novamp
the token that stole the name is not the token you meant to buy
A clone resolver for pons v2 launches on Robinhood Chain.
Vamping is copying somebody else's launch: same name, same picture, minutes later, a wallet that has done it two hundred times before. On a chain sealing a block every 100 ms with tens of thousands of launches a day, a name that catches gets copied inside a minute, and most of what retail loses is not lost to the rug they bought. It is lost to buying number seven while reading number one's timeline.
novamp answers one question: out of every token fighting over this name, which one is real, and does being first still mean anything here?
No key. No signer. No transaction path. There is no
PRIVATE_KEYsetting in this repository and no code that would use one. CI fails the build if a signing primitive appears insrc/.novamp decides. Execution is deliberately somebody else's process:
swarm --exechands a finding to a program you wrote, holding your key, outside this tree. The tool cannot lose your money, which is the point of building it this way. See docs/EXEC.md.
git clone https://github.com/bored2boar/novamp && cd novamp
npm install
npm run doctor # config, fixtures and index, no network
npx novamp vamp NOVA --demo # a real clone fight, from the bundled fixtures
npx novamp board --demo # the same thing as a page on 127.0.0.1source demo 11 launches from 3 fixture file(s) · SYNTHETIC, not chain data
demo mode fixture data, not a live read of Robinhood Chain
NOVA · key "nova" FARM
3 launch(es) under this name · 2 copies · first copy 1m 3s after the original
# SYMBOL VERDICT SCORE RISK LAG TOP10 DEV SMART BUYERS TOKEN
─ ────── ───────── ───── ───── ─────── ───── ───── ───── ────── ─────────────
3 NOVA CONTESTED 92 GREEN +2m 28s 12% 2.8% 4★ 18 0x23ca88…8e12
1 NOVA TAINTED 0 RED first 48% 14.9% 0 4 0x23c8ef…1dc9
2 N0VA DEAD 0 RED +1m 3s 39% 11.2% 0 0 0x23c919…a740
→ first born is suspect: dev buy 14.90%; 7 wallets exempt from the opening tax;
deployer has 22 launches and no graduation; top 10 hold 47.5%; 1 other launch
in this cluster was funded from the same wallet
→ flow is on NOVA (rank 3), not on the first born
→ 2 launches in this cluster were funded from 0x2c88fa10…: one operator, several wallets
why NOVA scores 92:
-22 copy, 2m behind the first
+16 4 proven wallets, converging inside 27s
+5 18 distinct buyers in the first minute
+8 2.79 ETH of early buys
+6 early money spread out, top 3 are 30%
+10 dev buy 2.80%, inside the 1-6% band
+6 top 10 hold 11.9%, spread out
+14 deployer graduated 2 of 4 launches
+6 3 social link(s)
+8 graduated to the pool
That is the case worth building a tool for. The first launch under the name is the operator's own bait, the second is his too, and the one people should be looking at is number three.
| The problem | What novamp does | Command |
|---|---|---|
| fourteen tokens, one name, sixty seconds | groups every launch fighting over a name, matching on the ticker and the token name, with plurals, filler words, leetspeak and lookalike characters folded | vamp |
| "is this the real one?" | birth order by (block, txIndex, logIndex), then five labels: ORIGINAL, TAINTED, CONTESTED, VAMP, DEAD |
vamp |
| first is not always real | a first born loaded with a bundle, a serial deployer, or a wallet funded by the same hand as the copies is called TAINTED, not ORIGINAL | vamp |
| you are already in one of them | full breakdown of one token, plus the cluster it turns out to be number six of | scan |
| "how concentrated is this?" | top-10 holders with the curve and the pool excluded, green under 20 %, red over 30 % | vamp, scan |
| who actually knows something | wallets with a real record, and an alert when three of them land inside ninety seconds | smart, vamp |
| how bad is it out there | vamp ratio and vamp lag for the whole window | stats |
| a word is trending | which fights exist under it, from free RSS feeds | narrative |
| a name just caught fire and there are eleven of them | catches the burst, checks whether a headline in the last hour landed before it, ranks the field, and names the one that is not a copy - or says no pick | swarm |
| you are already holding some of these | balance-checks your wallet against the index and tells you which positions are copies, with the original beside each one | wallet |
| who keeps producing the copies | deployers ranked by copies shipped, names touched, and how many other wallets share their funder | farms |
| a terminal is the wrong shape for this | the same engine behind a page on 127.0.0.1, with filters and sorting | board |
| tell me when somebody copies my bag | watchlist plus Telegram, fired on a new live copy or a cluster flipping to CONTESTED | watchlist |
| "your scoring is wrong" | CSV with every reason and flag, so you can re-run it against your own weights | --csv |
| how far back can I actually look | a local append-only index that grows as you use the tool | index, --since |
| "the dev is dumping" | the sell side: whether the deployer sold and how fast, whether one sale took a bite out of the reserve, whether more is leaving than arriving | vamp, scan |
| a wall of tiny wallets faking flow | early money in quote, not in addresses, plus the share held by the three biggest buyers | vamp, scan |
Full list in docs/COMMANDS.md.
