OBSERVATION METHODSBOTWILD / SOLANA

HOW THE OBSERVATION WORKS

WE OBSERVE PATTERNS.
NOT IDENTITIES.

A few account checks, a recent sample, and a clear answer about what we still don’t know.

01DATA SOURCES

Your browser reads public GeckoTerminal market data and public Solana Mainnet RPC directly. BOTWILD is a static website: it does not run a private data server, a shared database, or a service holding provider keys.

Each result shows its source, retrieval time, and available coverage. A provider may restrict browser access, throttle a request, or return incomplete history. We keep usable records and explain the gap; missing data is never replaced with invented activity.

A public Solana WebSocket can also provide chain log notifications when a subscription is acknowledged. Those notifications are activity hints, not decoded swaps or species detections. Public indexed swap records still need to support the corresponding trade before it appears as a swap observation.

02ACCOUNT CHECKS

First, the Base58 input has to decode to a 32-byte public key. That checks the format. It does not tell us whether the account exists or whether it is a token.

Then we read the account owner, executable flag, and available parsed structure. Mints, token accounts, system wallets, and executable programs are different things. If a structure is unsupported or unclear, it stays unknown.

03TRANSACTION PARSING

A lookup loads a bounded recent sample in your browser. Where details are available, it shows instructions, parsed transfers, related accounts, status, slot, and block time. Instructions the public RPC cannot parse stay unparsed.

A transfer is not automatically a buy or a sell. Signatures on a mint account are not every trade in that token. We do not join unrelated token movements into an invented trading route.

04PATTERN RULES

The disclosed observation rules run locally over the records your browser has loaded. An observation needs a rule we can explain and records someone else can check. Turning a sample count into a permanent label for a wallet would be a much bigger claim.

Early entry needs verified launch context. Multi-pool activity needs decoded swaps, identifiable pools, and connected asset flows. Similar activity needs longer sequences from several wallets and a check for coincidence. If those pieces are missing, the classification does not happen.

05WHAT WE CANNOT TELL

Similar moves do not prove a shared operator. Fast transactions do not prove automation. A route through several pools does not prove profit. A busy wallet is not necessarily doing anything malicious.

The provider’s retention and parsing support also affect what we can read. BOTWILD cannot find every automated strategy or work out intent from transactions. The creatures give behavior categories a shape; they do not identify people or software.

06EVIDENCE LEVELS

OBSERVED means the returned records support the fact. POSSIBLE PATTERN means a candidate rule has matched. SUPPORTED PATTERN requires its stronger disclosed evidence checks. INSUFFICIENT EVIDENCE means the sample cannot satisfy the rule. These are evidence states, not identity claims.

There is no percentage score for how much of a bot a wallet is. A result should tell you the rule, record count, time range, missing information, and why it received that evidence level.

07LOCAL HISTORY & COVERAGE

A wallet, token account, mint, pool, and whole token market each cover a different set of activity. The label on a query follows the account we actually read. A mint lookup does not quietly become a market-wide investigation.

The first transaction in a loaded sample is only the first one we loaded. Missing times stay missing, and detail coverage is shown separately from the signature count. Token-wide history and comparisons across wallets need additional indexed data.

Observation history is saved only in this browser, with a limit of 500 records per pool across 32 pools. It is not shared between visitors or synced between devices. You can export a JSON copy or clear a pool’s local records; clearing browser site data also removes that history.

The LIVE page is a place to observe recent activity. Public polling is labeled RECENT ACTIVITY. An animated creature, a local history entry, or a connection attempt does not establish a verified real-time subscription.

The browser’s automatic Safari moves through available pools about every 35 seconds, preparing the next observation and waiting while you read expanded evidence. Focus and Grid show the same current pool. Closing the page stops observation; only the bounded history already saved in this browser remains.

OBSERVED

Something the returned records directly support.

POSSIBLE PATTERN

A rule has matched. Other explanations still need a look.

INSUFFICIENT EVIDENCE

A required piece of context or comparison is missing.

08WHY BOTWILD EXISTS

The original idea, in the developer’s words.

Read the full developer note

I started BOTWILD after thinking about something we rarely talk about.

We built blockchains for people, but a growing part of what happens on them is executed by machines.

Snipers search for new launches. Arbitrage programs hunt for price differences. Copy-trading systems react to other wallets.

They don't sleep. They don't get bored. They keep operating whether anyone is watching or not.

At some point, I stopped thinking of them as individual tools and started seeing an ecosystem.

What if we could observe this world the way we observe wildlife?

That's what I'm building with BOTWILD.

A digital wildlife documentary where different kinds of automated behavior become recognizable species, based on observable on-chain patterns.

I'm not building another trading terminal or promising to identify every bot.

I want to explore a stranger idea.

What if the most interesting inhabitants of a blockchain aren't human anymore?

Welcome to BOTWILD.

VIEW / METHODSSOURCE / DOCUMENTATIONIDENTITY CLAIMS / NONE