Mainnet
DOCUMENTATION

Know what the numbers mean.

Research provenance, transaction mechanics, and the limits of what this interface can establish.

Source-backed research

HASHFRONT is an independent read-only interface over publicly exposed Yukon benchmark data. It is not operated by Yukon and does not run or represent Yukon’s judges. Source records, timestamps and status are fetched through the public benchmark and submission APIs and cached for five minutes. Unavailable data is identified explicitly.

Only benchmarks linked to these source repositories are included. Submissions are runtime-validated. Unrecognized score policies do not enter the cost plots. Model names are submitter-reported, extracted from public attribution, and are not a controlled comparison of model capability.

Normalized time–memory cost

log₂(T × M′) = timeLog2 + max(0, memoryLog2Bytes − 20)

M′ = max(1, memory / 1 MiB)

The 1 MiB floor prevents a sub-mebibyte memory budget from artificially reducing the score. Units and computational assumptions come from the applicable source cost model. The Orchard field-RAM model is separate from the collision-frontier model used by the bitwise hash families.

For Pareto records, round count and cost are read by objective name from the bound evaluation. The top-level official score can be a round count and is not interpreted as normalized cost. Scalar records require the exact normalized time–memory policy markers.

Four views, distinct questions

  • Frontier: source-accepted, non-dominated round/cost points alongside provisional claims. Connections between points do not prove attacks at intermediate rounds.
  • All Rounds: qualified historical records, filtered by algorithm, status and round count.
  • Time Series: actual acceptance timestamps for accepted points; submission timestamps for provisional points. No synthetic historical dates.
  • By Model: counts of accepted and provisional model-attributed submissions. Counts are not a scientific ranking.

Safe, Wounded, Unsafe and Very Unsafe are source-specific classifications. HASHFRONT does not assign these labels when a compatible explicit classification is absent. The interface shows review status and nominal references instead of inventing security verdicts.

Scoring policies · inspiration & sources

Meet the review process

Screening, reproducibility, AI assessment and publication are different stages. Passing an automated evaluation does not constitute mathematical proof or manual acceptance.

  1. Structured claims and artifacts are submitted through the official source workflow.
  2. The source evaluates validity, execution evidence, assumptions and normalized cost.
  3. AI qualification such as “plausible_not_refuted” remains qualified language.
  4. Scalar acceptance requires accepted status and successful promotion. Pareto acceptance requires accepted status and acceptance in the bound human evaluation.

HASHFRONT does not execute uploaded research code. Participation links lead to the official workflow; there is no fabricated independent judging service.

Two real launch destinations

Pump.fun

Token creation uses the official Pump SDK’s createV2Instruction, SOL quote, Token-2022 mint, and optional live-enabled Mayhem or holder-reward modes. New cashback creation is disabled. Curve trades use v3 instructions and current SDK quotation; graduated SOL markets route to the canonical PumpSwap pool after it exists on chain.

Official Pump documentation · inspiration & sources

Meteora DBC

The default creator-funded configuration uses standard SPL tokens, 1 billion supply, 6 decimals, a constant 1% curve fee, 85 SOL migration threshold and DAMM v2 migration. Twenty percent of the supply is designated for migration; migrated LP is permanently locked to the creator. Protocol fees still apply. New configuration and pool initialization execute atomically in one wallet approval when they fit the packet limit; unusually long metadata falls back to separate, recoverable transactions. Existing configurations are decoded and checked before reuse.

Official DBC SDK · inspiration & sources

Execution and recovery

HASHFRONT checks the mainnet genesis hash, simulates before signing, checks packet size, persists the signed transaction’s public signature before submission, polls confirmation, and verifies the expected accounts. Pending outcomes must be reconciled rather than blindly resubmitted. No wallet seed phrase or private key is requested or transmitted.

Optional initial buys are separate transactions after launch verification, not atomic guarantees of an opening price. Metadata charges are reviewed separately. RPC services can throttle requests; the public fallback offers no capacity guarantee.

Zcash node connectivity

The Zcash network observer reads Zcash mainnet through a shared public Tatum JSON-RPC endpoint or your own public HTTPS node. It verifies the mainnet genesis hash before showing chain data. Block height, sync progress, peers, software version, block time and aggregate value pools are node-reported. Missing fields are marked unavailable.

Shared status refreshes every two minutes and may be briefly cached. Custom nodes support Basic authentication or bearer tokens; connection details remain in tab memory and are sent to the server only to query that endpoint. Private-network addresses and redirects are rejected. This feature reads network status and does not host a node, manage keys or send ZEC.

Zebra connection documentation

Durable metadata

The launcher supports wallet-funded Irys image and JSON uploads and a validated user-supplied HTTPS metadata URI. It checks the metadata name, symbol and image content before preparing a launch. PNG, JPEG and WebP images are limited to 5 MiB. Upload charges and any additional funding are shown before approval. No operator-funded storage account is assumed.

A creator-supplied HTTPS host can change or disappear; availability checks do not guarantee permanence. Research association resides in creator-supplied metadata and is not independent verification.

Disclosures & risks

Research results are not trading advice. Market prices do not measure research quality. Reduced-round attacks and normalized cost improvements do not demonstrate a real-world break of full-round SHA-256 or another standard.

Tokens can lose all value. Users bear transaction, rent, storage, slippage, protocol, liquidity, smart-contract and RPC risks. HASHFRONT is non-custodial and does not supply liquidity, deploy a proprietary token-launch program, or maintain an operator treasury.

DEX Screener and GeckoTerminal market data may be delayed, unavailable or incomplete. On-chain ownership is checked before a trade route is built. Listings are third-party markets, not claimed HASHFRONT launches. Wallet history is local to the browser; no exhaustive platform launch count is asserted.

Integration code and local tests do not establish funded-mainnet success. Only a real transaction signature and verified chain accounts establish an individual successful launch or trade.