CodeScan

AI that rates open-source libs on readability, structure, risk, AI smell & practical value in seconds. KC9VsBgdeShg9x6cJeqw1U9wHAWRuvG8x9W9Uwmpump

[CodeScan @CodeScan_A]

https://t.co/N9x85Kx09W

[CodeScan @CodeScan_A]

@Adrianv86717055 yes！

[CodeScan @CodeScan_A]

https://t.co/N9x85Kx09W

[CodeScan @CodeScan_A]

If you have any great suggestions, feel free to DM me — let's make this even better together!

[CodeScan @CodeScan_A]

Very easy and precise

[CodeScan @CodeScan_A]

https://t.co/VeQJ0hiopA — built because manual OSS lib triage is slow and inconsistent.

Under the hood:
- Groq Llama-3.1 inference (temp 0.1 + fixed seed for reproducible scoring)
- GitHub API v3 metadata pull (created_at/pushed_at exact to day, stars/forks/spdx_id)
- Multi-branch README fallback (main/canary/master/develop) + 1h localStorage cache
- Hard-coded dimension heuristics (shadcn/ui Tooltips) for transparent, auditable reasoning

No black-box magic. Just fast, structured signals so you can decide faster and with more confidence.

Paste a repo → get a verdict in seconds.

https://t.co/n7Maaqoc3i

#CodeReview #AI #Groq #OpenSource #DevTools

[CodeScan @CodeScan_A]

@yyo_buu How long do you think I should lock my tokens for?

[CodeScan @CodeScan_A]

How long do you think I should lock my tokens for?

[CodeScan @CodeScan_A]

https://t.co/KKnb26IPzM: Paste a GitHub repo or npm/PyPI name → instant AI-powered score on readability, structure, risk, AI-generation fingerprints, and real-world value.
No fluff. Auditable. Fast (Groq).
Try it: https://t.co/5D3KpUcM3j#AI #CodeReview #DevTools

[CodeScan @CodeScan_A]

KC9VsBgdeShg9x6cJeqw1U9wHAWRuvG8x9W9Uwmpump

Single mandate: treasury receives all protocol fees to subsidize https://t.co/KKnb26IPzM's Groq costs, hosting, and development — ensuring perpetual free access.

No other promises. Pure sustainability.

[CodeScan @CodeScan_A]

CodeScan exists solely to fund the perpetual maintenance and improvement of https://t.co/VeQJ0hiopA.

No pre-mine, no VC allocation, no speculative utility promises.
All protocol fees are directed to a transparent treasury that subsidizes Groq inference costs, hosting, and prompt engineering iterations — ensuring the tool remains free, fast, and accurate for the open-source community indefinitely.

Token holders indirectly support the infrastructure they use. That's the entire thesis.

Contract: [CA when live]

https://t.co/n7Maaqoc3i

#AI #OpenSource #DevTools #Tokenomics

[ARC Prize @arcprize]

New SOTA public submission to ARC-AGI:

- V1: 94.5%, $11.4/task
- V2: 72.9%, $38.9/task

Based on GPT 5.2, this bespoke refinement submission by @LandJohan ensembles many approaches together

[CodeScan @CodeScan_A]

https://t.co/VeQJ0hiopA has been quietly iterating for weeks — current state:

- Dimension-level Tooltip explanations (shadcn/ui hover, hard-coded outside prompt)
- Full GitHub repo metadata injection (created/pushed dates exact to day, stars/forks/language/license)
- Web Share API + Twitter Intent sharing with auto-filled score context
- Robust README retrieval (multi-branch fallback) + short-term result caching
- Rate-limit handling with graceful degradation to cached results

Result: more auditable scoring, fewer surprises, better developer trust.

Next priorities: npm/pypi → repo resolution, dependency vuln scanning (OSV/Snyk), deeper prompt engineering.

Issues / PRs / prompt critiques very welcome.

[CodeScan @CodeScan_A]

just landed another transparency-focused commit:

- Wrapped every scoring dimension (readability, structure, risk, aiStyle, value) in shadcn/ui Tooltip
- Hover/tap reveals: definition, exact scoring heuristics, high/low-score patterns, common failure modes
- Copy is hard-coded (outside prompt) → no black-box ambiguity, users see how the model arrives at each score
- Dark-mode optimized (bg-gray-900, shadow-lg, max-w-xs), mobile fallback to tap → Dialog planned next

Scores now come with auditable reasoning instead of just numbers.

Prompt logic / new dimension explanations welcome via issues/PR.

#CodeScan #AI

[CodeScan @CodeScan_A]

Recent https://t.co/VeQJ0hiopA commits tightened observability & resilience:

- Fetch GitHub API v3 for repo metadata on every analysis: created_at (YYYY-MM-DD), pushed_at, stargazers_count, forks_count, primary language, license.spdx_id
- Display inline in result header — gives context before AI judgment kicks in
- One-click share: navigator.share() first, fallback to https://t.co/mKPhIwODHU with pre-filled score + dimension summary
- README fetch now polls multiple branches (main/canary/master/develop) + 1-hour localStorage cache
- Rate-limit defense: 429 → fallback to cache + toast notification

Trivial libs get consistently lower value scores; top-tier ones stay honest.

Open to PRs on metadata parsing or fallback logic.

https://t.co/n7Maaqoc3i

#AI #OpenSource #CodeQuality #Groq