Nothing stops anyone reusing a ticker, so most copies simply reuse it, and catching those is string equality. That is not where a matcher earns its keep. It earns it on everything that is nearly the same, which is most of what an exact match drops on the floor.
A launch carries two strings a copy can borrow, and novamp indexes both:
# SYMBOL joined by
1 PEANUT first born, name "Peanut the Squirrel"
2 PEANUТ folded: the last character is Cyrillic Te, not T
3 PEAN0T folded: leetspeak, then one edit away
4 PEANUTCOIN folded: filler word stripped
6 PNUT different symbol, same token name
7 PEANUTS folded: plural
PNUT is the row that matters most in practice. Once the obvious ticker is
taken, a copy takes a different one and keeps the name, and a symbol-only
matcher never sees the fight at all. Membership is transitive: A shares a symbol
with B, B shares a name with C, all three are one cluster.
The lookalike-character case is real but it is the tail, not the headline. It shows up where a copy is trying to beat automated filters — scanner blocklists, alert bots, anything doing an exact string compare — rather than a human eye. novamp folds Cyrillic, Greek, fullwidth forms, zero-width joiners, accents and leetspeak because it is cheap to fold them, and every join is labelled in the table so you can see when it happened.
novamp swarm --watch
A clone war has a signature in time. For hours a name does not exist; then in a few minutes eleven launches appear carrying it, from wallets that have never touched each other. Nobody coordinates that. It is what a crowd looks like when it reads the same sentence at the same moment, and it is visible long before anyone knows which of the eleven is going to be the one.
PEANUT FARM · 6 launches in 7m 1s · 6 deployers · 0.9/min
burst started 2026-09-05 14:12:00 · the name did not exist before this
news Peanut the Squirrel seized by state wildlife officials, owner says
t.me/fastwire · 39m before the first launch · matched "Peanut" → PEANUT
# SYMBOL VERDICT SCORE RISK LAG TOKEN
─ ────────── ──────── ───── ───── ─────── ───────────
1 PEANUT ORIGINAL 96 GREEN - 0xa0c5…f6b5
2 PEANUТ VAMP 0 RED +41s 0xa1fd…2253
3 PEAN0T VAMP 0 RED +1m 36s 0xa253…736a
4 PEANUTCOIN VAMP 12 GREEN +3m 7s 0xa317…5565
5 РEANUT DEAD 0 RED +5m 42s 0xa4a4…d3c1
6 PNUT VAMP 0 RED +7m 1s 0xa581…a365
PICK PEANUT 0xa0c54ffbe2ea6f151468fd40d4281d807fc2f6b5
+16 4 proven wallets, converging inside 24s
+14 first launch under this name
+10 dev buy 3.10%, inside the 1-6% band
Three things about that headline check, because they are the parts that get
skipped. The item must have a timestamp — an undated one proves nothing and
is dropped, not assumed fresh. It must have landed before the first launch.
And when it did not, novamp says AFTER the first launch rather than quietly
counting it. A headline before a burst is a coincidence in time, not a cause.
Headlines come from free RSS and from public Telegram channels read through their
t.me/s/ web preview — no bot token, no API id, nothing joined.
When nothing clears the bars, the answer is NO PICK with the reasons and the
near miss. That is a normal outcome and it is most of them. The full rule set,
including why a copy can never win on points, is in
docs/SWARM.md; the --exec hand-off is
docs/EXEC.md.
novamp wallet 0xYOURS --since 7d
0xdeadBEEF…dDeEfF 2 of your 4 position(s) are copies · checked 11 launches · window 7d
YOU HOLD VERDICT RANK LAG SCORE RISK THE ONE YOU MEANT
PEANUТ VAMP #2/7 +41s 0 RED → PEANUT 0xa0c54f…f6b5
PNUT VAMP #6/7 +7m 1s 0 RED → PEANUT 0xa0c54f…f6b5
NOVA TAINTED #1/3 first 0 RED -
KETTLE ORIGINAL #1/1 first 75 GREEN -
Copies first, because the reason you ran it is the bad news. It reads; it cannot sell anything for you.
