Product Controls

This page captures the current product decision for clone controls and strategy building blocks.

Related:

• Clash Of Clones Product Spec
• Domain Language
• Capacity And Model Tiers
• Backend Contract

Product Surface Map

The strategy builder is one part of the larger clone lifecycle:

Pick model/capacity -> pay Audition Fee -> public identity -> private strategy -> audition -> Arena

The product has two distinct user-facing configuration surfaces:

Surface	Purpose	Primary owner
Public identity	Name, description, image, model tier, public assets/data streams	Product/profile UI
Private strategy	Data streams, assets, trading prompt, cadence, execution behavior	Strategy builder and trading runtime

The private strategy is visible only to the creator. Public identity and selected capability signals are visible to backers on clone cards, race views, and Hall of Fame.

Clone Capacity And Model Controls

Before strategy configuration, users choose asset capacity and model strength. These determine the paid audition tier and the entitlement that the strategy builder and runtime enforce.

Planned tiers:

Tier	Active assets	Enabled prompts	Max data links / asset	Max total expanded data links	Model strength
Starter	1-3	1	3	9	Cheapest decision models
Builder	4-6	3	5	30	Medium decision models
Pro	7-10	5	8	80	Best decision models, such as GPT-5.5 and Claude Opus

In v1, active assets also cap expanded Asset Block -> Trading Prompt Block assignments.

Current behavior:

• The Economy module quotes and verifies audition-fee deposits, then grants a clone entitlement.
• The backend still returns a mock Starter entitlement when no purchase-backed row exists, mostly for local/dev and legacy clone paths.
• The Clone Block model selector shows stronger locked models with an upgrade prompt placeholder.
• The workflow canvas locks Asset Blocks beyond the current entitlement capacity.
• The Pipeline API and runtime paths reject active assets, prompts, or context-link topology above the active entitlement.
• Real checkout should use /api/v1/economy/audition-fee/quote and /api/v1/economy/audition-fee/submit; race ticket pricing is still future lifecycle work.

See Capacity And Model Tiers for the detailed plan.

Runtime topology caps:

• Assets: hard v1 cap is 10 active assets per clone.
• Trading prompts: tier-derived enabled prompt cap.
• Context links: tier-derived expanded Data Stream Block -> active asset cap.
• Runtime expansion: one decision run per Trading Prompt Block + active asset per due tick.

Basic creation can still default to one Trading Prompt Block. Custom and power-user flows can expose multiple Trading Prompt Blocks within tier limits.

Public Identity Controls

Clone identity is a first-class backer-facing surface:

• Name: required, max 32 characters, unique across active Arena clones.
• Description: required, max 280 characters.
• Image: required, 512 x 512, PNG or JPG, max 1 MB.
• Model tier badge: Haiku, Sonnet, or Opus.
• Public asset selections.
• Public data stream selections.

Current moderation is report-based. Public profile submit auto-publishes; users can report abuse, and admins can retire or reinstate a profile. If an operational moderation strike invalidates an in-flight audition, admins can void that round.

During race accumulation, clone cards should visually prioritize:

1. Image and name.
2. Description.
3. Model tier badge.
4. Career 1st-place count.
5. Career top-10 rate.
6. Current streak.
7. Last-race finish.
8. Total tickets sold for this race and top backer concentration.

Current Workflow Shape

The persisted strategy graph has three block types:

Data Stream Block -> Asset Block -> Trading Prompt Block

The workflow canvas also shows a visual terminal block:

Data Stream Block -> Asset Block -> Trading Prompt Block -> Trading Clone Block

The Trading Clone Block is not a strategy block. It holds clone-level run controls such as model, trading on/off, and trading history. Its canvas position and visible prompt-to-clone connection are saved as layout metadata so the workflow reopens in the same shape without turning the clone block into an executable graph node.

Each Trading Prompt Block defines one strategy. At runtime, the backend expands that strategy across the connected assets:

one strategy branch -> selected assets -> one decision run per due asset

Data Stream Blocks add context to each due asset run. They do not create additional model runs.

Control Granularities

All modes save to the same backend graph shape. The difference is how much of that graph the user sees.

Mode	User-facing shape
Basic	Generated graph with simple defaults
Custom	Same graph with per-asset/source presets exposed
Power User	Direct graph editing with advanced source and prompt controls

Basic - Simple Defaults

Basic users should not need to understand the graph.

• One generated Hyperliquid OHLCV Data Stream Block scoped to the selected/default asset.
• One or more generated Asset Blocks from category or symbol selection.
• One generated Trading Prompt Block.
• One Trading Clone Block for global model and trading on/off.
• Retrieval depth defaults to Balanced.
• Prompt behavior can use a short custom behavior prompt.
• No manual channel, lookback, granularity, datapoint, leverage, or risk guardrail tuning.

Custom - Per-Asset Presets

Custom users can choose assets and tune broad behavior without editing raw retrieval settings.

