tekmaker
← all projects

Clocked

The trenches lie detector: is that chart's volume and holder base real?

Open the app → Read the research
Token
$CLOCKED, reserved name, not launched. Planned venue: pump.fun.
Problem
Pump.fun traders pick coins off volume ranks, holder counts and fast charts, and all three are routinely faked. On 2026-10-09, 9 of the top 10 Solana pools by raw volume were farms, and 29% of new graduates were instant self-bonds.
What the MVP does
Paste a coin and in about 5 seconds get a 0–100 realness score with reasons. It covers wash and robotic trading in the last 300 trades, turnover and buyer:seller skew, holder concentration and near-identical bags, fresh wallets, shared funders traced from each top wallet's first transaction, and the launch pattern. A backend snapshots the board every 10 minutes for a Farm Index history and keeps a verified public scan feed.
Token role
Creator fees pay Clocked's data bill (RPC credits, market data, an indexer). Holding $CLOCKED unlocks watchlist alerts, deeper two-hop traces and higher limits, checked read-only with no signing.
What works today
  • Realness score (0–100) with CLEAN / MOSTLY REAL / SUS / CLOCKED / FULLY FARMED verdict and plain-English reasons
  • Live-tape forensics on the last 300 trades: churn, top-5 share, identical-size bots, dust trades
  • Money trail: first-funder tracing for top holders and traders, shared-funder clusters, dev-funded wallets, coordinated funding through busy wallets
  • Holder forensics with pools, curves and program vaults removed (on-curve check), near-identical bags, fresh wallets, dev bag
  • pump.fun launch checks: instant self-bond, serial deployer, transfer tax, authorities, risky Token-2022 extensions
  • Real Board: trending and top-volume pools re-ranked by distinct real traders with REAL / SUS / FARM badges
  • Farm Watch: server snapshots the board every 10 minutes and tracks the Farm Index over time (share of top-20 raw volume that's farmed), plus a 24h Hall of Shame
  • Recently clocked: a public feed of scans that the server re-verifies itself (clients can't post scores), with instant cached reports for shared links
  • Share card PNG, deep links (?ca=), and a prefilled X post
Shipped 2026-10-09 · safety · factory conviction 8/10 · data: Solana RPC via /api/rpc (Helius), Helius DAS, GeckoTerminal, DexScreener, pump.fun public coin data, Clocked backend (/p/clocked/api: board snapshots, history, verified scans)

TL;DR

On pump.fun in October 2026 the contract-level rug checks are solved and useless: almost every coin has mint and freeze authority revoked and its liquidity locked in the curve or PumpSwap pool, so RugCheck-style scanners say "good" to nearly everything. What actually burns traders now is manufactured activity: wash-traded volume, sybil "buyer" swarms, bundled launches that self-bond in the first minutes, and fresh wallets holding the supply on behalf of one operator.

Clocked answers the question degens actually ask before aping: "is this chart real?" Paste a coin and in about 5 seconds it reads the live tape, the 24h market, the top holders, who funded those holders' wallets, and the launch curve. It returns a 0–100 realness score with plain-English reasons and a share card. A free Real Board re-ranks today's trending Solana pools by distinct real traders instead of raw volume. On the day we built it, 9 of the top 10 pools by raw volume were farms.

Token: $CLOCKED (reserved, not launched). Creator fees fund the data bill. Holding unlocks alerts and deeper traces. The project page will publish Clocked's own scan of $CLOCKED, live.

The user and the pain

Who: the pump.fun and PumpSwap trader who buys coins between $50K and $5M market cap off trending lists, X posts and Telegram calls, usually on Axiom, GMGN, Photon or a Telegram bot, with a burner wallet and a 0.2–2 SOL preset.

The pain: every signal they use to find coins is gameable and gamed.

  • Volume and trending ranks are bought with volume bots. A round-trip buy and sell counts twice and costs the farm almost nothing.
  • Holder and trader counts are padded with thousands of sybil wallets that buy dust and never sell.
  • Launches are bundled. The dev and their wallets take the curve in the first block or minutes, the chart prints a fast "ATH", and the bundle unloads on whoever arrives next.
  • The free tools don't say it plainly. Terminals show "bundlers %", "insiders %" and "top 10 %" as raw numbers with no verdict. Bubblemaps shows a graph you need to interpret. RugCheck focuses on authorities and LP. Nobody answers "is this volume real?" in one line you can screenshot.

The moment of use is short and emotional: a coin is moving and they have 30 seconds. Then, after a rug, they want to post proof ("clocked: 9 of the top 10 holders were funded through the same wallet").

