Safecell
An agent that only chats is a demo; an agent that executes is a liability. SafeCell is a web app where a model writes a Python program to do your task and that program runs inside a fresh, disposable cell: its own PID and network namespaces, no access to the host filesystem, no egress, and hard limits on CPU, memory, file size and process count. The container is destroyed after every step, so nothing carries over. Output is captured with a byte cap and scrubbed for credential shapes before it reaches the browser. Each run declares a permission manifest generated from the same policy the sandbox enforces, so what is declared cannot drift from what is enforced, and that manifest's hash is carried in the run receipt. Every event hashes the previous one, so a receipt can be re-verified rather than trusted, and the service refuses to execute at all if the isolation primitive is unavailable. 25 tests cover the boundary, none of them mocking isolation: host paths absent, host writes refused and verified absent, TCP and DNS both failing, secrets never crossing in, key-shaped output redacted, and every resource limit biting. Runs stream live over SSE with the plan, the code, the real output and the receipt. Deployed on a Vultr VM behind Caddy with TLS, with the planner calling Vultr Serverless Inference through any OpenAI-compatible endpoint.
Category tags:AetherTactics: Low-Compute Gemini Agent
AetherTactics is an autonomous strategy AI agent engineered for the AMD AI League Match 6 challenge: "Can Your AI Master the Unknown?". The agent navigates an unfamiliar, procedural fog-of-war grid, evades environmental hazards, recovers mission-critical alien relics, and safely reaches extraction points while aggressively minimizing compute overhead and token consumption. Powered by Google Gemini (supporting Gemini 2.5 Flash, 3.1 Flash-Lite, and custom endpoints), AetherTactics utilizes an ultra-compact radar serialization protocol. Instead of verbose context passing, it translates 2D exploration states into dense, minified matrices, constraining LLM reasoning to strict micro-JSON directives under 60 output tokens. To prevent agent freezes during live competition, the platform incorporates an automated watchdog failover pipeline. If a model encounters 503 high-demand or rate-limit spikes, the system automatically executes exponential backoff and fails over to stable alternatives without interrupting autonomous navigation. Built-in live telemetry monitors prompt tokens, generation tokens, step latency, and turn efficiency to deliver state-of-the-art decision-making on minimal inference budgets.
Solo Builder XP
VeriFetch ROCm: Cite-or-Refuse Web Agent
VeriFetch ROCm — Cite the Exact DOM Substring, or Refuse to Guess. Built for AMD AI League Match 4 (Intelligent Web Information Retrieval), VeriFetch ROCm is a high-precision web and HTML table value extraction agent engineered to run locally on AMD Instinct MI300X (192 GB HBM3) inside the official rocm/pytorch container with zero paid external API keys. WHY STANDARD WEB RAG AGENTS FAIL When navigating dense technical specification pages, OS compatibility matrices, or financial filings, standard web retrieval pipelines break in three critical ways: 1. Naive HTML stripping (such as BeautifulSoup .get_text()) destroys <table> row and column alignment, merging numbers across columns (for example, confusing Peak FP16 Dense TFLOPS with Peak FP8 Sparse TFLOPS). 2. Cold-loading a 4B or 8B parameter LLM on every CLI invocation consumes 14 to 18 seconds of the strict 30-second per-query evaluation budget. 3. Generative models paraphrase or guess when a metric is absent, failing exact-match automated grading. HOW VERIFETCH ROCM SOLVES THIS ON AMD INSTINCT MI300X 1. Resident Unix-Socket ROCm Daemon (/app/daemon_rocm.py): Boots once inside rocm/pytorch and keeps our extraction engine warm in HBM3 VRAM behind /tmp/verifetch.sock. Every CLI invocation (/app/app.py --query-id <id> --query "...") completes in ~0.42 seconds—35x faster than cold model boot. 2. Header-Contextualized DOM & Table Linearizer (/app/extractor.py): Parses HTML <table>, <dl>, and JSON-LD structures into self-contained "[Row Header | Column Header]: Cell Value" tuples before hybrid scoring. 3. Deterministic Cite-or-Refuse Substring Verifier: Before writing /app/output/<query-id>_output.json, VeriFetch verifies character-for-character that the extracted atomic value exists in the raw source DOM. If unverified, it deterministically abstains (answer: "", citations: []), achieving 100% exact-match accuracy and 0.0% hallucination on adversarial traps.
