Skip to content

Cookies are off until you say otherwise

Non-essential cookies stay off unless you accept them, and rejecting them leaves the site fully usable. What runs either way is only what the site cannot work without: the anti-abuse check on our forms, the record of the choice you make here, and the access cookie on internal routes. Site analytics is cookieless and loads only after you accept. Cookie policy

Quant24
Start in Sandbox — free
Pre-launch · waitlist open

You set the goal. Our agents test it.

Describe a trading goal in plain English. Two language agents turn it into a testable strategy; an engine in code runs the validation; you approve every deployment.

Start in Sandbox — freeSee how it works

US regulated futures · BYOK · Human approval before every dollar

The whole thing, in five steps

  1. You type a goal in plain English.current
  2. It becomes a Goal Spec: one metric, one benchmark, a risk budget, a universe.next
  3. Six engines run in parallel over a training split, a test split and a sealed holdout.next
  4. The posterior reports the probability of meeting the goal, with a credible interval — never a single number.next
  5. That probability sets the starting rung on the capital ladder.next

Probability of meeting the goal, with credible interval

not available in this build

Starting rung on the capital ladder

  1. SANDBOXthe walk-forward did not beat the benchmark, or the holdout degraded severelycurrent
  2. PAPERvalidated, running without orders
  3. LIVE_MICROa small slice of the allocated sizelockedrequires broker connection
  4. LIVE_SCALEDfull allocated sizelockedrequires broker connection
  5. LIVE_FULLnever assigned from synthetic resultslockedrequires broker connection

These steps show the mechanism, not a run. Where the probability of meeting the goal and its credible interval would appear, nothing is shown: the validation engine that computes them does not exist in this build and arrives in a later phase. No number in this section came out of an engine, because there is no engine.

CFTC Rule 4.41(b)(1)(i) — Hypothetical performance disclosure

These results are based on simulated or hypothetical performance results that have certain inherent limitations. Unlike the results shown in an actual performance record, these results do not represent actual trading. Also, because these trades have not actually been executed, these results may have under-or over-compensated for the impact, if any, of certain market factors, such as lack of liquidity. Simulated or hypothetical trading programs in general are also subject to the fact that they are designed with the benefit of hindsight. No representation is being made that any account will or is likely to achieve profits or losses similar to these being shown.

The ladder, in numbers

The posterior from the validation engine reports the probability of meeting your goal with a credible interval of 90%, never as a single number. Where that probability lands is what picks the rung:

  • Below 55%, the strategy runs in paper.
  • Between 55% and 70%, it starts at a micro allocation.
  • Above 70%, and only if the stress test is survivable, it scales.

Both thresholds are calibrable parameters, not constants of nature, and no synthetic result reaches the top rung. Underneath every rung sits the same risk guardian:

  • A drawdown of 2% in a day stops new positions.
  • A drawdown of 4% in a week halves size.
  • A drawdown of 8% in a month pauses everything for a manual review.

Why validation, and not a louder promise

Around 70% of day traders lose moneyPROXY — an indirect figure standing in for one we do not haveF16F17F18. That is the segment, measured by others, and Quant24 does not claim to reverse it. What the ladder does is narrower and checkable: it ties the amount at risk to the evidence that exists, and it moves only after you say so.

What this build is

Pre-launch. No broker connection, no capital and no exchange market data — and no validation engine either: it is built in a later phase, so the product produces no performance numbers at all today. The numbers on this page are product parameters we declare and market figures we cite, each one with its source. When results do exist they will be computed on synthetic data and labeled SYNTHETIC on the number itself.

Who does what, and who says yes

You
The goal, and the approval
You write the goal and you approve the deployment. Nothing reaches a market without an explicit yes from you. BYOK — bring your own broker account: the trade happens there, and Quant24 never holds your money.
The Strategist
Language agent · designs and interprets
Turns your sentence into a measurable contract, researches the mechanism, proposes a handful of candidate strategies with the economics declared up front, and reads what the engine returns. It never produces a number itself.
The Executor
Language agent · runs what you approved
Runs the approved package in three layers: execution in code, a risk guardian with a mathematical veto over every order, and a strategic layer that reviews context on a fixed cadence. Its adjustments never go straight to execution; they go to the guardian first, which can shrink them or refuse them outright.

