; your reads of it become hypotheses, stances, or tenet evidence downstream.
what the system does: Scope declares the source and cadence; the due-sweep fires checks on schedule, and readings land as dated observations. When you interpret a reading aloud, capture offers the interpretation as its own tracker, linked back to the monitor so the lineage is structural. It never resolves — retirement is its only end.
what you might say: “Keep an eye on the weekly EIA petroleum report for me.”“Track EUR/USD — I care when it moves past 1.15.”

`condition`

what it is: A perpetual situation under active management — ongoing, not aimed at a finish line.
use when: Something recurring needs managing rather than solving: a chronic illness, a recurring workplace pattern, an ongoing constraint. A monitor is passive watching; a condition is active management.
what the system does: Scope carries escalation triggers and a re-evaluation cadence — the system watches for the thresholds you set and prompts when they trip; the result field accumulates the management history. Rarely closes.
what you might say: “My back is acting up again — third time this month.”“The standup keeps overrunning; I'm going to start timeboxing it.”

`stance`

what it is: A third party's position, posture, or stated belief, tracked as an attribution — the holder is someone other than you.
use when: You're tracking what an external actor holds or believes: a fund's disclosed holding, a central bank's policy posture, an analyst's view. The split keeps their beliefs out of your own prediction queries, and vice versa.
what the system does: The headline names the actor and the attribution; scope sets the refresh cadence and sources; evidence accrues the same graded way as a hypothesis. Closes when the actor's posture verifiably changes.
what you might say: “Berkshire is still long AAPL — about $200B as of the last 13F.”“The Fed is signaling hawkish through Q1.”

`tenet`

what it is: A standing, regime-conditional generalization held and under test — never resolved, only more or less supported.
use when: You believe something general about how the world works — “gold hedges uncertainty risk, not inflation per se” — that should accumulate evidence for and against rather than settle. Dated bets it implies become hypotheses, linked underneath.
what the system does: Accrues two-sided evidence — confirms and refutes — and can decompose into regimes where the claim holds or inverts; the composition rolls both sides into one gauge. It never closes: supersession is the only removal. Theses cite it, and the drift watcher flags any thesis whose evidence has moved since it was written.
what you might say: “Gold protects you when things get scary — though it dumped in 2008.”“Good infra teams pay down migration debt early; it compounds either way.”

The relations — how things connect

Links are typed edges between rows, written by skills as structure is detected. The kinds a reader will actually meet:

`cites`

what it is: A plain citation — this note, journal entry, or composition rests on that source.
use when: Anything you write that draws on a source gets a cites edge; capture writes them when you mention where a claim came from.
what the system does: Citations render as marginalia on the reading surfaces and keep provenance walkable — from any claim you can get back to what it stood on.
what you might say: “Per the FT piece this morning, the fab expansion is delayed.”

`observed`

what it is: A dated reading or event recorded against a tracker — “this happened, and it bears on that.”
use when: A monitor check lands a reading; a day's events bear on an open goal or condition; anything time-stamped a tracker should remember.
what the system does: Observations against an evidence-bearing tracker carry a graded sidecar — provenance, strength, credibility, staleness — and a polarity. They are the raw series the composition function reads.
what you might say: “SPR came in at 278M today — below the threshold I was watching.”

`cited_evidence`

what it is: A graded evidence edge from a belief to what supports or undercuts it — every row carries provenance, strength, credibility, and staleness.
use when: A hypothesis, stance, or tenet rests on a specific source or reading; written by capture and the check skills when evidence is named.
what the system does: The composition function weighs each edge by its grades and polarity to produce the tracker's confidence; the citation-cleanliness membrane keeps position-bearing belief graphs free of unusable provenance.
what you might say: “The Q3 filing backs the margin-expansion thesis.”“That 2008 drawdown cuts against the gold tenet.”

`derives_from`

what it is: Structural lineage — this object exists because of that one.
use when: A position opens under a hypothesis; an interpretation springs from a monitor; a hypothesis is implied by a tenet.
what the system does: Lineage is what lets the worldview page roll the stack up — tenet to thesis to hypothesis to position — and what keeps an interpretation traceable to the feed that prompted it.
what you might say: “Opening the trade on the back of the AAPL thesis.”

