The reserve isn't SOL. It's a 2x/3x/5x perp on BTC, HYPE, or a stock. Your bag has a heartbeat now — it ticks with the underlying, even when nobody trades. Right regime → the curve graduates almost by itself. Wrong regime → time bleeds you out. la riserva non è SOL. è un perp leverato Hyperliquid. il tuo bag ha un battito anche quando nessuno scambia.
Three things. Don't scroll past if you'll skip the rest. tre cose. se salti il resto, almeno leggi queste.
On pump.fun the reserve is spot SOL. Here it's a BounceTech Leveraged Token — an ERC-20 that wraps a leveraged perp on Hyperliquid. The token's price moves even when nobody trades.
The reserve moves with the underlying. Trend in your favor → graduation can fire without a single new buyer (curve almost frozen). Choppy market → vol decay drains the reserve: graduation drifts away and everyone bleeds.
Hyperliquid (oracle, CLOB) · HIP-3 (builder-chosen oracle) · BounceTech (NAV, redemption capacity) · alt.fun (smart contract, MEV). Safe? No. More transparent than standard pump.fun? Yes.
Same math (constant product, bonding curve). Different physics. stessa matematica, fisica diversa.
y/x, trade flow). On alt.fun you get two: y/x (trading) + p_LT (underlying × leverage, ticking on every oracle update from the Hyperliquid perp). That single line is the entire whitepaper.
Click a layer to expand. The stack is vertical and additive: every layer adds functionality, and also a dependency you inherit. clicca un layer per espanderlo. ogni layer aggiunge una dipendenza.
A token's USD price is the product of a "curve ratio" and the LT's NAV. Sounds obvious. It is the entire mechanism. due motori spingono il prezzo: trade flow e underlying drift.
Moves only when someone buys or sells on the curve. The classic "pump.fun behavior".
The LT's NAV. Ticks on every oracle update of the Hyperliquid perp, amplified by leverage.
p_token,0 = (V/p_LT)·p_LT / S = V/S, so MC₀ = V = $4,000. Initial price-per-token is identical across all launches. It's a deliberate design choice — tokens are comparable from the second they exist.
Move the sliders. Watch trade flow and underlying drift combine. Leverage only amplifies engine 2 — not engine 1. muovi gli slider. vedi come i due motori si combinano.
The "twist" is the virtual reserve: the contract accounts for more liquidity than it actually holds, giving the curve a sensible opening price without needing real USDC seed. le virtual reserve permettono alla curve di partire a un prezzo sensato senza liquidità reale.
First you mint LT from BounceTech with USDC, then spend it on the curve. The UI bundles it in a single tx.
ΔLT is atomically redeemed for USDC at NAV. If the LT's idle USDC buffer is depleted, the sell tx reverts.
Slide tokens-sold and current LT NAV. See price, reserve, and how the two graduation triggers race each other. muovi gli slider: vedi prezzo, riserva, e come i due trigger competono.
First one to hit wins. Graduation is permissionless — anyone can close it, the protocol can't block it. due strade alla graduation, in OR. chiunque può triggerare.
The LT held in the curve crosses the USD threshold. Even with a near-frozen curve, if the underlying pumps, p_LT rises and the product breaks the threshold.
All 750M curve tokens have been sold. The curve empties regardless of p_LT. Classic pump.fun-style "buy frenzy".
t_LP = ℓ · r₀ / r₁ forces the same reserve ratio onto the new pool.
The protocol holds 25% of supply outside the curve. It's not arbitrary. It's the maximum of a parabola — proven, not hoped. il 25% non è scelto, è il massimo di una parabola dimostrata.
Maximize by differentiation: d/ds = (S − 2s)/S = 0 → s* = S/2. Plug back in: t_LP(S/2) = S/4 = 25%.
For any path to graduation — few buyers + big underlying pump, or zero pump + many buyers — the protocol has always enough tokens to seed the AMM. Anything not needed gets burned.
Notable: p_LT doesn't appear in the final formula. It cancels out in the derivation — sufficiency is pure unit math.
x: tokens sold (s) · y: tokens needed to seed AMM (t_LP)
Five design choices that move trust away from the team. cinque scelte di design che spostano la fiducia via dagli operatori.
Trades gated for the first 3 blocks via EIP-1153 transient storage. Bots can't buy in the same block as the deploy → narrows (doesn't kill) the sniper-vs-human asymmetry.
Every alt-coin deploys to an address that ends in a fixed suffix. The contract rejects non-matching deploys. Frontends and indexers filter fakes at the structural level — no whitelist lookup needed.
LP tokens minted at graduation get sent to a custodian with no withdraw(). Liquidity locked forever. No governance recovery path. Anti-rug, structurally.
Anyone can close graduation once the trigger fires. The protocol runs a backstop keeper, but it's not a privileged actor. The protocol can't censor a graduation.
The AMM's first tick = last curve price. No bot snipes the graduation arbitrage gap. Sounds like a detail: in practice it redistributes MEV to late-curve buyers.
Whitepaper-listed. Here with the numbers that hurt. cinque rischi che con la riserva spot non avevi.
A constant-leverage instrument that doesn't liquidate must rebalance. Rebalancing in chop converts realized vol into NAV bleed. Approx: decay ≈ −½·L·(L−1)·σ²·dt.
The LT never liquidates, but in a sustained crash NAV converges asymptotically to zero. The curve's reserve loses USD value — even with zero sells on the curve.
Distance to USD trigger can grow back. At 80% of threshold, a −20% move in p_LT drops you to 64%. No ratchet. Sitting in a marginal position is negative-EV.
LT redemption is atomic, capped by the LT's idle USDC buffer. In a mass-sell event the buffer depletes → sell txs revert. Sellers face asymmetric execution risk.
Hyperliquid downtime → oracle frozen. HIP-3 builder compromised → bizarre mark price. BounceTech vault exploit → LT goes to 0. 4 layers = 4 single points of failure.
Drift (trend) compounds in your favor. Vol decay bleeds you in chop. See how much NAV remains in 30 days. drift vs decay. quanto il tempo mangia (o nutre) la riserva.
½·L·(L−1)·σ² per day.
Same whitepaper, three players, divergent optimal moves. That divergence is where the edge lives. stesso paper, tre attori, mosse ottimali diverse.
alt.fun isn't "pump.fun on Hyperliquid". It's a different mechanism: the reserve is an active financial instrument. Buying an alt-coin means buying two things at once wrapped in a single ERC-20, a price that ticks even with zero trades, and the chance that time works for you or against you depending on the macro regime of the underlying the creator picked.
What makes alt.fun an existential bet isn't the curve mechanics — it's whether memecoin traders will accept that time is no longer free. On pump.fun the median holder survived by waiting. Here they don't. For real degens: more edge for the chart-readers, less edge for the pure Twitter-trackers.
x·y=k, p_token = y·p_LT/x, t_LP(s) = s(S−s)/S verified by re-derivation.T = $12,000 (default); docs.alt.fun (snippet) mentions $9,000. Verify in-app.≈ −½·L·(L−1)·σ²·dt is the classical leveraged-ETF approximation, not BounceTech-specific. Real numbers depend on exact rebalance schedule.(1+r)^L — accurate for instantaneous moves with rebalanced position. The vol-decay calculator (sandbox 03) uses the continuous-time log-drift approximation.