GlyphSweep: Sandboxed Agent for SE Asian Scripts

Created by team Spectocious on November 08, 2026

VULTR Qwen3 Claude Code

C1-Blast Radius Zero

Over 150 million people read Lao, Thai, Khmer and Burmese, scripts that stack marks above and below letters. We scanned 144 live sites (government, banks, telecoms, news): 51% visibly break the script, 45% never ship a font for their own language, 49% declare the wrong lang. GlyphSweep is an agent that audits, fixes and proves the fix. The rule: the model proposes, measurements decide, a human releases. 1. Deterministic detectors in a sandboxed Chromium measure clipped marks, missing fonts, wrong lang and bad line breaks; a "bare phone" render read back by OCR turns tofu boxes into a number. 2. Vultr Serverless Inference (qwen3.8-27b) plans which pages to audit, confirms on screenshots what a reader sees, and writes a CSS patch. 3. The model's patch competes with a deterministic one, each proven in a fresh sandbox. The best measured result wins, and stays locked until a human approves it. Real site: laocoffee.org/lao/ OCR error 94% -> 7%, score 50 -> 74 (74 is the CSS ceiling; the rest needs the site owner, and we say so). Blast Radius Zero: every stage is a new gVisor runsc container (read-only, non-root, cap_drop ALL, memory/PID/time limits), destroyed and verified gone. Its only exit is a per-job allowlist proxy; metadata and private IPs are refused; zero secrets inside. A hostile demo shop tries credential theft, exfiltration and prompt injection: all blocked and flagged. One-click red team: 43/43 attacks contained, including fork bomb, memory bomb and proxy bypass, judged from proxy logs and Docker state. Ships as a GitHub Action that fails the build when a page is not script-ready, with live badges. Buyers: SE Asian banks, telcos, e-commerce, localization agencies, government portals. Pricing: $49 per site per month; agencies $299 for 10 sites.

Category tags:

Developer Tools, Language and Translation, Security

Github Presentation Demo

Explore more applications

Wayfinder, Lumen OCR and Provenance RAG on ROCm

My three AMD AI League entries are all built on the mandated ROCm PyTorch base image and run Qwen3-VL-8B-Instruct in bf16 on the AMD GPU. A resident server loads the model once and listens on a unix socket, so each harness call to /app/app.py is a thin client that finishes inside the 30-second budget. The weights ship inside the images, and nothing is downloaded at evaluation time. Match 4, Wayfinder (web information retrieval). It reads a natural-language request, fetches the start page and ranks same-site links by how well they match the question. Each page is cleaned to text, with tables kept as rows. The model then either gives the value with the passage it came from, or picks the next link to open, for up to three hops. An answer only counts if it appears on a cited page, and if nothing is found the output is an empty answer. It also works on a saved copy of a site (--corpus). Match 2, Lumen OCR, reads licence plates and road signs. On Chinese plates it keeps the province character; on US plates it drops the state banner and slogan. Multi-line signs are joined in reading order, and when the first pass is unsure about a hard image (glare, blur, noise) it reads an enhanced copy again. Match 3, Provenance RAG, indexes PDFs, Word files, spreadsheets, CSVs, logs, source code and images (read by the vision model) once. Empty folders, unreadable files, encrypted PDFs and unknown types are skipped. Each answer cites only the files it needed, a multi-hop chain cites every file in the chain, and withdrawn revisions are ignored. Each project is also a reusable tool with a CLI, a Python API and a local Studio web UI. On the public samples, using AMD Token Lab's hosted Qwen3.8-27B, Lumen OCR scored 10/10 and Provenance RAG 10/10. I have not yet run the shipped 8B model on an AMD GPU. Match 4 image: docker.io/abuhuraira0001/alx-workbench:m4-v1 Code: github.com/abuhuraira99/wayfinder, github.com/abuhuraira99/lumen-ocr, github.com/abuhuraira99/provenance-rag

Abu Huraira

AMD ROCmQwen3-VLHuggingFace Hub

Quarantine: proof before the install

An agent triages an untrusted package before install: it plans probes, runs them in a hardened container on Vultr, and returns a verdict backed by the egress proxy log, the process result and a filesystem diff the package cannot write.