VeriFetch ROCm
Verified Web Retrieval Agent on AMD ROCm
Most web agents hallucinate. This agent is built around one rule: never ship an unverified citation. Given a start URL and a natural-language request, it fetches the page, extracts readable text, and picks the windows most likely to contain the answer using IDF-weighted term scoring instead of blindly taking the first 6000 characters. It then asks a Token Lab LLM (DeepSeek-V4-Flash) for exactly one short verbatim quote. Every quote is checked against the fetched page text; if verification fails, the agent falls back to a guarded extractive answer and records fallback_reason instead of inventing evidence. When the page truly lacks the answer, it returns an honest not-found result. Results: 11/11 evaluation questions pass, 9/9 citations verify verbatim against the source page, 0 fabricated citations, 14 pages fetched, and a multi-page demo cites the correct page. A live Token Lab run also passes 11/11 (9 LLM answers, 2 guarded extractive fallbacks). 22 unit tests pass offline. The AMD stack is load-bearing: developed on ROCm 7.2 on an AMD Instinct MI300X VF GPU (wavefront 64) with HIP kernels for vector add, histogram and reduction, rocBLAS GEMM (73x over naive), a tiled LDS transpose (14x), and rocprofv3 counter profiling. The agent itself is pure Python standard library plus the AMD Token Lab API, so it runs with one command.
PHAM MINH CANH
Chronos: OpenGTM
OpenGTM is Chronos's open-source go-to-market skills library for AI agents. It gives an agent reusable guidance for defining an ideal customer profile, researching accounts, writing cold emails, planning LinkedIn outreach and multichannel sequences, handling replies, and improving SEO and visibility in AI search. The public repository contains 17 skills in the SKILL.md format, installation options for agent tools, and a zero-dependency MCP server for loading the skills. The project separates knowledge from execution. The skills can be used for free without an account to help an agent think, plan and draft. An optional connection to the Chronos MCP provides the execution layer for contacts, messages and campaigns. OpenGTM is built by Chronos and released under the MIT License. This submission covers that public repository, not Chronos's private production code. Chronos's broader goal is an environment where companies can connect their systems, understand their data and act through coordinated agents. OpenGTM contributes a reusable GTM knowledge layer toward that goal. No AMD hardware use or completed AMD Match is claimed.
Chronos
TaskFlow - Smart Agile Kanban Board
TaskFlow is a streamlined, client-side Agile Kanban workspace engineered to eliminate friction in daily sprint tracking and hackathon task management. Modern project management software has become bloated, slow, and overly reliant on complex cloud infrastructure. TaskFlow addresses this issue by delivering an instant, distraction-free board built entirely with native web standards and local persistence. Key Features: 1. Dynamic Sprint Pipeline: Effortlessly categorize tasks across To Do, In Progress, and Completed columns with live badge counters. 2. Fluid Interaction: Full desktop drag-and-drop mechanics combined with quick-shift tap buttons for touch and mobile viewports. 3. Priority Tagging & Live Filtering: Categorize items by urgency (High, Medium, Low) and instantly filter large backlogs using the built-in search bar. 4. Zero-Latency Local Persistence: All board updates, task movements, and edits automatically serialize into browser local storage, safeguarding state across browser refreshes without requiring external logins or APIs. 5. Instant Evaluation: Ships with pre-configured sprint demo tasks so evaluators can test the complete lifecycle right away. Problem Solved: Builders and agile teams often waste critical time setting up heavy ticketing software instead of executing. TaskFlow provides a zero-setup, privacy-first management board that runs instantaneously anywhere.
NewOne
For Innovators & Creators
AetherTactics: Low-Compute Gemini Agent
AetherTactics is an autonomous strategy AI agent engineered for the AMD AI League Match 6 challenge: "Can Your AI Master the Unknown?". The agent navigates an unfamiliar, procedural fog-of-war grid, evades environmental hazards, recovers mission-critical alien relics, and safely reaches extraction points while aggressively minimizing compute overhead and token consumption. Powered by Google Gemini (supporting Gemini 2.5 Flash, 3.1 Flash-Lite, and custom endpoints), AetherTactics utilizes an ultra-compact radar serialization protocol. Instead of verbose context passing, it translates 2D exploration states into dense, minified matrices, constraining LLM reasoning to strict micro-JSON directives under 60 output tokens. To prevent agent freezes during live competition, the platform incorporates an automated watchdog failover pipeline. If a model encounters 503 high-demand or rate-limit spikes, the system automatically executes exponential backoff and fails over to stable alternatives without interrupting autonomous navigation. Built-in live telemetry monitors prompt tokens, generation tokens, step latency, and turn efficiency to deliver state-of-the-art decision-making on minimal inference budgets.
