Saentinel

Saentinel Trader · Active paper evaluation

Bounded AI.
Deterministic risk.

A focused intraday trading research system for US stocks and ETFs. Market data finds the candidates. One constrained AI decision evaluates them. Deterministic controls decide what may actually happen.

Explore the system See current research
  • Paper first
  • Long only
  • Flat by close
09:40 ET DECISION CYCLE PAPER
01RankDeterministic
02AdviseOne AI call
03ControlRules decide
OUTCOMENo recommendation can bypass risk controls.

A clean restart

Built to trade less—and know more.

Saentinel Trader replaces an overbuilt investment-research platform with a deliberately small operating loop. It scans liquid US equities, records the evidence behind every cycle, asks one model for bounded buy-or-abstain advice, and keeps position sizing, risk, execution, monitoring and exits outside the model.

The operating model

Intelligence where judgment helps.
Code where rules must hold.

MARKET DATA

Observe broadly

Alpaca assets, snapshots, daily history and completed intraday bars create a time-bounded view of tradable US stocks and ETFs.

DETERMINISTIC SCANNER

Reduce the universe

Liquidity, momentum, spread, history and intraday context rank candidates before AI sees anything. Missing or invalid inputs are excluded, never invented.

BOUNDED AI

Recommend—or abstain

At most one configured model call evaluates a bounded shortlist. Strict JSON validation applies. Zero recommendations means zero buys; there is no fallback or automatic retry.

DETERMINISTIC EXECUTION

Control every action

Fresh quotes, position capacity, allocation ceilings, risk budgets, daily-loss state and ownership checks independently constrain every paper order.

One bounded cycle

From market scan to a recorded outcome.

Each cycle is claimed once and persisted. Restarts cannot repeat the AI request, stale results cannot enter, and execution rechecks the world before acting.

  1. 01

    Scan

    Filter and rank eligible US stocks and ETFs from current market observations.

  2. 02

    Enrich

    Add completed daily and intraday context, with explicit rejection of stale or incomplete evidence.

  3. 03

    Advise

    Send a bounded shortlist to one OpenAI-compatible model and accept only a complete, validated response.

  4. 04

    Constrain

    Apply deterministic capacity, capital, risk, spread, timing and asset-eligibility rules.

  5. 05

    Monitor and exit

    Manage owned positions through stop, take-profit, thesis, optional stagnation, daily-risk and end-of-day rules.

  6. 06

    Journal

    Persist scans, advice, token use, orders, fills, ownership and decision provenance in SQLite.

Risk before activity

The model proposes.
The system disposes.

AI is never the authority for capital, order safety or exits. Those decisions remain explicit, testable and restart-safe.

01

Allocation is a ceiling

An AI-suggested allocation can be reduced by deterministic controls. It can never enlarge itself into more exposure.

02

Risk reacts without chasing

Recovery mode reduces size and demands higher confidence. A hard daily-loss stop blocks new entries and flattens owned exposure.

03

Ownership stays exact

Saentinel monitors and exits only quantities it can prove it owns. Manual positions remain outside its authority.

04

No overnight positions

New entries stop before the close. Entry orders are canceled and owned positions are flattened within a bounded liquidation window.

Current paper research

The system works.
The strategy must still earn trust.

Forward paper runs exposed an important reality: safe software is not the same thing as a successful strategy. Saentinel is now using its own evidence to study where losses occur and improve only through explicit, deterministic changes.

OBSERVING

Earlier market context

Opening and one-minute telemetry record what candidates and SPY were doing before normal intraday history became available.

VALIDATING

One 09:40 opening decision

A single bounded exception evaluates the persisted 09:35 candidate pool—without adding a second call, backfilling missing data or weakening normal controls.

IN DEVELOPMENT

Actionable market regime

SPY conditions are being converted from observational labels into deterministic entry-risk controls for unfavorable tape.

IN DEVELOPMENT

Protect gains from reversal

A restart-safe rule is being designed to prevent positions that became meaningful winners from later turning into full losses.

What failure changed

No victory lap.
No hidden rewrite.

“A system can execute correctly and still fail its investment objective.”

That is now part of Saentinel’s design discipline.

The original Saentinel became too complex to prove useful. The replacement deliberately narrows the problem, preserves real paper evidence, and treats disappointing outcomes as inputs—not as results to explain away. New controls are developed from observed failure modes, while profitability remains unproven.

The honest boundary

A research system.
Not a promise.

Follow the experiment

Built in the open spirit of evidence.

Saentinel Trader is private software under active paper evaluation. Questions about the project are welcome.

contact@saentinel.com