A block on this chain seals every 100 ms, so thirty days is around 26 million blocks and no public RPC will sweep that. Any tool offering a 30 day filter straight off an endpoint is either paying an indexer or lying.
novamp does neither. Every live run appends what it read to plain append-only
files under .novamp/, and --since 6h|24h|7d|30d|all is answered from the
accumulation. Day one gives you hours; a month of running gives you a month. When
the window you asked for is wider than the index covers, the command says how far
short it fell instead of quietly returning less:
! the local index only goes back 3d, 4d short of the 7d you asked for.
Keep running novamp and the window fills in.
novamp index shows yours. docs/INDEX.md has the rest.
Node 20 or newer. Two runtime dependencies: viem and commander.
git clone https://github.com/bored2boar/novamp && cd novamp
npm install
cp .env.example .env # only needed for live reads
npm run doctorEverything works with --demo and no .env at all. For live reads, point
RPC_URL at an endpoint you trust. The public Robinhood RPC works and rate limits
log reads hard; a private provider makes scan and stats comfortable.
This is v0.4. Being straight about the line is more useful than pretending there isn't one.
Proven. Name folding, clustering on both the ticker and the token name, birth
order, the taint check, the five labels, risk flags, the score, convergence, the
size-denominated flow and sell-side rules, the realized-results registry, the
operator ranking, the local index, the CSV export and the watchlist rules. All of it is exercised end to end
by --demo, which drives exactly the same code the live path drives, and all of
it is covered by 177 offline tests.
Run live. doctor --probe, vamp, scan, wallet, farms and stats
have been run against Robinhood Chain mainnet and returned real data, which
covers most of src/read/: launches, holders, buyers, sells and the funding
trace.
Written, not yet run live. swarm, whose live half also reaches news
sources over the network; watch and watchlist run, which are polling loops
around a reader that has run; and smart build, which rate limits before it
finishes on a public endpoint. If one of these throws on you, that is a bug
worth an issue.
Known gaps. Exempt wallets are not recovered from launch calldata yet, so a
live run underestimates bundles. Funding traces are bounded and often
inconclusive. A wide --since is only as deep as your local index.
All of it, in detail and without softening: docs/LIMITATIONS.md.
Every input the score reads is visible in the first two minutes of a launch: birth order, dev buy, exemptions, concentration, who bought and how fast. Two minutes of flow does not tell you what a coin does in an hour.
So the honest reading of a 90 is "if any launch under this name goes anywhere,
it is probably this one", and never "this one goes somewhere". novamp orders a
cluster. It does not predict a price, it has no buy button, and it is not going
to grow one. swarm --exec does not change that: it pipes a finding to a program
you wrote, and the score it pipes is still an ordering, not a forecast.
What would make the number worth more is sell-side flow, and what would make it trustworthy is testing every rule against what the chain actually did to each launch afterwards. Both are on the roadmap, and until that test exists this paragraph stays here rather than quietly disappearing.
Every rule, with its points: docs/RULES.md.
src/
├── chain/ constants, ABI slice, the RPC gate
├── read/ six readers: launches, holders, buyers, sells, funding, wallet
├── source/ live chain or fixture file, one interface
├── store/ the local append-only index that makes --since real
├── vamp/ confusables → normalize → cluster → flow → risk → potential → verdict → farms
├── smart/ the proven-wallet registry and convergence
├── narrative/ free RSS and public channel previews in, keywords out
├── swarm/ burst detection, news correlation, the selection rule
├── exec/ the hand-off to your own executor. Nothing here signs
├── alerts/ watchlist and one outbound POST to Telegram
├── export/ CSV, written by hand rather than pulled from npm
├── board/ node:http on 127.0.0.1 and one HTML file
├── ui/ table and render
└── commands/ thin wiring
fixtures/ what --demo reads, with an honest origin field
docs/ rules, limitations, architecture, commands, wallets, index, fixtures, swarm, exec
test/ 177 offline tests over the rules
One rule shapes the whole tree: nothing that makes a decision touches the
network. Reading is one layer, deciding is another, and they meet at a plain
object. That is why --demo is a real exercise of the tool rather than a mock of
its output. See docs/ARCHITECTURE.md.
- no signer, no key setting, no write path, not for convenience and not behind a flag.
swarm --execruns a program you wrote; novamp still holds nothing and signs nothing - no hosted service holding anything of yours
- no paid alpha channel, because a sharper private version means the public rules are worse on purpose
- no promise of returns, on the tool or on anything else
- no closed core: if a future page computes something this repository cannot, the repository is decoration
- no licence change
The most useful thing you can send is a launch novamp got wrong: a captured fixture plus one sentence about what the table should have said. See docs/FIXTURES.md and CONTRIBUTING.md.
Second most useful: a glyph the confusables table misses. That is a one line fix
in src/vamp/confusables.ts and a test in test/normalize.test.ts.
MIT. Runs on your machine, not anyone else's.