MARKET RESEARCH WORKSTATION · BUILT WITH OVERLIFT

Saros.ai

See what was knowable when a decision was made.

Investigate markets using the sources and information available at a chosen point in time. This case study shows how Saros keeps revisions, assumptions, and calculations attached to an analysis so it can be reviewed later.

See the research workflow See project screenshots

Saros.ai source registry with governed market, macroeconomic, and evidence providers — **Source registry** The provider registry distinguishes executable adapters, captured fixtures, policy-gated sources, evidence rights, and citation-preserving headline routes.

Recorded project scope

Implemented

Browser research workstation, governed source registry, replay, and private research memory

More in this category (2)

Provider-neutral evidence contracts, local vector surfaces, and deterministic analytical functions
Portable state and evidence receipts

Demonstrated

As-of research workflows using captured and labeled fixtures

More in this category (2)

Source rights, citation preservation, provider disagreement, and revision visibility
Market charting connected to evidence, scenarios, and deterministic tools

Planned

Broader live-provider promotion with credentials, rights, and freshness checks

More in this category (2)

Tauri desktop packaging and local Qwen-family inference
Microsoft MiniLM semantic bridge and expanded offline vector indexing

Research

Model-assisted thesis generation constrained by deterministic quant proof

More in this category (2)

Quantum-inspired scenario exploration across uncertain market states
Causal receipts that explain which evidence changed an analytical conclusion

MARKETING

Who it helps and why.

Market research is easy to rewrite by accident. Prices are adjusted, economic data is revised, filings arrive later, and headlines disappear behind new information. A chart made today can quietly use facts that were not available at the moment the original decision was made.

What makes it useful: Most market tools mix current knowledge with historical analysis and make hindsight leakage hard to detect. Saros freezes the as-of evidence, keeps semantic organization separate from deterministic math, and makes a thesis replayable so later conclusions can be compared honestly.

BUSINESS / EXECUTIVE

Business value.

Saros is designed to help an analyst revisit a thesis with its original sources, assumptions, and calculations intact. That supports review and comparison; it is not evidence of investment returns or predictive advantage.

Explore the business outcomes

More defensible research
Faster thesis testing
Clearer provider and evidence accountability

ARCHITECTURE

How the system fits together.

Embeddings, classifiers, hybrid retrieval, reranking, and model-assisted reasoning organize filings, macro releases, news, events, and research notes. Deterministic C17/WebAssembly functions own chronology, feature construction, event studies, regressions, risk, scenarios, state hashes, and tool policies.

A representative workflow

01
Choose an instrument, a research question, and the point in time the analysis should represent.
02
Inspect permitted sources and their revision dates, then compare calculations against that recorded state.
03
Save the sources, assumptions, and result in a receipt that another reviewer can revisit.

Explore the architecture details

I built Saros.ai as an offline-first quantitative research workstation. It combines governed provider adapters, as-of evidence, local vector memory, market and macro research surfaces, deterministic econometrics, scenario tools, replay, and portable receipts that preserve the sources and assumptions behind a result.

ENGINEERING

Engineering choices.

I made the time a fact became knowable part of its identity. Retrieval can organize the research, while deterministic code owns the calculations. The additional source and revision bookkeeping lets a reviewer inspect a thesis without silently replacing its historical evidence.

Explore the technology and engineering details

Technology stack.NET 10, C17/WebAssembly, PostgreSQL, pgvector, hybrid RAG, ONNX Runtime Web, deterministic econometrics

As-of market replay tied to one evidence clock
Hybrid RAG and local vector memory for source-grounded research
Deterministic quantitative functions with portable receipts

OVERLIFT METHOD

How this product applies the OverLift method.

The OverLift method freezes the as-of evidence, separates semantic organization from deterministic math, explores scenarios without rewriting history, and saves a receipt that makes the thesis reproducible. A later result can be compared honestly with what was knowable earlier.

Explore the shared OverLift architecture

The method is reusable. Each product supplies its own facts, rules, execution, and validation. These tiers describe the shared architecture; the project scope above records which capabilities are implemented here.

Why determinism matters here: Research is invalid when later information leaks backward into the decision. Determinism preserves the selected time boundary, calculations, evidence, and thesis path.

T0
Exact baseline
Define the authoritative result before adding acceleration, semantics, or agents.
T1
Measured evidence
Instrument state, latency, quality, and failure so improvement is observable.
T2
Bounded ownership
Move expensive or high-frequency work behind clear worker and memory boundaries.
T3
Deterministic transport
Version messages, snapshots, identities, deadlines, and replay inputs.
T4
Semantic assistance
Let local retrieval and models propose meaning only inside admitted evidence.
T5
Candidate paths
Compare several allowed next actions with explicit risk, cost, fit, and uncertainty.
T6
Shadow verification
Challenge the production route without giving the challenger authority.
T7
Governed production
Promote only measured, replayable paths with fallback and operational receipts.
Tier-S
Experimental challenger
Run advanced evaluators in shadow, compare them with the control, and graduate only repeatable gains.

CATS

Context-Aware Tier Selection

CATS chooses a qualified execution tier for the evidence, device, latency, permissions, and failure conditions of the task.

SEMANTIC BRIDGE

Meaning becomes typed product state

The Semantic Bridge maps a goal to explicit entities, evidence, uncertainty, tools, and permissions. Missing or contradictory information remains visible.

DETERMINISM

Validate the action and retain the result

Models may retrieve, compare, or propose. Exact software checks facts, permissions, state changes, and execution, then records what happened.

Read the OverLift MethodOverLift.ai See other products

PROOF

Evidence and limits.

The case-study screens show the research workstation, source registry, quantitative tools, and replay workflow. Provider modes and evidence boundaries explain which inputs belong to a result. This portfolio presents the project and its recorded scope; it does not run a new provider-data or investment-performance evaluation.

Back to the project overview ↑

Discuss Saros.ai View resume All projects