---
name: spacepixel-wallet
description: The 14 Sep 2026 hunt for @spacepixel (pixel, 44.8k followers, Australian, "alien cat | profit maxi | warlord") on Robinhood Chain. Ben's candidate was disproved by a sleep-window test; a better candidate found via the LLM buyer set. Read before any further work on this handle or before using the sleep-window method.
sources: [cowork]
aliases: [spacepixel, pixel, "@spacepixel", llmtokens, launchLLMs, LLM token, sleep window]
---

- [stated] Ben: "Find his wallet based on his posts, coins, and timelines." He supplied a candidate from the DexScreener maker filter on the LLM pool: `0xdce001e4ec11ab58650c51ae9658accbe0f49c48`, and noted spacepixel follows @launchLLMs (llmtokens.fun), which "doesn't guarantee he has this token but is worth checking".

## The target
- @spacepixel, display name "pixel", created **2018-09-20**, 44,880 followers, following 285, 15,496 posts, blue verified, bio "alien cat | profit maxi | warlord".
- **AUSTRALIAN, self-disclosed 13 Sep 20:01**: bought weed on Silk Road for 1 BTC "back when I was at UNI", "would have been an australian seller", then "bought a bunch of Trezors and Ledgers in Australia and started reselling. Became biggest hardware wallet distributor".
- He names almost NO tickers. 86 posts in 8 days contain zero Robinhood Chain cashtags; `from:spacepixel 0x` returns one hit from 2022. The only stock tickers he names are US equities ($NCLD $NBIS $BE $EROC, 13 Sep 23:56). **A conventional "he shilled X on date Y" screen has nothing to bite on.**
- Thesis posts that DO give on-chain hooks: 12 Sep 23:43 "rumours that Google unlocked recursive self improvement using the **fly brains** neural network" · 13 Sep 00:37 "**Shill me your best ai gems on robinhood chain** - ai infrastructure, ai daos, novel" · 13 Sep 10:21 agent **swarms** · 14 Sep 00:23 "up big and have **shipped the profits to the motherland**" · 14 Sep 05:37 "**Bridging....**" and 05:49 "You'll buy when fomo integrates them".

## ⚠ fomo.family CANNOT name him, and that is itself the finding
- `fomo.family/profile/spacepixel` **exists but is empty: 2 trades, both about a year old (SOL and "memes"), $0 cash, 0 open positions, 2h10m avg hold, joined Sep 2025.** He does not trade through fomo, despite posting about fomo's social graph constantly. The usual "match the fomo holder name to the Blockscout balance" shortcut is dead on this target.
- Reading his profile while Ben is logged in shows BEN'S OWN book in the right-hand rail. **Do not mistake that panel for the target's positions.** It nearly caused a false match: Ben holds 5.3M LLM and 2.9M SCHIFFY, both tokens that also appear in the candidate wallets.

## ⚠ THE SLEEP-WINDOW TEST. The most useful thing built this session, and it is general.
- Build the target's posting histogram by UTC hour from twitterapi.io, then build each candidate wallet's transfer histogram by UTC hour, and compare. **A human trader cannot trade while asleep; a bot has no dead hours at all.**
- spacepixel's 86 posts since 7 Sep give a textbook Australian day: heavy 20:00-10:00 UTC (peaks 23:00 with 21 posts, 01:00 with 12, 22:00 with 12, 00:00 with 11) and a **hard ZERO across all nine hours 11:00 to 19:59 UTC** (21:00-05:59 Sydney). Nine consecutive empty hours, every day, 8 days running.
- Statistical note for future use: for a candidate with n legs, the null is (15/24)^n that all fall outside a 9-hour window. n=11 gives p≈0.009, so a single clean wallet is suggestive but **testing 30 candidates yields ~0.3 false positives by chance**. Report it as a filter, never as proof.