Five pieces. The combination is the point.

Adversarial validation, not one backtest
Six engines, a holdout sealed on the most recent data, and an auditor that reviews the process — look-ahead, survivorship, trial count, costs — and can send the whole package back to design. A mediocre result does not block delivery; a dishonest process does.
A capital ladder governed by evidence
The starting capital level comes out of the posterior, not out of a good week. The top rung is unreachable on synthetic evidence alone, by design and by type.
The Strategist and the Executor are separate systems
The agent that designs is not the agent that trades, and neither of them can overrule the risk layer. That separation is architectural, not a promise in a prompt.
US index futures, and nothing else
ES, NQ, RTY and YM. One asset class means one microstructure, one margin regime and one set of session times to model properly instead of four done badly.
Honesty as a component, not a footer
Every number carries a badge for where it came from. Every market figure carries its evidence level and its source. Where a number cannot honestly be computed, the page says so instead of showing one.

The capital ladder

Five rungs — SANDBOX, PAPER, LIVE_MICRO, LIVE_SCALED, LIVE_FULL. The posterior decides where a strategy starts. Real results, accumulated in a real market, decide whether it moves. A streak decides nothing.

  1. SANDBOXthe walk-forward did not beat the benchmark, or the holdout degraded severelycurrent
  2. PAPERvalidated, running without orders
  3. LIVE_MICROa small slice of the allocated sizelockedrequires broker connection
  4. LIVE_SCALEDfull allocated sizelockedrequires broker connection
  5. LIVE_FULLnever assigned from synthetic resultslockedrequires broker connection

What you can hold us to

  • A process you can audit and reproduce: every number can be recomputed from the seed, the skill versions and the hash of the data.
  • Metrics reported with credible intervals instead of point promises.
  • A starting capital level proportional to the available evidence, and never above it.
  • Falsification conditions declared before anything runs, not explained after it fails.
  • Your explicit approval before a single dollar is at risk.

Risk

What nobody serious can guarantee

  • Future returns. Nobody serious can promise them, and you should be wary of anyone who does.
  • That the odds of the segment change. Around 70% of day traders lose money, and validating a strategy does not reverse that.PROXY — an indirect figure standing in for one we do not haveF16F17F18
  • That a stop loss is a price. Gaps, thin liquidity and slippage are real; historical buffers and circuit breakers reduce them, and the residual is never zero.
  • That a simulation equals a market. That is why every synthetic number is labeled, and why the top rung asks for a real track record.
  • That the language agents add value by construction. Their contribution is measured against the same system without them, and if it stops adding value the kill switch drops to the engine alone.

What we're built on

  • Cloudflare Workers, D1, R2, Queues and Durable Objects
  • Workers AI behind an AI Gateway, with the model chosen per task
  • Two speeds in the design — an interactive path for the session, a batch path for validation — neither of them running in this build
  • No market data of any kind in this build — no exchange feed, no synthetic bars, no broker connection
  • Stripe in test mode, Resend for mail, Turnstile wired for the forms Phase 5 will publish
See the full stack

Start free. Pay when there is something to pay for.

  • Sandbox: free
  • Live: from $79 (waitlist)
See all plans

Get involved

Founder Program
A closed cohort. The Live plan at no cost while the cohort runs, in exchange for an anonymized track record, weekly feedback and pricing that locks in for life.
Creator Program
A small first group of creators who publish the process — the goal, the evidence, the approval — and never a profit-and-loss screenshot.
Investors
The thesis, the moat and the market, every figure at its evidence level. No round terms, no projections, no offer.

Frequently asked questions

FAQ

Start where nothing is at risk

Sandbox costs nothing, needs no card and needs no broker account. One strategy, validated end to end, with the whole evidence trail open to read.

Start in Sandbox — free

Sources

  1. F16BrokerChooser — Day Trading Statistics 2026 (survey, N ~ 89,606) (opens in a new tab)2026
  2. F17Gitnux — Retail Investors Statistics (opens in a new tab)2025
  3. F18QuantifiedStrategies — Day Trading Statistics 2026 (opens in a new tab)2026