Live directory

Rank it before
the judges do

See the field before the only submission that counts. DryRun reads the public directory, scores public repositories on their content, places yours on a speculative board, and audits your live site for AI readiness.

Check my repo, free See the field
No sign-upRepos are read, never executedPay only for a deep review

Speculative board live

  1. Waiting for scored repositories.

Track share live

…entries
    Top content score…Waiting for the board.
    Most crowded tracks
    GET /robots.txt live
    …
    Tracks with room…under 12% of entries each
    Repos matched / scored… / …

    Reading the directory.

    Speculative reading of the published rules against public repositories. Not an official result and not affiliated with the organizers.

    Here is how
    DryRun reads
    a repo

    Read the directory

    One page after another, with a pause between them. A snapshot is kept when the count changes.

    Score the content,
    not the files

    Clone, read, delete. README substance, real code vs scaffolding, tests that assert, NEAR used in code, commit spread, a demo link that opens. File presence and rule hygiene are 10 of 100 points. Nothing is run.

    Place it on the board

    The rank is among the repositories actually scored, with the fixes in order.

    Audit the live site

    Crawler access for 17 AI crawlers, readable HTML, JSON-LD, llms.txt and sitemap, metadata, trust.

    How the content score is weightedpoints of 100
    • Code that is real, not scaffolding25
    • README explains problem, setup, demo22
    • NEAR used in code15
    • Tests that check behaviour10
    • Work spread over time10
    • Rule hygiene (files, license, secrets)10
    • A demo link that opens8

    lib/quality.js · lib/assess.js

    The field,
    as far as
    public pages
    go

    Pulled from the public directory about every 10 minutes. The numbers on this page are live.

    Open the field
    … entriesin the DryRun reading of the directory

    Generic words like "agent" or "NEAR" barely count. Every match says which phrases hit.

    IDEA CHECK · BM25 + MODEL RE-RANK     … scored… matched to public repos

    With a NEAR AI Cloud key, the AI judge cites file paths and lines. An uncited score is capped at 2 of 5.

    AI JUDGE · DOES NOT MOVE THE BOARD     17 AI crawlerschecked in robots.txt, plus a live firewall test

    File presence alone no longer scores 100. A template README earns almost nothing.

    FREE CHECK

    A triage queue
    for organizers

    Paste up to 20 public repository URLs. Each one is cloned, read, and deleted. You get rule failures in triage order, duplicates, the content score and an AI-ready column.

    Open the triage queue

    What the queue puts first

    • failWill not clone
    • failPossible secrets in the tree
    • warnMissing license
    • warnREADME commands that do not resolve
    • warnCommits outside the window
    • thenDuplicates, content score, AI readiness

    Pricing

    Free to check.
    Pay per repo
    for depth.

    Every builder tool on this page is free. The deep review is a paid read of one repository, settled in USDC on NEAR through NEAR Intents.

    Builders

    $0forever
    No account, no wallet
    Check a repo
    • Content score: README, code, tests, NEAR in code, history
    • Submission form and demo post checks
    • Track crowding and name clusters
    • Speculative leaderboard
    • Estimated rank and a fix list
    • Idea check against the live directory
    • AI readiness audit of your live site

    Deep review Most useful

    …per repository
    Paid from any supported token via NEAR Intents
    Run a deep review
    • Everything in Builders, plus:
    • A model on NEAR AI Cloud reads the README against the code
    • Startup lens inspired by Garry Tan's published questions
    • Claims in the README checked against the tree
    • A prediction of where a fresh clone fails
    • One payment, one repository, no wallet connection

    Counted only when the payment settles in USDC on NEAR.

    Organizers

    $0in this build
    Up to 20 repositories per run
    Open the triage queue
    • Paste a list of repository URLs
    • Rule failures in a triage order
    • Duplicate URLs flagged
    • Content and AI-ready columns
    • Larger runs and other hackathons Coming soon
    • Rulebook and directory adapters per event Coming soon

    Questions

    Is the leaderboard the official result?

    No. It is a reading of the published rules against the public repositories this process could match and score. Judges decide. The order will move as more repositories are scored.

    Do you run the code in a repository?

    No. A repository is cloned, read, and deleted. "Commands resolve" means the scripts and files named in the README exist. It does not mean they succeed. Symlinks are skipped. Secret values are not printed.

    What does a match mean?

    High confidence means the first GitHub link in that project's public demo post. That link can still be a dependency rather than the submission. Medium confidence means a GitHub name search cleared a conservative threshold, and it can be the wrong repository. Entries with no match are left unscored.

    What if my repository is still private?

    It cannot be scored, and the screeners cannot open it either. Make it public, then paste the link.

    What does the deep review cost, and what if there is no API key?

    The price on this page comes from the server configuration. Without NEAR_AI_API_KEY the free checks, the board, rank, and the idea check still run. The model is not called. A gateway signature, when the API returns one, is shown with the kind the API sends. It is not described as an enclave attestation.