`superseded_by`

what it is: Replacement-by-newer: the old row stays on the record, pointing at what replaced it.
use when: A belief is reframed rather than resolved; a note is rewritten; a tenet gives way to a sharper formulation. History is never deleted to make room.
what the system does: Surfaces walk the chain to show the current formulation while keeping every prior one reachable — supersession is the only removal verb for things that never close.
what you might say: “That framing was too broad — the real claim is about uncertainty, not inflation.”

`constituent_of`

what it is: Part-of structure between positions — a leg that belongs to a larger holding.
use when: A composite position decomposes: a pair trade's two legs, a property inside a portfolio position.
what the system does: The parent values itself through its constituents at read time, so a leg can't be deleted out from under the aggregate — the membrane blocks it.
what you might say: “The short leg is half of the pair trade, not its own bet.”

`regime_of`

what it is: A regime decomposition edge: a child tenet naming the conditions under which the parent claim holds — or inverts.
use when: A generalization is true only sometimes: gold rises while risk is unrealized, gets sold once risk is realized. Each regime becomes its own evidence-accruing child.
what the system does: Each regime child composes its own confidence from its own evidence, with a stance toward the parent — confirms or inverts; the parent's gauge combines both sides. A claim weighed both ways, never resolved.
what you might say: “It holds while fear is rising — once the crisis actually hits, it flips.”

`contests`

what it is: A declared tension between standing claims — two tenets or stances that cannot comfortably both hold.
use when: You notice two of your standing beliefs pull against each other, or an actor's posture contradicts one of your tenets.
what the system does: The tension renders on both sides, so neither claim is read in false isolation; reviews can surface contested pairs for deliberate resolution.
what you might say: “Hold on — my cash-is-trash tenet sits badly with the liquidity-premium one.”

`input`

what it is: The declared input set of a composed document — exactly which rows a thesis, briefing, summary, or essay was drafted from.
use when: Written automatically when the compose family drafts a document; the declared-inputs discipline is what keeps a composition honest about its sources.
what the system does: Each input edge to a belief stamps the confidence at draft time as a baseline — the drift watcher and the reading surfaces compare it against now, so a thesis quietly outgrown by its evidence gets flagged instead of silently going stale.
what you might say: “Draft the gold thesis from the tenet, the open position, and this quarter's observations.”

A handful of further kinds are internal plumbing — media containment, encoding variants, provenance breadcrumbs — and are deliberately not documented here.

Composing — frozen prose from declared inputs

The compose family drafts documents from rows you already have. Four prose forms, one discipline: declared inputs, citations, immutability.

`thesis`

what it is: A frozen prose argument — the case for a belief or holding as of the day it was written, drafted from declared inputs.
use when: A hypothesis, position, or tenet has accrued enough structure that you want the argument in one readable piece.
what the system does: The draft cites its inputs through typed edges and stamps each cited belief's confidence at draft time; from then on the drift watcher compares baseline to now and flags the document when the evidence moves. The document never silently updates — a new thesis supersedes the old.
what you might say: “Write up the AAPL case as it stands.”“Pull the gold tenet, the position, and the monitors into one argument.”

`briefing`

what it is: A situational summary on a topic — what's known and current, gathered into one piece.
use when: You need the state of play before a decision, a meeting, or a re-entry into a domain you've been away from.
what the system does: Drafted from declared inputs across notes, sources, trackers, and journals on the topic, with citations as marginalia; immutable once published, so the record shows what you knew when.
what you might say: “Brief me on where the semiconductor export-control situation stands.”

`summary`

what it is: A period or scope retrospective — what happened across a window or a project, in prose.
use when: A month, a quarter, or a project closes and the raw journals and reviews deserve a narrative pass.
what the system does: Pulls the period's rows in as declared inputs and freezes the retrospective; future reviews can cite it rather than re-deriving the period.
what you might say: “Summarize what happened across the renovation project.”“Write up Q2.”

`essay`

what it is: Open-ended exposition — thinking in prose, without a tracker or a period as its spine.
use when: An idea wants development beyond a note but isn't an argument for any one belief.
what the system does: Same discipline as the rest of the family: declared inputs, citations, immutability — exploratory writing with the same provenance as everything else.
what you might say: “I want to develop that thought about taste and tooling into something longer.”