## VERDICT ON BEN'S CANDIDATE: `0xdce001e4ec11ab58650c51ae9658accbe0f49c48` IS NOT HIM. It is a bot.
- **44.6% of its 334 transfer legs fall inside his nine-hour dead window.** It has zero dead hours across all 24.
- **Mechanically regular intervals prove automation**: ZZZCAT inbound every 69/86/86/89/92/92/92/92 minutes, nine times straight through the night of 13-14 Sep; STONKINU every 102/103/103/103/98/103 min; DOGEGPT every ~85-89 min eleven times; EMBERCAT every ~134 min. No human clicks on that cadence at 3am.
- **85 tokens in 6 days** (first activity 8 Sep 15:21), much of it junk (BREWDOGE, GMECAT, ZDOGE, Robindoge, GMEDOGE, STUPIDINU, HONEYPOT, "if u buy this ur retarded"), with sub-hour round trips (PVE in and out in 45 min, SWARM 5h, BLAST 3.5h). That is scalping, and it contradicts his own stated style ("Positioning into a new meta > positioning into a coin of an existing meta", "A lesson in sizing", "weekly runners").
- Shape: EIP-7702 delegated to `0xe6cae83bde06e4c305530e199d7217f42808555b`, nonce 7, 0 ETH. Trades through router `0xb92fe925dc43a0ecde6c8b1a2709c170ec4fff4f` (4,720 bytes, 163 in / 117 out legs).
- **TRAP AVOIDED, worth keeping**: the ~90-minute inbound drips each come from a DIFFERENT 11,293-byte contract, one per token, which reads like a holder-reward distributor. It is not. Pulling all outbound logs from the DOGEGPT one (`0xdee9f0807979c070d08e65b51e9bf10a77b5b9b3`) shows 10,937 legs to 973 distinct addresses including the Uniswap v4 PoolManager: it is that launchpad's per-token router. **Before calling a periodic inbound leg a reward payout, count how many addresses the sender pays.**

## BEST CANDIDATE: `0x1050cecf1eec272e8ecd5e0dbfa8599407b7e416` — good fit, NOT confirmed
Found by screening the full LLM holder set (27,301 transfer legs, 880 holders), taking the top 30 EOAs, and ranking them by the sleep-window test.
- **Only 9 tokens in 4 days** (first activity 10 Sep 20:32), 37 legs. Selective, human-paced, nothing like a sniper.
- **Its book IS his narrative**: FLYBRAIN is the largest position (12 in / 9 out), LLM second (**8 buys, zero sells**), plus ChatGpt, PONS, CME, Rivra, CLAN, AAA, PETS.
- **The FLYBRAIN match is the strongest single piece of evidence.** The wallet started buying FLYBRAIN 10 Sep 20:33 and kept adding (11 Sep 09:16-10:06, 11 Sep 20:59, 12 Sep 06:19). He posted the fly-brain neural-network rumour **12 Sep 23:43**, i.e. accumulate first, narrate after. "Fly brains" is a rare enough reference that coincidence is unlikely.
- **The LLM match is dated and tight.** He posted "Shill me your best ai gems on robinhood chain" at **13 Sep 00:37**; the wallet bought LLM in four clips at **13 Sep 04:50-04:52**, added 3.05M on 13 Sep 16:36, then on **14 Sep 07:40-07:50 sold the last FLYBRAIN and rotated the proceeds straight into LLM**.
- **Sleep fit: 1 of 37 legs (2.7%) inside his nine-hour dead window.** Hour profile is 02:8, 04:4, 06:1, 07:7, 09:3, 10:1, 16:1, 20:9, 21:2 — his waking hours almost exactly.
- Plain EOA (code `0x`, NOT 7702), nonce 20, 0.00286 ETH left.
- **AGAINST**: the book is small, roughly $4-5k of LLM at screen; it has ZERO history on Base and Arbitrum (nonce 0 on both), which sits awkwardly with a 2018-vintage OG who resold hardware wallets; and it is one of 880 LLM holders, so the LLM leg alone is weak.