Solo Builder XP
VeriFetch ROCm: Cite-or-Refuse Web Agent
VeriFetch ROCm — Cite the Exact DOM Substring, or Refuse to Guess. Built for AMD AI League Match 4 (Intelligent Web Information Retrieval), VeriFetch ROCm is a high-precision web and HTML table value extraction agent engineered to run locally on AMD Instinct MI300X (192 GB HBM3) inside the official rocm/pytorch container with zero paid external API keys. WHY STANDARD WEB RAG AGENTS FAIL When navigating dense technical specification pages, OS compatibility matrices, or financial filings, standard web retrieval pipelines break in three critical ways: 1. Naive HTML stripping (such as BeautifulSoup .get_text()) destroys <table> row and column alignment, merging numbers across columns (for example, confusing Peak FP16 Dense TFLOPS with Peak FP8 Sparse TFLOPS). 2. Cold-loading a 4B or 8B parameter LLM on every CLI invocation consumes 14 to 18 seconds of the strict 30-second per-query evaluation budget. 3. Generative models paraphrase or guess when a metric is absent, failing exact-match automated grading. HOW VERIFETCH ROCM SOLVES THIS ON AMD INSTINCT MI300X 1. Resident Unix-Socket ROCm Daemon (/app/daemon_rocm.py): Boots once inside rocm/pytorch and keeps our extraction engine warm in HBM3 VRAM behind /tmp/verifetch.sock. Every CLI invocation (/app/app.py --query-id <id> --query "...") completes in ~0.42 seconds—35x faster than cold model boot. 2. Header-Contextualized DOM & Table Linearizer (/app/extractor.py): Parses HTML <table>, <dl>, and JSON-LD structures into self-contained "[Row Header | Column Header]: Cell Value" tuples before hybrid scoring. 3. Deterministic Cite-or-Refuse Substring Verifier: Before writing /app/output/<query-id>_output.json, VeriFetch verifies character-for-character that the extracted atomic value exists in the raw source DOM. If unverified, it deterministically abstains (answer: "", citations: []), achieving 100% exact-match accuracy and 0.0% hallucination on adversarial traps.
VeriFetch ROCm
Verified Web Retrieval Agent on AMD ROCm
Most web agents hallucinate. This agent is built around one rule: never ship an unverified citation. Given a start URL and a natural-language request, it fetches the page, extracts readable text, and picks the windows most likely to contain the answer using IDF-weighted term scoring instead of blindly taking the first 6000 characters. It then asks a Token Lab LLM (DeepSeek-V4-Flash) for exactly one short verbatim quote. Every quote is checked against the fetched page text; if verification fails, the agent falls back to a guarded extractive answer and records fallback_reason instead of inventing evidence. When the page truly lacks the answer, it returns an honest not-found result. Results: 11/11 evaluation questions pass, 9/9 citations verify verbatim against the source page, 0 fabricated citations, 14 pages fetched, and a multi-page demo cites the correct page. A live Token Lab run also passes 11/11 (9 LLM answers, 2 guarded extractive fallbacks). 22 unit tests pass offline. The AMD stack is load-bearing: developed on ROCm 7.2 on an AMD Instinct MI300X VF GPU (wavefront 64) with HIP kernels for vector add, histogram and reduction, rocBLAS GEMM (73x over naive), a tiled LDS transpose (14x), and rocprofv3 counter profiling. The agent itself is pure Python standard library plus the AMD Token Lab API, so it runs with one command.
PHAM MINH CANH
Chronos: OpenGTM
OpenGTM is Chronos's open-source go-to-market skills library for AI agents. It gives an agent reusable guidance for defining an ideal customer profile, researching accounts, writing cold emails, planning LinkedIn outreach and multichannel sequences, handling replies, and improving SEO and visibility in AI search. The public repository contains 17 skills in the SKILL.md format, installation options for agent tools, and a zero-dependency MCP server for loading the skills. The project separates knowledge from execution. The skills can be used for free without an account to help an agent think, plan and draft. An optional connection to the Chronos MCP provides the execution layer for contacts, messages and campaigns. OpenGTM is built by Chronos and released under the MIT License. This submission covers that public repository, not Chronos's private production code. Chronos's broader goal is an environment where companies can connect their systems, understand their data and act through coordinated agents. OpenGTM contributes a reusable GTM knowledge layer toward that goal. No AMD hardware use or completed AMD Match is claimed.
Chronos
TaskFlow - Smart Agile Kanban Board
TaskFlow is a streamlined, client-side Agile Kanban workspace engineered to eliminate friction in daily sprint tracking and hackathon task management. Modern project management software has become bloated, slow, and overly reliant on complex cloud infrastructure. TaskFlow addresses this issue by delivering an instant, distraction-free board built entirely with native web standards and local persistence. Key Features: 1. Dynamic Sprint Pipeline: Effortlessly categorize tasks across To Do, In Progress, and Completed columns with live badge counters. 2. Fluid Interaction: Full desktop drag-and-drop mechanics combined with quick-shift tap buttons for touch and mobile viewports. 3. Priority Tagging & Live Filtering: Categorize items by urgency (High, Medium, Low) and instantly filter large backlogs using the built-in search bar. 4. Zero-Latency Local Persistence: All board updates, task movements, and edits automatically serialize into browser local storage, safeguarding state across browser refreshes without requiring external logins or APIs. 5. Instant Evaluation: Ships with pre-configured sprint demo tasks so evaluators can test the complete lifecycle right away. Problem Solved: Builders and agile teams often waste critical time setting up heavy ticketing software instead of executing. TaskFlow provides a zero-setup, privacy-first management board that runs instantaneously anywhere.
NewOne