STACC

Core-platform developer at the Solana memecoin launchpad during its hyper-growth phase — hired out of the community after building vaults on top of their bonding curves. Remote, then embedded with the team in the UK.

• Frontend PRs on the main app, including wallet-flow fixes reviewed into production
• Built transaction-inclusion metrics (Grafana + custom server): “% of transactions submitted that actually land on-chain”
• Supabase schema migrations; moderation tooling (ban-term automation, bulk actions, unban flows)
• Coordinated the Pashov security-audit pipeline
• Product vision later shipped by the platform: a native AMM (“an AMM that is literally just entirely a bonding curve” → PumpSwap), perpetual creator revenue sharing (→ Project Ascend), referral-fee-by-default (→ multi-wallet fee splits), treasury-into-LST value accrual (→ PUMP buybacks) — dated team-chat receipts, Apr–May 2024

Algo-trading SaaS — strategy engine, exchange integrations, and the live alpha fund.

• Shipped the first production strategy (“longshort”) to paying users — ~$16k across 5 users in alpha, reported 3.5–6% average daily returns over the sample window
• Owned the Binance broker API relationship and Deribit partner-program integration
• Won both the futures and spot/margin buckets of Binance’s 2020 BNB trading competitions — share of the 40k + 10k BNB prize pools — with his open-sourced bot
• Strategy provider on Zignaly — verifiable third-party-tracked performance

Six years of public, verifiable output across Ethereum then Solana.

• 1,400+ public repos & gists — 667 repos, 787 gists (github.com/staccDOTsol): MEV searchers, Jupiter AMM implementation, flash-loan tooling, LUT multi-swap arbitrage, binary options on Token-2022
• Merged/PR contributions: solana-labs/solana #30380, solana-program-library #6071, GlassEaters/hydra #21/#26/#30, blockworks-foundation/autobahn #16, raindrops-protocol #30
• ETH-era public arbitrage bots (dexArb, 2020); RAY (Staked.us Robo-Advisor for Yield) governance contributor
• Ran mainnet and testnet validators; dedicated RPC/geyser infra for low-latency strategies

Oracle-free leveraged synthetic-pairs AMM on Solana. Deposit USDC, receive a Token-2022 receipt token representing a leveraged pair — deployed as a triangle of three CP-Swap pools in one atomic transaction, rebalanced by “minting the loser” instead of trusting an oracle. Rebalance MEV accrues to LPs, not arbs.

• Full mechanism spec + on-chain program design (Anchor/Pinocchio, tx-introspection enforced, mandatory LUTs)
• Production-grade design system for the trading terminal: tokens, components, lint-adherence config, content/voice guide
• Live at magicalinternet.money — mainnet alpha, unaudited (honesty-first labeling, by design)

Risk-managed yield infrastructure designed for conservative operators — built so that even the most cautious treasury (up to a nation-state’s) can outpace inflation without taking speculative risk. Android release build shipped; state-level implementation deck available on request.

A presale where the launch price is simultaneously the all-time-high and the floor: half the raise buys the token, half locks as a USDC quote pool — one atomic bundle, zero sellable supply at t=0. The only holders are presalers, and they literally cannot sell into air.

• Live tracker wired to real mainnet state (JSON-RPC + WebSocket account subscriptions)
• Full thesis memo + mechanism design published on stacc.art

Mirror-able Solana bonding-curve launchpad: referral fees, Orca liquidity adds, buy-and-burn, and one-click Vercel mirror deployment so communities can run their own front ends. Public repo: staccDOTsol/pumprecovery.

Deflationary liquid-staking token on Sanctum’s audited stake pool — burn-loop value accrual, designed as savings infrastructure, not a casino chip.

Agent-native game design: an MCP-native “last bidder wins” game where AI agents register faucet-funded play wallets and bid via tool calls — shipped as an OpenAI Apps SDK widget (playable inside a chat client) with snipe-extend rounds and settle-to-claim payouts.

Browser extension that rebates attention: XP for ad impressions with a Bitcoin-style halving schedule, three-level referrals, and prize draws — Firebase/Firestore backend with security-rule-clamped batched writes. Privacy-safe by construction: never reads page content.

A trained LoRA adapter (rank 32, ~1 GB of weights) teaching a 32B instruct model a specific human writing voice — end-to-end: raw archive → cleaned train/valid/test splits → PEFT training → ChatML-templated artifact.

