Connected research
Keep local experiments, their inputs and recorded results together in Research Lab
TradingOS
The intelligence behind every trading decision
TradingOS is being built to connect market research, specialist AI analysis, quantitative models and controlled execution in one trading platform
Free registration for future TradingOS access
The product is in development and registration does not unlock trading software today
Full product launch currently targeted for February 2028 to May 2028
Already registered? Sign inThe approach
The goal is to investigate an opportunity from the first signal to the final outcome, bringing together the evidence behind a trade, its uncertainty, its effect on the portfolio and the conditions required to act
Understand what supports a trade, what could invalidate it and how the decision changes as the market moves
From research to decision
Planned workflow
A historical strategy signal is one possible starting point. This map describes the intended connected process; it is not a recorded trade or a completed replay
- Observe an opportunity
Identify the candidate strategy and retain the data with the time it became available
- Examine the case
Bring together supporting evidence, counterarguments and any model output with its horizon and uncertainty
- Compare and decide
Consider other candidates, costs, liquidity, exposure and permissions before proposing action, waiting or rejection
- Review what happened
Link a later outcome to the original evidence without changing what was known then
Capabilities
Local capabilities are not public product access or confirmation of live trading readiness
Implemented locally
Strategy validation
Test a research idea with temporal holdout and walk forward, keeping failed and inconclusive results visible
Reproducible research
Revisit a local experiment using its recorded strategy version, configuration and dataset references
Execution awareness
Inspect configured spread, slippage, commission and lagged volume limits in historical simulation
In development
Decision evidence
Connect source quality, candidate assessments and reasons to the same decision; the complete operational chain is still being integrated
Portfolio context
Bring exposure and available capital into candidate assessment; the current account to allocation connection remains incomplete
Opportunity selection
Compare eligible candidates and retain reasons to act or wait as the decision and capital services are connected
Controlled execution
Build on deterministic permissions, risk checks and order controls while the operational execution path is validated
Decision history
Link what was known at a decision to its later review; local evidence stores exist, while the complete user workflow is still being connected
Planned
Specialist AI analysis
Explore an idea through defined research, review and explanation roles with evidence you can inspect
Research
Probabilistic forecasts
Inspect horizon specific model outputs and calibration in bounded research; trading value has not been established
Validated improvement
Evaluate frozen model versions on separated data before considering a change; experiments do not grant execution permission
AI analysis
TradingOS is being designed around specialist analytical roles that examine an opportunity from different perspectives and connect their findings to the same decision record
Evidence and interpretation
Inputs
Market structure, news, forecast results, strategy evidence and portfolio exposure each answer a different question and the system is intended to make agreement, disagreement and missing information visible
Responsibilities
Quantitative models calculate measurable outputs while AI helps investigate context and explain the evidence within defined permissions
Models and controls
Quantitative models
Implemented locallyThe local backtest and indicator engines calculate measurements. Forecast models are a separate research programme
AI agents
PlannedThe documented roles below would use restricted research and citation tools. They are not confirmed as an integrated agent team
Permissions, risk and orders
In developmentDeterministic controls own permissions and the order path. AI output cannot bypass them or promote an experimental strategy to live trading
Structured AI shadow review
In developmentA bounded AI integration can return a thesis, invalidation conditions, evidence references and uncertainty. Its limited provider proof does not establish a connected decision workflow or autonomous tool use
Planned agent roles
These are the roles specified for the future research environment. Each produces an advisory result or a research draft; none can submit orders, change risk limits or approve its own work. Multiple roles do not establish independent judgement or better accuracy
Sources and hypotheses
Source Curator
Reviews approved source leads and provenance, contributing a source recommendation with citations or a quarantine reason
Knowledge Extraction Agent
Extracts draft claims and contradictions from approved documents, contributing references to the exact source passages
Hypothesis Agent
Develops falsifiable research hypotheses, contributing assumptions, a proposed test and rejection criteria
Strategy and validation
Strategy Engineer
Drafts strategy code and tests in a research sandbox, contributing a reviewable draft and its limitations
Bias Auditor
Examines leakage, overfitting and selection bias, contributing a report that preserves adverse findings
Experiment Agent
Coordinates an approved research job, contributing versioned attempts, results and failure diagnostics
Portfolio and explanation
Portfolio Analyst
Interprets authorized portfolio projections, contributing exposure and scenario reports with nonexecutable recommendations
Performance Supervisor
Reviews performance and health projections, contributing anomaly reports and recommendations for human review
Explanation Agent
Explains authorized records, contributing a cited account of facts, interpretations, uncertainty and missing evidence
Quantitative tools
Established indicators and deterministic simulation are implemented locally for comparing ideas. Their value depends on the data, validation and how the results are used
Read the market
Trend, momentum and volatility
SMA and EMA measure trend, RSI measures momentum and ATR measures volatility. These established indicators are implemented in the local deterministic strategy and backtest framework
Price references and ranges
VWAP provides a volume weighted price reference. Highest and Lowest identify rolling range levels. Their interpretation depends on the source data and chosen window
Test and compare strategies
Deterministic backtesting