Evidence

  • Volume is widely faked, with forensic proof. Bitquery traced Solana DEX data from late April 2026 to one wallet that seeded 200 bots with 0.5 SOL each in 52 seconds. The bots wash-traded a single memecoin to more than $500K of fake volume in 12 hours. Bitquery's detection used trades per trader, trade size, two-sided wallet overlap and tracing each wallet's first inbound SOL, which is the same family of signals Clocked computes. (Bitquery)
  • Scale claims are large, if contested. A Flip Research report summarized by Benzinga argues up to 93% of Solana trading activity may be non-organic. (Benzinga) An earlier case study found 91% of a DexScreener top performer's $3.7M volume was likely wash. (elovianoo)
  • Coordinated early-buyer groups are persistent. A 2026 study cited by PumpView identified 1,012 wallet groups of 2–12 wallets that repeatedly buy early across launches, from data on 160,000+ pump.fun launches. (PumpView)
  • Bundling is an industry with stealth products. "Best Solana bundler in 2026" comparisons are published openly (The Coin Republic, 2026-09-25), and open-source "stealth bundlers" advertise that they're undetectable on DexScreener, GMGN and Bubblemaps (GitHub). Detection is an arms race, and a free referee has an audience.
  • Losses are the norm. Talos found 81% of a sampled set of memecoins fell at least 90% from peak, while pump.fun's protocol revenue held at $18.6M for the week to October 7, 2026. (CoinMarketCap) The widely repeated "98.6% of pump.fun tokens are rugs" figure is really a liquidity-below-$1,000 proxy, which shows how badly traders want a better signal. (CoinEdition)
  • Naive signals fail. A 2026 dataset on "dev-wallet silence" as a rug predictor found it doesn't work: 96% of rug dev wallets went silent, but so did 99% of the controls. (Zenodo) Traders need composite, explained signals, not one heuristic.
  • Existing checkers have known blind spots. A 2026 review of seven Solana rug checkers found they can't catch slow soft rugs where contract flags stay clean, and that Birdeye lacks holder data in the first ~10 minutes. (uwuu.ai)

Market now

Live numbers from the factory's pump.fun digest (2026-10-09 07:30 UTC) and Clocked's Real Board (about 12:30 UTC the same day):

  • Launches: about 28,500 per day. Graduations of coins created in the last 24h: 1,050, a bond rate of about 3.7%.
  • Instant self-bonds (ATH within 2 minutes of creation, then −50% or worse): 305 of 1,050 graduates (29%). Only 51 organic graduates were still at ≥50% of their ATH.
  • Raw volume #1 on the board was a pool trading $151M in 24h on a $4.97M market cap. By real traders it ranked #23. A clone "TikTok" pool did $29.7M of volume on a $206 market cap. Clone pools wash-traded $33.6M and $26.9M under the ticker of a real runner (PQC). 9 of the top 10 pools by raw volume carried farm signatures.
  • Farm fingerprint: many farmed pools show almost exactly ~2,000 buyers vs ~35 sellers in 24h, the output of a sybil tool, not a market.
  • Real runners exist, and the board surfaces them: the top real-trader coins that day had 9K–27K distinct wallets and normal 0.4–10× turnover.
  • Narrative lift: among organic graduates, product words are over-represented ("app" lift 10.3, "trading" 3.2, "research" 4.1), so a working tool with a token fits what's bonding.

Context: pump.fun made $18.6M in protocol revenue in the week to October 7 and has spent about $476M on PUMP buybacks. Memecoins were about 42% of Solana DEX volume in August. (CoinMarketCap, CryptoTicker)

Competitors and the gap

Tool What it does Gap Clocked fills
GMGN / Axiom / Photon Terminals show bundlers %, snipers %, insiders %, top-10 %, dev holdings Raw numbers with no verdict, inside a trading UI. No volume-reality check on the tape, no shareable result.
Bubblemaps Visual wallet clusters; paid tiers and its own token Great graph, but slow to read in 30 seconds. Doesn't judge volume or trades per wallet; reading it takes skill.
RugCheck Authorities, LP, Token-2022 risks, insider networks Most pump.fun coins pass. Says little about wash volume or sybil buyers.
Trench.bot data (Qocaine extension) Bundle % overlay on terminals Bundle-only; the extension listing is ~1.7 years old (Chrome Web Store).
Raiden terminal Paid API for bundle % and funding clusters (blog) API key, paid, developer-facing.
DexScreener / GeckoTerminal trending Lists coins by volume and boosts This is the gamed surface. Clocked's Real Board is its honest mirror.

The gap: one free page that combines tape forensics (wash, churn, robotic sizes, dust), market reality (turnover, trades per wallet, buyer:seller skew), holder forensics (concentration, equal bags, fresh wallets, dev bag) and a first-funder money trail, and returns one verdict in plain words with a share card. Plus a public board that shames farmed volume.

The MVP

Live at /p/clocked/app/. Paste a CA, a pump.fun link or a DexScreener/GeckoTerminal link; results deep-link as ?ca=<mint>.