Game theory you can play: a Solana remake of Fomo3D (Token-2022 LSTs, staking, bounties), gobbler.fun fee-split games, a casino-grade slots prototype fusing scatter-pays with last-key-wins pot economics, and fully on-chain Clockwork-thread games.

• ffragmented — trading UI for ~1,284 long-tail Orca whirlpools that orca.so doesn’t list: same protocol, same fees, infinitely more pools
• Prompta — voice-tracked AI teleprompter: realtime transcription follows the speaker, subscription tiers, per-user spend-capped API keys
• ProofFront — manifest-driven dApp frontend framework: declarative scenes, wallet gating, theming (12+ demos)

The problem. Every Solana validator pays a fixed vote-transaction burn — 5,000 lamports × ~410,000 votes ≈ 2.05 SOL per epoch. Below ~5k SOL stake, that fixed cost exceeds the inflation reward: small validators run at a loss by design, and the network centralizes around whoever can absorb the bleed.

The patch. Batch up to 31 votable banks into a single TowerSync transaction (the tower already accumulates the lockouts). Opt-in; default N=1 is bit-for-bit identical to today’s behavior. Force-flushes on fork switches and right before the validator’s own leader slots; an adaptive mode retunes N every epoch from current on-chain stake. Two files touched: replay_stage.rs, tvu.rs.

The receipt. Live on mainnet on his own validator — stacsol.app, a 2,341.72 SOL single-validator LST — since 27 May 2026 at N=31: measured 94.6% vote-fee reduction (2.05 → 0.11 SOL/epoch), vote credits healthy across epoch 978, validator economics flipped from −1.33 to +0.61 SOL/epoch. A +1.94 SOL/epoch swing — ≈ +283 SOL/year per small validator.

The candor. The PR flags its own most aggressive consensus choice (not force-flushing on new_root) and asks reviewers to attack exactly that line, with a documented conservative fallback. That is how you write an upstream PR.

In plain words: validators vote ~410,000 times per epoch and pay a fee per vote (~2 SOL total). Small validators pay more in fees than they earn — so they lose money and quit, and the network centralizes. This patch packs up to 31 votes into one transaction — one fee instead of 31 — cutting the cost 94.6% on his own live validator. Off by default; nobody is forced to use it.

The problem. Legacy Raydium AMM pools hold liquidity whose LP tokens were burned by third parties. A burn is forever by design — so that liquidity is stranded: visible on-chain, unusable by anyone, forever. Dead weight on the protocol’s books.

The patch. An authority-gated rescue path: Raydium — and only Raydium — can withdraw stranded liquidity from pools with burned LP, without burning any LP tokens to do so. First written August 2024, shelved after a friend’s critique (“releasing the liquidity undoes the burn”), revived April 2025 once cp-swap/CLMM shipped proper liquidity locking — the burn-vs-lock distinction the ecosystem was missing.

The governance instinct. Proposed safeguards (e.g. restricting rescue to pools under ~$5k), asked for a token-holder vote, floated a 1% bounty for the treasury. Raydium’s infra lead replied on the record: “Decentralization is a spectrum.”

The callback. Reopened July 2026 after pump.fun shipped “BOOST” — in his words, “exactly what I already built and pitched to you.” A second documented case of his mechanism work shipping elsewhere 10–20 months later. (See the dated receipts in EX-04.1.)

In plain words: on Raydium, people prove their share of a pool with LP tokens. Some burned theirs to signal “this money can never be touched” — and so it sits there, frozen forever, unusable by anyone. This patch gives Raydium alone a key to rescue that frozen money, with safeguards and a community vote. Two years later pump.fun shipped the same idea (“BOOST”).

Why this matters. Anyone can claim foresight. Below, every left-hand cell is a verbatim, dated message he wrote in the pump.fun team Telegram (Apr–May 2024) or a dated public PR — every right-hand cell is what shipped publicly afterward. Six for six. The seventh — the page they skipped — is the one he is building himself.

Source: verbatim dated messages, pump.fun team Telegram (Apr–May 2024), screenshotted 17 Jan 2025; ship dates are the public pump.fun timeline. He laid it out publicly himself — the “Pump’s roadmap reads like my 2024 notebook” thread (10 Jul 2026).