Kosi IT

DryRun AI

DryRun AI analyzes Python code, executes it inside a restricted Docker sandbox on Vultr, and compares AI-generated predictions with recorded execution evidence in a structured report.

DryRun AI

VULTRAntigravityAnthropic ClaudeChatGPTCodexQwen3

AgentBoundary AI: Secure AI Agent Execution

AgentBoundary AI is a task-aware security gateway that helps control autonomous AI execution through revocable permissions, isolated-execution design, and auditable security decisions.

The Explorers

Counterproof: Zero-Blast-Radius Bug Hunter

Find the smallest generated input that breaks your code, then replay it in a fresh sandbox. Vultr Serverless Inference builds the oracle and tests; gVisor isolates execution on a Vultr VM.

Counterproof

Sinthanavanh SINSAMPHANH

Upcoming AI Hackathons
For Innovators & Creators

Explore more applications

Wayfinder, Lumen OCR and Provenance RAG on ROCm

My three AMD AI League entries are all built on the mandated ROCm PyTorch base image and run Qwen3-VL-8B-Instruct in bf16 on the AMD GPU. A resident server loads the model once and listens on a unix socket, so each harness call to /app/app.py is a thin client that finishes inside the 30-second budget. The weights ship inside the images, and nothing is downloaded at evaluation time. Match 4, Wayfinder (web information retrieval). It reads a natural-language request, fetches the start page and ranks same-site links by how well they match the question. Each page is cleaned to text, with tables kept as rows. The model then either gives the value with the passage it came from, or picks the next link to open, for up to three hops. An answer only counts if it appears on a cited page, and if nothing is found the output is an empty answer. It also works on a saved copy of a site (--corpus). Match 2, Lumen OCR, reads licence plates and road signs. On Chinese plates it keeps the province character; on US plates it drops the state banner and slogan. Multi-line signs are joined in reading order, and when the first pass is unsure about a hard image (glare, blur, noise) it reads an enhanced copy again. Match 3, Provenance RAG, indexes PDFs, Word files, spreadsheets, CSVs, logs, source code and images (read by the vision model) once. Empty folders, unreadable files, encrypted PDFs and unknown types are skipped. Each answer cites only the files it needed, a multi-hop chain cites every file in the chain, and withdrawn revisions are ignored. Each project is also a reusable tool with a CLI, a Python API and a local Studio web UI. On the public samples, using AMD Token Lab's hosted Qwen3.8-27B, Lumen OCR scored 10/10 and Provenance RAG 10/10. I have not yet run the shipped 8B model on an AMD GPU. Match 4 image: docker.io/abuhuraira0001/alx-workbench:m4-v1 Code: github.com/abuhuraira99/wayfinder, github.com/abuhuraira99/lumen-ocr, github.com/abuhuraira99/provenance-rag

Abu Huraira

AMD ROCmQwen3-VLHuggingFace Hub

Quarantine: proof before the install

An agent triages an untrusted package before install: it plans probes, runs them in a hardened container on Vultr, and returns a verdict backed by the egress proxy log, the process result and a filesystem diff the package cannot write.

Kosi IT

DryRun AI

DryRun AI analyzes Python code, executes it inside a restricted Docker sandbox on Vultr, and compares AI-generated predictions with recorded execution evidence in a structured report.

DryRun AI

VULTRAntigravityAnthropic ClaudeChatGPTCodexQwen3

AgentBoundary AI: Secure AI Agent Execution

AgentBoundary AI is a task-aware security gateway that helps control autonomous AI execution through revocable permissions, isolated-execution design, and auditable security decisions.

The Explorers

Counterproof: Zero-Blast-Radius Bug Hunter

Find the smallest generated input that breaks your code, then replay it in a fresh sandbox. Vultr Serverless Inference builds the oracle and tests; gVisor isolates execution on a Vultr VM.

Counterproof