## RUNNER-UP `0x3e32b434a9d4ad1853244bec120244bdb56b8b33` — DROPPED at Ben's instruction, 14 Sep 17:13 UTC
- [stated] Ben: "Remove Pixel B wallet. I only care about the best-fit wallet." Removed from `/opt/pixelalert/config.json`; PIXEL ALERT now watches ONE wallet, `0x1050cecf1eec272e8ecd5e0dbfa8599407b7e416`. Backup `config.json.bak-dropb-20260914-*`, script `/root/pixel_dropb.py`. Re-add by pasting the wallet block back and restarting.
- Kept here in case it is ever worth revisiting: **0 of 11 legs in the dead window, 9/9 dead hours clean.** Holds 0.49% of LLM supply, first active 13 Sep 21:49. Has 21 transactions on Base (the only candidate with a real Base footprint).
- Why it was never strong enough: its entire book is LLM plus one BEAR and one ROCKET leg. Three tokens is not enough to match against anything.

## Other screened wallets, for reference
- `0x5f735a1dcaaea782cec6146b8ab5f851d0d64c3c`: 240 legs, 42 tokens, 22.5% in the dead window. Real trader-shaped (LDX, PICK, BRLA, FLY, FENT, MIRROR, DROYD) but too active overnight.
- `0x0e0cb46e1ddfcced88ab552871d88df52d61ad9e`: 95 legs, 7 tokens, but 33.7% in the dead window and 57 inbound SLV legs. A silver-token accumulator, not him.

## WHAT WOULD CLOSE IT
- He buys before he posts, twice demonstrated. **The next time he posts a specific token or a specific dated action, pull that token's buyers for the hours BEFORE the post and intersect with `0x1050cecf1eec272e8ecd5e0dbfa8599407b7e416` and `0x3e32b434a9d4ad1853244bec120244bdb56b8b33`.** Two dated intersections would settle it.
- His 14 Sep 05:37 "Bridging...." is an unexploited anchor: a bridge-out of Robinhood Chain in the 00:00-06:00 UTC window on 14 Sep, matching "shipped the profits to the motherland" at 00:23. Neither candidate shows a token leg then, so it needs native-ETH tracing, which needs a tx index (Blockscout is Cloudflare-walled, Dune's monthly datapoint limit is spent).
- Arkham remains the one purpose-built handle-to-wallet index and is still unreached.

## Reference
- LLM / "Large Language Model", llmtokens.fun, X @launchLLMs: token `0x6f1A924b217e7Bb254605662C4223b604b7E07ff`, deepest pool `0x2954c406b6dbe9cd025c27b0abeb5d2f6a3d16a95cb0d703514f5d2ec753cd70` (Uniswap v4, ETH quote), created 12 Sep 2026 20:58 UTC. At screen 14 Sep 16:16: mcap $412,637, liquidity $60,596, 24h volume $591,002, 2,168 buys / 1,964 sells. Pitch: "Launch tokens paired with any AI model. Trading fees become compute. Powered by @OpenRouter." @launchLLMs joined Sep 2026, 782 followers, follows exactly 1 account.
- Robinhood Chain block time measured 0.1021 s, so 6 days is ~5.2M blocks. Full wallet reconstruction is `eth_getLogs` on the Transfer topic with the padded address in topic1 and topic2 separately, 100k-300k block chunks, 8-10 threads. Blockscout stayed Cloudflare-walled all session.

