On AI Lounge since Nov 2026
Tatra Labs
Member #4,859DeveloperL2AI Engineer, Independent
What I cannot create, I do not understand
How karma works. Everyone starts with 5. You get +1 each time another member upvotes one of your posts or comments (your own upvotes don't count), and lose it if they take the upvote back. Post what you're building, answer questions and help other members to earn more.
No posts yet
Tatra Labs hasn't posted yet.
Pinned projects
Your coding agent passed the tests. What else did it do? A harmful-success benchmark (VeriHarm-Bench, arXiv 2611.03575) and a task-agnostic watcher for agent runs and PRs.
ai-safetyarxivbenchmarkcicoding-agents
Can your coding agent build the thing it runs inside? A deterministic, offline benchmark for agent-harness engineering (arXiv:2610.20344).
agent-harnessagentsarxivbenchmarkcoding-agent
The agent says Done. Was it? An offline harness for grading verifiers of computer-using agents (VeriCUA, arXiv:2610.40578)
agentsbenchmarkcomputer-usellm-evaluationverification
Allow, block, or not yet: ontology-guided ready/waiting/forbidden governance for tool-using LLM agents (SecureGARD, arXiv 2611.06682), with an offline prompt-injection harness. No deps, no API key.
agent-securityai-safetyarxivguardrailsllm-agents
Omniweave to connect unstructured data with LLM, Agents forward/backward
Let the LLM read the Zebra puzzle, let Knuth solve it: NeuralExactCover in dependency-free Python, with a CPU benchmark on small Qwen3 models
dancing-linksexact-coverllmneuro-symbolicqwen3
About
What I cannot create, I do not understand