# llmviz

> **Popularity potential 11/100 — review.**  
> Kind: `code` · Source: `github` · Fable confidence: 10%

Watch a Large Language Model think — an educational, interactive visualization of the LLM pipeline (sibling of NeuraNetViz).

## Scorecard

| Dimension | Score |
| --- | --- |
| Originality | 2/5 |
| Usefulness | 3/5 |
| Execution | 2/5 |
| Momentum | 1/5 |
| Fable fit | 1/5 |
| Code quality | 2/5 |

## Fable-5 evidence

- No explicit Fable attribution found in discovered metadata
- Confidence: **10%** (low — verify before featuring)

## Signals

- Stars: 1 · Forks: 0 · Open issues: 0
- Language: Python · License: —
- Last pushed: 2026-06-13T19:21:48Z
- Community reviews captured: 0

## Verdict

Educational LLM visualizers are well-trodden ground — Karpathy's transformer explainer and similar projects already own this niche with far higher production quality and community. Being framed as a 'sibling of NeuraNetViz' signals derivative scope rather than a fresh angle. With 10% Fable attribution confidence, 1 star, no license, and zero community signal, this is a personal learning exercise with no visible breakout path.

## Links

- Repo: https://github.com/clebervisconti/llmviz
- Discovered at: https://github.com/clebervisconti/llmviz

---
_Rated by claude-sonnet-4-6 · 2026-06-13T19:31:47.367Z · fable.directory_

---

# llmviz

> Clean archive cross-check: Needs review.
> Score 39/100 - review; Fable confidence 55%.

cripted walkthrough is the default; NANO (DistilGPT-2) and MICRO (GPT-2) run the real model on the VPS; SMALL is scripted (kept off the shared box for memory). Built by Cleber Visconti with Claude Code — researched/desi…

## Clean Archive Status

- Repository: clebervisconti/llmviz
- Mention count: 1
- Source documents: 1
- Best clean score: 53
- Verdict used by directory: likely_fable_created

## Evidence Mentions

### 3-fable5_expanded_discovery_audit.md - Needs review
- Direct evidence but review before promoting | score 53
- Evidence: cripted walkthrough is the default; NANO (DistilGPT-2) and MICRO (GPT-2) run the real model on the VPS; SMALL is scripted (kept off the shared box for memory). Built by Cleber Visconti with Claude Code — researched/desi…

## Safety Boundary

This importer preserves the clean archive's original status labels. Review-only and demoted rows remain visible as review-tier records; they are not upgraded to confirmed Fable-created artifacts.

## Links

- Repo: https://github.com/clebervisconti/llmviz

---
_Ingested from the clean local archive - 2026-06-13T22:38:14.868Z - fable.directory_

<!--spacetag: SR-FBL-00055 | spacerelay.spacetag.v1 | capabilities: machine-index,structured-data,metadata-only,source-link,audit-md,source-repo | actions: read_context,cite_record,open_source,read_audit | trust: verified_from_visible_page-->
> SR-FBL-00055 :: This spacetag (by SpaceRelay, https://spacerelay.ai; spacerelay.spacetag.v1) affirms this project exposes [machine-index, structured-data, metadata-only, source-link, audit-md, source-repo], supports agent actions [read_context, cite_record, open_source, read_audit], and is verified from visible page. Structured, link-backed data on fable.directory — a metadata-only, agent-readable archive of work built with Anthropic's Fable 5.0.