Vision | Quantify Terminal

The problem

Information is not the problem.

Every terminal tells you what happened. The work that matters starts after that, and today it is still done by hand: what it affects, and what to do about it. Quantify carries all three, automatically and with provenance.

01 · What happenedA filing lands, guidance moves, a supplier warns, a rate decision prints, a headline accelerates.

02 · What it affectsWhich cell in the model, which assumption in the thesis, which exposure two suppliers deep, and whether it is already priced.

03 · What to doResize, hedge, review or exit, inside the mandate, with the reasoning and evidence recorded.

Any terminal

Reported, ranked, summarised

Does not know it is your position

Out of scope

Analyst, by hand

Read and re-read

A spreadsheet, a memory and an afternoon; stale by the next print

Rationale in a chat thread, mandate in someone's head, audit trail in a calendar

Quantify

Compiled into facts, claims and evidence with provenance

Propagated through the model, the thesis graph and the Portfolio Twin

Decided inside the constitution, recorded in the ledger, replayable

Search, summaries and research assistants are now baseline features of every terminal. They answer questions. They do not maintain state.

The investor's actual state, the assumptions inside a model, the health of a thesis, the exposure hidden two suppliers deep in a portfolio, still lives in spreadsheets, notes and memory. It goes stale the moment the world moves, and the work of bringing it back into line is done by hand, after the fact, under time pressure.

Quantify is the system that maintains that state. Public and private information compiles into structured financial knowledge with canonical identifiers, point-in-time timestamps and machine-readable rights. Change propagates through causal relationships into live models and machine-readable theses. Portfolio-level consequences are recomputed. Decisions are presented inside explicit mandates and human approval gates, and every step keeps its provenance.

It is not built by adding screens. The desktop terminal already carries the desk: live markets, research, portfolios, risk, derivatives, funds, macro and geopolitics. What sits underneath it is one shared event-to-decision system: one Data Fabric, one evidence graph, one model registry, one thesis graph, one Portfolio Twin, one Decision Ledger. A change at the bottom reaches a decision at the top with its sources intact.

Win on speed of reasoning, memory, reproducibility and personalised portfolio impact. Not on the number of tabs.

Sequence

Built in this order.

Each stage has an exit criterion. Nothing downstream is built on a foundation that has not passed its own, and the first production slice proves the whole thesis on one market and one sector with real users.

Data Fabric and rights. Canonical identifiers, universal fields, source records, point-in-time versioning, entitlements and raw storage. Exit: every terminal field passes through one canonical read API.
World Model and Pulse. Entity edges, event schema, signal pipeline, personalised ranking. Exit: portfolio and watchlist users receive explainable, ranked changes.
Deep Research and Memory. Evidence graph, deterministic finance tools, notebook, verifier. Exit: a research report is reproducible and saved as structured memory.
Model Compiler and thesis graph. Model registry, cell-level lineage, assumptions, health. Exit: one sector's models update from earnings with analyst diff approval.
Portfolio Digital Twin. Hidden exposure graph, scenarios, failure search. Exit: a portfolio answers dependency and failure-path questions.
Workflow Engine and Decision Ledger. Plans, Investment Constitution, approval gates, replay. Exit: a natural-language objective executes as an auditable multi-step workflow.
Investor Chat. Rooms, live financial objects, gated data rooms, track records. Exit: private and public collaboration with permission-aware retrieval.
Financial Git and Research CI/CD. Dependency invalidation, financial unit tests, research QA. Exit: an amended source automatically marks downstream work stale.
Institutional hardening. Private Brain connectors, enterprise auth and audit, Agent Runtime, private deployment. Exit: a fund's research team measures analyst hours saved.

Standards

Measures it is held to.

Success is an analyst saying: this used to take several hours after earnings; Quantify updated the model, showed exactly what changed in the thesis, and quantified the portfolio impact in minutes. These are the engineering targets underneath that sentence, each tested against golden sets before a capability is enabled.

Measure	Target	How it is tested
Analyst time after earnings	Hours become minutes	The model updates, the thesis re-scores and the portfolio impact is quantified before a spreadsheet is opened. Measured per analyst against the same workflow done by hand.
Citation coverage	≥ 95% of material claims	Every factual claim links to the evidence span that supports it. Below the bar, the output returns to the verifier instead of shipping as a report.
Model compiler precision	≥ 99% on reviewed cells	Compiler output is scored against analyst-reviewed golden models by sector. Automatic proposals are enabled for a sector only once it clears the bar there.
Permission leakage	0	Restricted rooms and documents are attacked continuously by automated ACL tests at retrieval and after retrieval. A single leak blocks a release.
Point-in-time leakage	0 future-data reads	Backtests and historical research are audited for any access to data unavailable at the time, including restated fundamentals.
Workflow reproducibility	Same snapshot, same result	Dataset versions, source IDs, query plan, calculation, model and prompt versions and permissions are stored with every report. Reproduce regenerates it exactly.
Market-stream availability	99.9% at launch	Raised with redundant feeds. Canonical quote freshness is alerted on p95 lag against the vendor contract.
Incident tracing	100% of external calls	Every external call carries a request and correlation ID, so any number on the desk can be traced to the call that produced it.

Discipline

Failure modes designed against.

The ways a system like this goes wrong are known in advance. Each one has a structural answer in the architecture rather than a policy.

Feature count as strategyShip connected state transitions, not dozens of disconnected tabs.
Arithmetic in language modelsAll financial calculation runs in deterministic, versioned code. Models plan, interpret and explain.
Magic scoresEvery score exposes its components, inputs, uncertainty and evidence.
Unlicensed dataA rights matrix and entitlement enforcement precede any external launch. A cheap retail plan is not a redistribution licence.

Trading by accidentExplicit human and broker gates on every live order, with limits and a kill switch.
Vector leakagePermission filters at retrieval time and again after retrieval. Embedding access is data access.
Look-aheadavailable_at is mandatory throughout research and backtests.
Agent theatreIndependent agents must use different evidence paths and tools, and their disagreement must be visible and measurable.

Leadership

Founder-led, engineering-first.

Quantify Terminal is led by founder and Chief Executive Officer Aaryan Saroha, who directs architecture and product across the Data Fabric, the finance engine, the desktop terminal and the systems that connect them. The company is building institutional financial infrastructure with engineering across quantitative research, data infrastructure and desktop systems.

Leadership Join the Team

Get started

Run the terminal.

Install the desktop build for macOS, Windows or Linux, or contact us for institutional evaluations, partnerships and platform enquiries.

Downloads Contact Us