Historical OHLCV runs preserve resolved configurations and fingerprints for exact reruns on the same inputs. Configured transaction costs and volume based liquidity assumptions can be inspected; order book depth, queue position and real broker fills are not reconstructed
Strategy Ensemble
Local strategy versioning and immutable evidence support contextual routing between eligible strategy versions. The service uses supplied competence evidence; it does not discover a profitable strategy or prove an operational edge
Record and validate the evidence
Research Lab
Versioned experiments record configuration, dataset references, attempts and measurements. Completed, failed and cancelled runs remain distinct from a successful validation
Robust Validation
Predeclared temporal holdout and walk forward plans separate development from evaluation. PASS, FAIL and INCONCLUSIVE distinguish a passed check, a failed check and evidence that cannot support a conclusion
Forecasting research
Local model pipeline
The local model training pipeline supports classification and regression, calibration and out of sample evaluation. One bounded frozen direction model has been exercised with a defined short horizon and recorded outcomes; it remains a research observation before costs, not a trade recommendation
What is being evaluated
The research direction is to study up/down probabilities, expected return, move size and uncertainty. These outputs still need evaluation against transaction costs. Only information available at the forecast time should enter the comparison, with separated evaluation data, walk forward and strong reference models
Implementation and calibration do not establish predictive advantage. No profitability or accuracy claim is made for these models
Development direction
Market understanding
Interpret the conditions in which an idea may work
- Regime, liquidity and market structure
In development
Connect contextual assessments and source quality to the decision
- Cross asset relationships, crowding and event shocks
Research
Investigate shared exposures and conditions that may undermine a candidate
Forecasting and research
Turn an idea into a testable comparison
- Research Lab, Strategy Ensemble, Robust Validation
Implemented locally
Local tools retain experiments, routing evidence and validation outcomes
- Probabilistic forecasting
Research
Evaluate horizon, calibration and uncertainty against separated data and reference models
Decision and capital
Consider which opportunity deserves capital
- Candidate selection and capital allocation
In development
Connect candidate comparisons, portfolio context and hard risk constraints; current account integration is incomplete
Execution and supervision
Keep the route from proposal to order inspectable
- Order lifecycle and execution controls
In development
Validate executable market state, execution risk and position monitoring within explicit permissions
- Shadow validation
Research
Observe research forecasts and later outcomes without giving the model order authority
Learning and change assurance
Understand what should be tested next
- Outcome analysis and model evaluation
Research
Compare observed outcomes and candidate models before considering a change
- Adversarial AI review
Planned
Use the Bias Auditor role to surface counterarguments and unresolved findings for human review
Product FAQ
What is TradingOS
TradingOS is DeVcore’s flagship product in development: a platform intended to connect market research, specialist AI analysis, quantitative models, risk and controlled execution around a traceable trading decision
Can I use the platform today
The trading platform is not publicly available. Local implementations and research trials are part of development; joining the waitlist does not unlock trading software today
What does joining the waitlist mean
You request a free registration and confirm it through the email link. It lets you follow development and be informed about future product access. It does not activate a subscription, guarantee a test place or set an invitation date
Which features are built and which are planned
Features are grouped or labelled by their implementation status. Implemented locally means an implemented local capability, with the scope described beside it. In development, Planned and Research describe unfinished integration, documented future roles and experimental work. None of these labels promises public or live trading access
Does TradingOS guarantee trading results
No. Models, AI analysis and historical tests do not guarantee future results. Forecasts remain uncertain, and no proven trading advantage is claimed for the research described here
Early access
Join the free waitlist to follow development and be informed when product access becomes available
Waitlist registration does not give access to the trading platform today or guarantee a place in a limited test
Founding Access
Founding Access reserves the applicable Founding price, opens the private Hub and is credited toward the first subscription invoice. The subscription does not start today
Market data, exchange fees, broker fees and optional third party services are not included unless explicitly stated
What happens next
Your reservation
- Verify your email and choose an Individual, Team or Company account
- Review your plan and explicitly accept the applicable conditions
- Pay the one time reservation through Stripe Checkout. Once payment is confirmed, your Founding identifier and Hub access are activated
Your referrals
Verified direct referrals add to Access Score. Account groups rank separately, and rank shows relative priority rather than a guaranteed date
Founding Hub
The Founding Hub contains selected updates, progress and Labs. The current launch target is February 2028 to May 2028
View the Founding RegistryFounding FAQ
What do I pay today?
Only the one time reservation shown for your plan. The confirmation page sets out the amount and applicable terms before Checkout
When does the subscription start?
After product launch and activation, not with the reservation. The reservation is credited toward the first invoice
How does the Founding price work?
Individual Founding pricing continues while the qualifying subscription remains uninterrupted after launch. Team and Company protection follows the plan and the terms confirmed at purchase
Can Teams and Companies join?
Yes. The organization owns the account, its shared Access Score and the agreed seat allowance
How do referrals work?
Verified direct referrals add to Access Score. They do not create a commission, and a refunded purchase loses its paid bonus
What is not included?
Market data, exchange and broker fees are excluded unless stated. TradingOS is software and does not guarantee trading results
How will Founding customers receive updates?
Through the Founding Hub, with selected build updates, a progress ledger, product decisions and Labs. Important account information is sent by email