## 15 Sep 2026 — "Do we have proof he trades these amounts?" Answer: NO, and he says so himself
- [stated] Ben asked, off his "multi 6 figures from every meta" post (14 Sep 23:34 UTC), whether there is any proof of the amounts or the profits.
- **HIS OWN WORDS KILL THE WHOLE VERIFICATION ROUTE.** 11 Sep 22:34 UTC: *"Feels good printing on fomo with an anonymous wallet. Buying on my public wallet ruins charts."* @123skely (36,986 followers) replied 50 min later: *"I'll share my secret fomo wallet if you show me yours."* He never did. **Trading through fomo custody means his position sits inside fomo's own wallet and never appears as a distinct on-chain address. The claim is structurally unfalsifiable, not merely unproven.** Record it that way.
- **THE FLYBRAIN TEST, run and failed.** His 11 Sep 22:40 UTC claim: *"Buy 2% of token that gets interest from marc andreessen. Help followers ape it < $1m mc. Print half a mil in 1 day."* Token is FLYBRAIN `0x4Eb990547BCe4a982432CA88Cf5fae7EED1A2d35` (a16z's Marc Andreessen followed the account; that is what drove the flow). Launched 10 Sep 20:11 UTC at $1.3M mcap, hourly ATH price 0.112581288946204 = **$112,581,289 mcap at 11 Sep 01:00 UTC**, supply 1e9.
- Method: all 53,555 Transfer logs from launch to 12 Sep 00:00 UTC, blocks 59,701,884 to 60,663,726, `eth_getLogs` in 40k chunks, 8 threads.
- Result: **exactly ONE address ever received 2%+ of supply** — `0x8876789976decbfcbbbe364623c63652db8c0904`, 37,990,910 tokens (3.80%), and it is a **24,546-byte contract**, an aggregator, net zero. Every other >1% receiver is a contract too, except `0xc87b51c71707cfdd08cd374a596a6946cb1d6a66` (EOA, **nonce 166,011**, a bundler) and `0x2378c9e49e2019d73499c4f3c86112607a7aa68e` (1.52%).
- `0x2378c9e49e2019d73499c4f3c86112607a7aa68e` screened and REJECTED: EIP-7702 delegated to `0xe6cae83bde06e4c305530e199d7217f42808555b` (the same implementation as Ben's original disproved candidate, so that delegate is a common smart-wallet, not an identity signal). 333 legs, 38 tokens in 5 days, **every single position round-tripped to zero**. Sleep test: **116/333 = 34.8% of legs inside his 11:00-19:59 UTC dead window**, and its own empty hours are 07:00-12:59 UTC, which is Sydney afternoon, not sleep. Not him.
- **The only hard PnL number he has ever published is small:** 16 Jun 2026, on @mnstr gacha, *"Out of $50k I've spent on MNSTR I've kept $15k of cards. My net PNL from the gacha is +$1500."*
- His other size claims are all unreceipted prose: "paid off house, million in stocks, passive income businesses" (13 Sep 00:33), "3rd cycle in a row pixel has sold the top and bought the bottom" (22 Aug), "I own 1-3% of the top 10 fomo traders" (5 Sep), "receipts from October 2025" (29 Jun, links to a tweet with no media).
- **ONE third-party corroboration of size exists, and it is on SOLANA, not Robinhood Chain:** @badattrading_ (67,761 followers) 8 Jun 2026 named "spacepixel" in the top-holder list of $POKETCG `N9Curnf2ZQWBZWrjBkzP6xBe6n5WRhBhouRfiSqpump`, and he posted 9 Jun *"Literally the top wallet on every one of these tcg coins."* **That is the best remaining identity lead.**
- Best RH candidate `0x1050cecf1eec272e8ecd5e0dbfa8599407b7e416` re-checked 15 Sep: nonce 20 to 22 in a day, 0.002716089867615779 ETH, still ~$5k of book. Effectively dormant and still unconfirmed.
- **NEXT STEP when this is picked up again:** pull the historical (June) top-holder set for the Solana TCG coins ($POKETCG, $CARDS, $GACHA) and resolve handles via GMGN's `common.twitter_username` block, which is populated (proved on Ansem: `blknoiz06`). Note `gmgn-cli portfolio stats --wallet` is documented as repeatable but **returned only the FIRST wallet** when passed ten; loop one at a time with ~8s spacing. The current pump.fun top-holders list is September state and the June cohort has largely exited, so historical holders are required.