What it checks, about 5 seconds per coin:

  1. Market: all pools for the token (GeckoTerminal, DexScreener fallback). Volume vs market cap, trades per wallet, buyers vs sellers. Thresholds are doubled for pools under 48h old, because day-one turnover is legitimately high.
  2. Live tape: the last 300 trades on the main pool. Wallets churning both sides (≥2 buys and ≥2 sells), top-5 wallet share, wallets repeating identical sizes (within 3%), and the share of trades under $2.
  3. Holders: top 20 token accounts with owners resolved. Pools, bonding curves and program vaults are removed using known pool keys, account owner programs, and an ed25519 on-curve check that separates PDAs from wallets. Then top-10 concentration, near-identical bags, and the dev bag.
  4. Money trail: for the top 10 holder wallets and top 6 traders, read up to 1,000 signatures, find each wallet's first transaction and its funder, then mark wallets under 3 days old. Shared private funders become clusters (union-find, including holder-funds-holder chains), and creator-funded wallets are flagged. Funders with 1,000+ transactions are treated as exchanges or services, and only count as coordination when 3+ wallets were funded through them within a 12-hour window. In testing this caught a live coin whose 12 top wallets were funded through one high-throughput wallet within 6 hours, each holding 0.16–0.24%.
  5. Launch: for real pump.fun launches only: minutes from creation to ATH and the drawdown since (instant self-bond), the creator's number of launches, transfer tax, a pump.fun ban, live mint or freeze authority, and risky Token-2022 extensions.

Score: starts at 100. Penalties are capped per category (volume 60, holders 45, launch 40). The verdict bands are CLEAN, MOSTLY REAL, SUS, CLOCKED and FULLY FARMED. Scans missing market, tape or funding data are marked partial and capped below CLEAN. Mature large coins (more than 14 days old and above a $20M valuation) don't get penalized for market-maker or arbitrage tape.

Calibration run (live, 2026-10-09): a wash-farmed coin (volume 985× market cap, 36 trades per wallet, 89% in one wallet) scored FULLY FARMED or CLOCKED. A bot-flagged $150M-volume pool scored SUS. pqc.market, a real runner, scored 94 CLEAN. WIF scored 90 CLEAN after the mature-coin rule.

Backend (v2): a sandboxed worker at /p/clocked/api snapshots the board every 10 minutes into SQLite. It serves the Real Board to every visitor from one cached snapshot, which spares the shared GeckoTerminal quota, and tracks the Farm Index over time: the share of top-20 raw volume that's farmed. In its first snapshot that was 97%, with 10 of the top 10 farmed. It also keeps a 24h Hall of Shame, and runs its own verification scans for a public "recently clocked" feed: visitors only queue a mint, and the server computes every score shown. Fresh verified reports load instantly on shared links.

Data: Solana RPC through the factory's read-only proxy (Helius), GeckoTerminal (pools, trades, trending), DexScreener (fallback), pump.fun public coin data, and Helius DAS for metadata. No wallet connection and no signatures.

Limits: heuristics, not proof. First-funder tracing stops at 1,000 signatures, so old wallets read as "established". CEX-funded farms are only caught by the time-window rule. GeckoTerminal's free tier is shared through one server IP, and a stale cache covers bursts.

Token design

$CLOCKED is reserved, not launched, with pump.fun as the planned venue. The token has jobs that already exist in the product:

  • Pays the data bill: pump.fun creator fees fund RPC credits, a paid market-data tier and an indexer as usage grows. This is said plainly on the page.
  • Holder tier (read-only check, no signing): wallets holding a set amount unlock watchlist alerts (Telegram pings when a watched coin's realness drops or a new funding cluster appears), deep traces (top 50 holders, two-hop funding), and higher rate limits.
  • "We clock ourselves": the project page shows Clocked's live scan of $CLOCKED. A public dev wallet, no bundle, and a disclosed dev buy make the trust story the marketing.

What it is not: no staking yield, no revenue share, no promises about price.

Launch plan

  1. Pre-launch (3–5 days): post daily Real Board screenshots ("today's #1 by volume is a $206 market cap farm") and reply to trending-coin threads with share cards. Seed 10–20 "clocked" receipts on coins that later dumped.
  2. Launch on pump.fun in a US-afternoon window (14:00–17:00 UTC) with a small disclosed dev buy from a single wallet, no bundle. Pin the token's own Clocked scan.
  3. First 30 minutes: a livestream clocking coins chat requests. Each viewer request becomes a share card.
  4. Week 1: a Telegram bot (/clock <CA>), a public "farm fingerprint" thread (the 2,000-buyer / 35-seller signature), and a weekly "most farmed" leaderboard.

Roadmap (if it gets traction)

A Telegram bot, then a browser extension overlay on pump.fun, DexScreener and Axiom. Watchlist alerts. Multi-hop funding graphs. A historical database of farm fingerprints by tool. A paid API for terminals and call groups. Our own indexer for trades, to drop the GeckoTerminal dependency.

Risks and kill criteria

  • False positives on legit coins: burner-wallet culture makes "fresh wallets" common. Mitigations: explained signals, softer weights, and an appeal link to report misreads.
  • Data limits: GeckoTerminal's free limits are shared on one IP, and Helius credits cost money. Needs a paid tier or an indexer once usage passes a few thousand scans a day.
  • Adversarial adaptation: farms will fund through exchanges and spread timing. The time-window and equal-bag rules partly cover this; it's an ongoing arms race.
  • Incumbents ship it: GMGN or Axiom could add a verdict line. Clocked's edge is being neutral, free and shareable, plus the board.
  • Reputational and legal: language stays "signals" and "looks", never accusations of intent.
  • Kill criteria: fewer than 500 scans a week after two weeks of posting, or persistent false positives on clearly organic coins that the method can't fix.

Sources