Starter mode vs full mode: why we gate diversification behind volume

Two modes, one difference

HyperMirror runs in one of two modes. In starter mode your capital mirrors a single leader. In full mode it mirrors up to ten leaders simultaneously, allocated in proportion to each leader's composite score, with one Hyperliquid sub-account per leader. Full mode unlocks at $100,000 of mirrored volume.

It is worth being precise about what does not change between the two. The custody model is identical: funds stay in your own Hyperliquid account, permission is a revocable trade-only agent approval, and there is no withdrawal or transfer right in either mode. The fee is identical: 0.1% of mirrored notional via Hyperliquid's native builder-fee mechanism, with no profit share and no subscription. The selection process is identical: the same five scoring factors, the same qualification floors, the same replacement logic. Sub-account isolation is present in both.

The only structural difference is how many leaders your capital is spread across. That is the entire distinction, and it is the reason the threshold is expressed in volume rather than in features.

What starter mode actually is

Starter mode is not a trial, a demo, or a degraded version of the product. It is the full execution stack pointed at one leader.

The mirrored leader is chosen by the same composite score used in full mode, so a starter-mode user is following whichever qualifying trader currently ranks highest on the same criteria — realized PnL consistency, win rate, profit factor, position discipline and account survivability. If that leader's score decays or a hard risk breach occurs, the same probation and replacement logic applies. Starter mode is not a leader you pick and are stuck with.

The risk controls are also unchanged. Position sizing is derived from allocated capital rather than copied from the leader's size, leverage is capped independently of whatever the leader chooses, per-leader notional ceilings truncate outsized positions, and the leader runs in a dedicated sub-account rather than in your main perps account. That last point matters even with one leader: isolation keeps mirrored activity separate from anything you trade manually, so your own positions and the mirrored ones cannot net against each other or share a liquidation price.

What starter mode does not give you is diversification across leaders. Your outcome is one leader's outcome, with all of the concentration risk that implies. We state that directly rather than framing single-leader mode as a feature.

What full mode unlocks

Full mode changes the allocation problem. Instead of one leader receiving all mirrored capital, up to ten qualifying leaders receive a share weighted by composite score. Higher-scoring leaders receive proportionally more; lower-scoring leaders still inside the qualification floors receive less. Nothing is split evenly, because an equal split would treat a marginal qualifier as equivalent to the strongest one in the set.

Each leader gets its own Hyperliquid sub-account. This is not an accounting convenience. Hyperliquid holds one net position per market per account, so two leaders taking opposing sides of the same perp in a shared account would cancel each other out — you would pay both sets of fees and hold close to no exposure. Isolation is what makes running ten leaders coherent instead of self-defeating: independent margin, independent liquidation price, and PnL attributable to the specific leader that produced it.

The practical effect is that no single leader's drawdown, style drift or disappearance determines the whole result, and tracking error becomes partly uncorrelated across the basket rather than concentrated in one strategy's execution characteristics.

The real reason for the $100,000 threshold

The honest answer is capital adequacy per sub-account, not monetisation. Diversification on Hyperliquid is not free at small size — it has mechanical costs that scale inversely with the capital behind each leader.

Three constraints drive this.

• Minimum order size and lot increments. Hyperliquid enforces per-market size and price increments and a minimum order value. When mirrored capital is divided ten ways and then sized proportionally to a leader's position, the resulting order can fall below the minimum or be rounded to a lot that is a large percentage away from the intended size. At that point you are no longer mirroring the strategy — you are mirroring a rounded approximation of it, and some signals are skipped entirely.
• Margin per sub-account. Isolation means each sub-account holds its own margin. Ten sub-accounts each need enough balance to open a position, absorb unrealised loss and survive normal volatility without being liquidated on noise. Splitting a small balance ten ways produces ten thinly margined accounts, each with a liquidation price closer to entry than the leader's own — which converts diversification into ten fragile positions rather than one robust one.
• Fee and spread drag relative to position size. Every mirrored round trip pays taker cost and the 0.1% builder fee. Those costs are proportional to notional, but rounding loss and the practical minimum position size are not. On very small per-leader allocations, execution frictions become a larger share of the position's expected outcome.

Why starting with maximum diversification is not better

The intuition that ten leaders must be safer than one is correct in a world where each allocation can be executed faithfully. It stops being correct when the allocation is too small to be executed faithfully.

Diversification reduces the variance of your outcome by combining exposures that are not perfectly correlated. That benefit depends on actually holding each exposure at roughly the intended weight. If four of ten leaders are producing signals too small to fill, two are being rounded to double or half their intended size, and the rest are running on margin thin enough that ordinary volatility risks liquidation, the portfolio you hold is not the portfolio the weights describe. You have added complexity and execution error without adding the diversification the complexity was supposed to buy.

There is a second effect that is easy to miss: at small size, tracking error stops being noise and starts dominating. The gap between a leader's result and a mirrored sub-account's result comes from latency, book depth, sizing and funding timing. Rounding on a thin allocation amplifies the sizing component specifically, because the error is measured against a smaller base. Ten thin allocations mean ten independent sources of amplified sizing error, and those do not cancel out — they accumulate as cost.

So the threshold is a sequencing decision. One leader with adequate capital behind it is a cleaner exposure than ten leaders without it. Below the threshold, concentration is the smaller of the two problems.

How the upgrade actually happens

There is no application, no form, no review and no payment. Full mode is not a plan you purchase — the 0.1% builder fee is the only fee in either mode, and it does not change when the mode changes.

The system tracks your cumulative mirrored volume, meaning the notional value that has actually been executed on your behalf. When that figure passes $100,000, the mode changes and capital begins being allocated across the wider set of scored leaders, each in its own sub-account. The current mode and your progress toward the threshold are visible in the dashboard, so the state is observable rather than something you have to ask about.

Two clarifications, since both are reasonable questions. Mirrored volume is not the same as deposited capital: a larger balance reaches the threshold faster because positions are larger, but the measure is executed notional, not what sits in the account. And the threshold is about execution history, not about verifying you as a customer — nothing about the upgrade involves us gaining additional permissions over your account. The agent approval and its limits are identical before and after.

Designing for progressive complexity

A system can be honest about the fact that its most sophisticated configuration is not appropriate for every account size. Score-weighted allocation across ten isolated sub-accounts is the design we think is correct — it addresses concentration risk, edge decay and position netting at the same time. It also requires enough capital per sub-account for each allocation to be executed as intended, and pretending otherwise would produce a worse outcome dressed as a better feature.

So the product starts at the configuration that works at smaller size, and expands to the configuration that works at larger size, with the transition driven by a measured figure rather than a sales conversation. That is the whole reasoning. No tier, no gate to negotiate, no upsell.

If you want the mechanics underneath both modes — how scores are computed, how allocation weights are derived, how isolation and replacement work — the documentation and how-it-works pages set them out in full, and the risk disclosure covers what neither mode fixes.