Model reports — recurring models with history

A model-report member is a skill producing a recurring, subject-keyed, closed-form model run — each run immutable, chained to its predecessor, readable as a series. A member's first run on a new subject doubles as that subject's front door.

DCF valuation

what it is: A discounted-cash-flow valuation run on a ticker — base, bull, and bear values with the assumptions on record.
use when: You ask what a company is actually worth. The first run on a new ticker doubles as its front door: the entity card, relationship notes, and paired trackers are offered in the same confirmation.
what the system does: Every run is an immutable report; recomputes chain to their predecessor, so the runs form a series readable as the model's history on the entity page. The living state stays in paired trackers — the valuation claim as a hypothesis, the filings feed as a monitor that can trigger recomputes.
what you might say: “What's ASML actually worth?”“Re-run the AAPL valuation with the new guidance.”

The namespaces — how retrieval accretes

Tags ride along on everything written, in a small set of prefixed namespaces. These are the public ones:

`s/`

what it is: Signals — a closed vocabulary of moment-in-time markers: insight, pattern, decision, action, blocker, risk, and a few more.
use when: Never by hand on day one — capture marks the moments as they happen on the log surfaces (journals, reviews, source margins).
what the system does: Signals stay anchored to their moment — the surfaces that accept them are enforced, stable claims reject them — so “every decision made in March” is one query.
what you might say: “Let's go with the second vendor — decided.” (lands marked s/decision)

`o/`

what it is: Operation classes — a closed vocabulary naming what kind of work produced a row: capture, synthesis, review, valuation, research, and the rest.
use when: You don't — skills stamp their own operation class on what they write.
what the system does: Lets the vault be sliced by activity rather than topic: everything synthesis produced, everything the review cycle touched.
what you might say: “Show me everything the review cycle produced last month.” (a retrieval, not a filing)

`t/`

what it is: Topics — open, fuzzy-matched domain tags: what a row is about.
use when: Automatically — capture tags what you say, and near-duplicates fold together by fuzzy matching so the axis stays clean.
what the system does: Topic pages collect everything on a subject across collections; the homepage's topics column and the /topics/ index derive from them.
what you might say: “…anyway, the whole tokamak funding question is heating up.” (lands tagged t/fusion-energy)

`e/`

what it is: Entities — open canonical tags for named public things: companies, assets, places, products, public figures in their public capacity.
use when: Automatically, whenever a named entity comes up. Listed companies use the ticker as the canonical form (e/aapl); everything else uses the common name, hyphenated. One entity, one tag — other names, tickers, and acronyms live on the entity's card as identities.
what the system does: Every entity accrues a page: its card, its relationship notes, its trackers, its model-report history. Mention something often enough and the system proposes carding it; renames redirect to the new name, and near-duplicate spellings are caught at write time with the card's context so you can say same-or-new.
what you might say: “TSMC makes basically all of Apple's advanced chips.” (both e/ tags, and the relation offered as a note)

`f/`

what it is: Folgezettel references — pointers into the hub notes that give the note graph its topical entry points.
use when: When a row belongs under a hub's umbrella; the value must resolve to a real hub note, reference-checked.
what the system does: Hubs collect their referencing rows, so the note graph stays navigable from a small set of curated entry points (/h/).
what you might say: “File that under the systems-design hub.”

The loops — what runs without you

Evidence. Observations and citations accrue against beliefs as graded edges — provenance, strength, credibility, staleness, each reading confirming or refuting — and the composition function rolls them into a confidence that updates as the evidence does. The gauges you see on belief surfaces are that arithmetic, not an opinion.

Drift. Composing a document stamps each cited belief's confidence at draft time as a baseline. The alignment review compares baseline to now and flags documents whose ground has moved — a thesis can't quietly outlive its evidence.

Reviews. Periodic reviews and the due-sweep resurface trackers on their own declared cadence — monitors fire their checks, horizons approach, conditions re-evaluate. Nothing depends on you remembering to look.

The surfaces — where things render

The homepage carries the urgent strip, pinned compositions, and the ranked feed. Topics (/topics/) and entity pages collect everything on a subject; notes and their hubs (/h/) form the reading graph; compositions (/compositions/) and model reports are the published prose and runs; published predictions live at /hypotheses/. Operator-tier surfaces — the full tracker pages, the worldview roll-up of the belief stack — stay private; what's public is the showcase, never the raw vault.