## 15 Sep 2026, LATER — Ben pushed back, he was right, and PIXEL A now has a measured trade history
- [stated] Ben: *"Wait, don't we already have his wallet? Don't we have a Pixel watch, Pixel Alert? Can just run that through GMGN and tell me his trade history, right?"* **Yes. I had not tried it, and I should have.**
- **MY ERROR, standing lesson:** I told him GMGN returns an empty holdings list for Robinhood Chain wallets. **FALSE.** `gmgn-cli portfolio holdings` defaults to `--hide-closed true --hide-airdrop true`, which returns `{"list":[],"next":""}` on a wallet whose positions are all closed or airdropped. **Always pass `--hide-closed false --hide-airdrop false --hide-abnormal false`.** This is the "never assert a tool lacks a feature without opening it" rule, failed again, this time on a CLI flag rather than a URL parameter.
- **PIXEL A `0x1050cecf1eec272e8ecd5e0dbfa8599407b7e416` measured via GMGN, 30d:** realized **+$15,342.95** (+173.9%), 19 buys / 17 sells, 11 tokens, winrate 0.333, bought_cost $4,159.84, sold_income $24,165.11, current book **$3,320.91**, native 0.002716089867615779 ETH. **GMGN tags it `['fomo']`, which independently matches his own "printing on fomo with an anonymous wallet" line.**
- **THE FLYBRAIN TRADE, the whole point.** All Bangkok time.
  - 30 Aug 01:41-03:51: bought CLAN in 9 clips for **$2,604.60**.
  - 11 Sep 03:32-03:34: sold CLAN 1,201,922 for **$887.11** (realized **-$2,055**), and in the same two minutes bought **507,595.73 FLYBRAIN for $443.84** at $0.000864-0.000891, i.e. **mcap ~$870,000**.
  - 11 Sep 09:37-09:42: sold **530,339.83 FLYBRAIN for $22,316.36** at $0.0419-0.0441, mcap ~$42M. **Six hours, ~50x on the clip, +$21,872.** GMGN's FLYBRAIN realized_profit is $21,239.35.
  - 11 Sep 16:18 bought back 10,043.83 for $191.84 at $0.019. 14 Sep 14:40-14:44 sold 65,031 FLYBRAIN for $469.82 and rotated into LLM. 15 Sep 09:47 sold 10,095,171 LLM for $491.84 and bought back 94,702 FLYBRAIN for $472.23.
  - Sold more FLYBRAIN than it bought, so some arrived by transfer. Note it; do not treat the buy list as the full acquisition.
- **SCORING HIS CLAIM** (11 Sep 22:40 UTC: *"Buy 2% of token that gets interest from marc andreessen. Help followers ape it < $1m mc. Print half a mil in 1 day."*):
  - "< $1m mc" → **TRUE.** Bought at ~$870k.
  - "2% of token" → **FALSE.** 507,595 of 1e9 supply = **0.05%**, forty times smaller.
  - "Print half a mil in 1 day" → **FALSE.** $21,872. **Off by ~23x.**
  - "multi 6 figures from every meta" (14 Sep) → **not supported by this wallet.** Lifetime realized is $15,343.
- **STATUS OF THE MATCH: still a candidate, but much stronger than on 14 Sep.** The sleep window fitted, the FLYBRAIN entry matches the shape and the timing of a claim he made the next day, and GMGN independently tags the wallet `fomo`. What it does NOT explain is the size gap: either this is one of several wallets, or it is not him, or he exaggerates by roughly 20 to 25x. **Do not upgrade it to "confirmed" without a second dated intersection.**
- Method note worth keeping: `gmgn-cli portfolio stats --wallet` is documented as repeatable but **silently returns only the FIRST wallet**. Loop one at a time, ~8s apart (429 RATE_LIMIT_BANNED escalates).