• Drag or select individual Asset Blocks.
• Attach Data Stream Blocks to Asset Blocks.
• A Data Stream Block can use a different context asset than the Asset Block it feeds, for example ETH TVL or OHLCV context feeding a BTC trading branch.
• Hyperliquid blocks are dropped by source/channel, then configured with the associated asset inside the block.
• DefiLlama blocks are dropped empty, then configured with channel and macro scope: global, chain, protocol, or stablecoin.
• Choose retrieval depth: Fast, Balanced, or Deep.
• Choose stream prompt inclusion: Off, Auto, Standard, or Required.
• Enable or pause each asset.
• Enable or pause each trading prompt.
• Edit trading prompt cadence, custom behavior prompt, and execution behavior.
• Use each Data Stream Block's View Data link to jump to the source site for the connected asset, market, chain, protocol, or stablecoin.

Power User - Direct Graph Editing

Power users can edit the strategy graph directly.

• Create/delete Data Stream, Asset, and Trading Prompt Blocks.
• Connect multiple strategy branches.
• Use multiple context blocks, with duplicates scoped by source/channel/target.
• Open advanced stream controls for prompt inclusion, lookback, and granularity.
• Keep Asset Blocks limited to asset identity, enabled/paused state, and exchange links.
• Use global prompt behavior, global prompt with per-asset overrides, or asset-specific-only prompt coverage when the backend/UI exposes those controls.
• Configure strategy, risk, and execution behavior on Trading Prompt Blocks.

Control Organization

Data Stream Controls

Data Stream Blocks answer:

What structured market memory should this trading branch make available to the prompt?

Current user-facing blocks are concrete context blocks, not source-family bundles. Examples:

• BTC / Hyperliquid / OHLCV.
• BTC / Hyperliquid / RSI.
• Solana / DefiLlama / Chain, protocol, and stablecoin flow changes.
• Polymarket / Specific Market, selected through keyword search.

The connected Asset Block is the asset being traded. The Data Stream Block's selected asset, market, chain, protocol, or stablecoin is the context being read. These can be the same, but they do not have to be: ETH macro or price context can feed a BTC decision branch when the user connects that stream to the BTC Asset Block.

Current top-level controls:

• Retrieval depth: Fast, Balanced, or Deep.
• Prompt inclusion: Off, Auto, Standard, or Required.
• View Data action scoped to the stream's configured context target.

Current advanced stream controls:

• Lookback.
• Granularity.

Max datapoints and prompt budget are managed automatically from lookback and granularity. The UI can show the derived datapoint cap, but users should not need to tune it directly in the default flow.

Retrieval is deterministic. There is no AI prompt for data retrieval.

Asset Controls

Asset Blocks answer:

Which individual asset is this strategy branch allowed to consider?

Asset selection is individual in the workflow canvas. The palette can still be organized by Hyperliquid categories and subcategories:

• High-level categories: All, Perps, Spot, Crypto, Tradfi, Trending.
• Crypto subcategories: AI, Defi, Gaming, Layer 1, Layer 2, Meme.
• Tradfi subcategories: Stocks, Indices, Commodities, FX, Pre-IPO.
• Optional spot subcategories: USDC, USDH, USDT.

Generic HIP-3 category exposure is out of scope.

Current Asset Block content:

• Asset display name, exchange symbol, and asset kind.
• Enabled/paused switch.
• View on Hyperliquid link.
• Delete block action.

Asset Blocks do not expose leverage, position sizing, trade direction, or risk tuning. The backend only reads the asset enabled/paused state from Asset Blocks. Strategy behavior and safety defaults belong to Trading Prompt Blocks.

Trading Prompt Controls

Trading Prompt Blocks answer:

How should the clone decide what to do for each due asset?

Current visible controls:

• Prompt enabled/paused.
• Custom behavior prompt.
• Decision cadence.

Current backend-owned prompt execution defaults:

• Max notional per action.
• Minimum model confidence.
• Cooldown.
• Max trades per day.

These defaults are safety rails around the prompt, not a separate strategy form. The user-facing strategy remains primarily prompt based.

The old trader archetype, market momentum sensitivity, and analysis timeframe controls are legacy migration inputs. They can still exist in compatibility data paths, but they are not the primary workflow controls in the current strategy builder.

Trading Clone Controls

The Trading Clone Block answers:

What global runtime settings apply to the clone?

Current visible controls:

• Model.
• Trading enabled/paused.
• Trading history.

The model is clone-level, not per Trading Prompt Block.

The prompt-to-clone connection is a visual runtime route. It should remain visible in the canvas and round-trip through pipeline layout metadata, but it is not part of the executable v1 graph that the backend compiler validates.

Data Review UX

Each Data Stream Block should expose a View Data action, but that action should be an external source link rather than an in-canvas raw-data drawer.

Examples:

• Hyperliquid blocks link to Hyperliquid trade pages.
• Polymarket blocks link to the selected market URL when available, otherwise to Polymarket search.
• DefiLlama blocks link to the configured chain, protocol, stablecoin, or source-family page.

Operational data completeness should live in a separate stats or ingestion-health surface. The strategy canvas should stay focused on block composition rather than database row inspection.

For DefiLlama, the viewer should not imply exhaustive per-token source coverage. New blocks use the configured macro scope: global, chain, protocol, or stablecoin. Legacy auto-scoped DefiLlama reads return global DeFi context across the configured lookback, plus matching chain and stablecoin rows.