Kestrel vs Trade-Ideas (Holly): AI Scanner vs Agentic-Trading Runtime

Answer card

Want an AI to scan the market, surface setups, and fire real-time alerts for a person to act on? Trade-Ideas (Holly) is a purpose-built, shipping tool. But if the trader is an LLM that must perceive a live market cheaply and act in milliseconds, that shape is wrong: it assumes a person watching a scanner and clicking. Kestrel is a language an agent authors, plus a runtime that renders markets as text and executes contingently. Different reader, different tool.

The two products are not competing for the same seat

Trade-Ideas is a scanner-and-alerts product for people. Its core artifact is a stream of surfaced opportunities: its AI, Holly, runs strategy simulations overnight and presents human traders with a curated set of setups and real-time alerts during the session. The author-and-actor is a human. The interface is a windowed desktop scanner with charts and alert feeds. The value is compressing a universe of tickers down to a shortlist a person can watch and trade discretionarily.

Kestrel starts from a different premise. The two failures of LLM trading are interface (perception and latency), not intelligence. A frontier model can reason about a market; it cannot see one through a screenshot without burning a context window, and it cannot act on one in the milliseconds a move demands. Those are interface problems, not IQ problems. Kestrel is the interface: a small language with four statement kinds over one lexical core, and a runtime that renders markets as text and executes contingently at machine speed.

• VIEW (perception): renders the market as a compact text Frame. The chart is in text: a vertical append-only tape, one row per candle, relative candles in basis points versus prior close, anchored by periodic keyframes. Perception cost is O(new bars), not O(screen), and the tape is KV-cache-friendly.
• WAKE (attention): a standing subscription over the trigger algebra. Event-driven, not polling. It spends attention (tokens, wakes), never risk. A Scan is a Wake over a whole universe, Kestrel's own analog to what a screener does.
• PLAN (latency): a standing, bounded-risk contingent program: trigger, actions, bracket, invalidation, TTL. The runtime fires it in milliseconds and wakes the agent in parallel. This is fire-then-inform: the agent is never in the hot path.
• GRADE (trust): the honest, counterfactual result of a run. Contamination-fenced: LLM authors are graded only on post-training-cutoff, date-blinded days, because weights leak what code fences can't stop.

The sharpest split is what happens after a setup is found. A scanner's job ends at the alert: a human still has to perceive the chart and place the order in time. In Kestrel, discovery is a Scan (a Wake over the universe); when it fires, the agent may author a new leaf (a Book with coverage, a thesis, a budget), and the PLAN it arms executes in milliseconds without the agent in the hot path. The alert is the beginning of a machine loop, not the end of a human one.

Where Trade-Ideas wins, plainly

This section is not a courtesy. If the work is any of the following, choose Trade-Ideas:

• A human is the trader, and the point is discovery for a person to act on. Trade-Ideas is built around a person watching a scanner and making discretionary calls. If your operator is a human who wants curated setups and alerts, that is exactly its seat, and Kestrel is the wrong direction.
• You want a shipping, polished scanner today. Trade-Ideas is a mature product with an installed base, a broad strategy library, and Holly's overnight simulations delivering a live shortlist. kestrel.markets is early (see Status, honestly below); if you need a working screener this afternoon, that is not a close call.
• Real-time alerting and charting UX for humans. The windowed desktop experience (charts, alert windows, filters a person tunes by hand) is designed for human eyes and attention. Kestrel renders to text for a model; it is not trying to be a better human dashboard.
• Discretionary trading where the human is the edge. If the judgment lives in the person and you just want the machine to narrow the field, a scanner is the right tool and an agent runtime is overkill.

If those describe the job, stop here. Trade-Ideas is the answer.

Where Kestrel is the fit

Kestrel is the fit when the trader is a model, the edge is judgment under live conditions, and the two interface failures (perception and latency) are what stand between the model and a market. Discovery is a Scan; the response is authored and armed by the agent.

IMPORT { fade-ladder } FROM "./armory/reversion.kestrel"
USING signal SPX exec SPY 0dte

WAKE universe-scan
  WHEN velocity(5m) >= p95 AND spot crosses above HOD
  DELIVER desk
  BUDGET 30 wakes/day

PLAN headline-chase budget 0.25R ttl +45m regime {intraday: trend}
  WHEN spot crosses above HOD AND velocity(5m) >= p95 held 120s
  DO buy 2 +1 C @ min(fair, mid) peg cap fair
  TP 2.5x frac 0.5 @ fair
  EXIT spot < VWAP @ fair
  INVALIDATE spot < HOD

This is illustrative, with generic instruments (SPX for signal, SPY 0dte for execution): not a strategy recommendation. It shows how an agent could express the pattern a scanner surfaces: the WAKE is the discovery Scan; the PLAN is the contingent response you deploy into your own pod and arm. The runtime holds the contingency and fires in milliseconds if the trigger hits, and the agent is informed in parallel. Prices are expressions, not constants: min(fair, mid) peg cap fair composes a VALUE anchor with BOOK state and a guard, because a quote is not a value.

Then the part a scanner cannot offer: honest grading of the judgment, not the alert.

VIEW desk budget 900
  skyline
  levels
  chain

GRADE plan headline-chase OVER 2026-04..2026-06 FILL conservative
  VS null
  BY regime

VS null and VS ungated are counterfactuals in the syntax: the grade tells you what the judgment earned over doing nothing, and whether the regime gate earned its keep. A scanner reports how often its alerts "worked"; that is not the same question as whether this agent's decisions, on days it could not have memorized, beat doing nothing. A contamination-fenced Grade answers that question; alert statistics cannot.

Comparison table

Standardized columns. Where a Trade-Ideas detail is uncertain, the cell describes the dimension honestly rather than inventing a spec.

Where Kestrel is not the fit

To be plain about it: if you are a person who wants an AI to hand you setups, Kestrel is not for you. It renders to text a model reads, not charts your eyes want. It has no windowed alert UX, no overnight human-facing shortlist, and no polished desktop app. It assumes the trader is software with an API and a wallet. If your loop is "AI narrows the field, I decide and click," a scanner is strictly the better tool, and Trade-Ideas is a strong one.

Status, honestly

Trade-Ideas ships today as a scanner product with a real installed base. Kestrel is the OSS language and runtime; kestrel.markets is the commercial platform (the Vercel to Kestrel's Next.js). As of mid-2026: anonymous trial sims, certified Grades, shareable proof URLs, and 402 Offers with Stripe settlement are live; always-on paper presence and the human-signed live path are in build. The free tier needs no signup. Host the scarcity, rent the genius: the scarcity is capital, licensed data, latency, deterministic execution, and provenance; the brain stays outside (the external-agent invariant). This is illustrative and impersonal, not investment advice.

Bottom line

These tools answer different questions. Trade-Ideas answers "what setups should a person look at right now?", and answers it well, for a human. Kestrel answers "can this model see the market cheaply, act on it in time, and be graded on judgment it couldn't have memorized?" Pick by your reader.

Trade-Ideas scans a market for a human to act on; Kestrel is the language an agent authors to see, act, and be graded. Different reader, different tool.