Roam
Sweepers visit sources inside the Solana ecosystem one tab after another and capture what is genuinely on screen, not a thin summary of it.
~/sweepnet ❯ watch --chain solana
Autonomous Sweepers wander the open corners of Solana, take in what the ecosystem has written, and fold it into clean study material for an AI built to understand one network in depth.
01 sweepers
Each Sweeper behaves like a patient scholar holding a genuine browser. It does not guess. It opens a source, studies it, and hands back only what survived cleaning.
Sweepers visit sources inside the Solana ecosystem one tab after another and capture what is genuinely on screen, not a thin summary of it.
Each capture is scrubbed, de-duplicated and cut into study-sized passages with its origin attached.
The Core studies the refined set. Once a cycle is financed, a new pass begins over a wider body of material.
What the Core learns is meant to be inspected. Methods and outputs are written down for anyone to check.
| lane | reads | hands back | status |
|---|---|---|---|
| protocol notes | specs, guides, reference material | plain passages | scoped |
| proposal threads | proposals, votes, rationale | decision records | scoped |
| source repositories | readmes, issues, change logs | annotated snippets | queued |
| builder forums | questions, answers, post-mortems | worked examples | queued |
| program metadata | open interface descriptions | structured entries | queued |
02 board
Pair activity on Solana, ranked by what changed hands over the last day. Values refresh on their own and rest while this tab is out of view.
| # | pair | price | 24h volume |
|---|---|---|---|
03 network
Depth beats breadth. Sweepnet studies a single ecosystem so closely that its Core can answer in that ecosystem's own terms.
04 roadmap
Phases are ordered by dependency, not by calendar. Each one opens only when the one before it has produced something worth keeping.
Chart the sources worth studying and set down the rules a Sweeper follows before it touches any of them.
Bring the opening Sweeper families online and grow the refined set lane by lane.
Complete the opening training cycle and publish how the Core did against plain questions about the chain.
Release methods, evaluations and results so anyone can repeat the experiment or challenge it.