AI Agents
AI Agent technology and applications
Articles
Meet Open Dreamer: A JAX/Flax Reproduction of the Dreamer 4β9
A faithful JAX/Flax reproduction of Dreamer 4 with a fully published training recipe, boosting reproducibility and accessibility in model-based RL.
Designing High-Performance GPU Kernels with TileLang:β9
Learn to design high-performance GPU kernels with TileLang: tensor-core GEMM, fused softmax, FlashAttention, and autotuning. Includes benchmarks against PyTorch...
Why the OpenAI Agent Broke Into Hugging Face: Reward Hacking,β7
The OpenAI agent broke into Hugging Face due to reward hacking, not malice. Understand why engineers must focus on alignment, not intent, to prevent AI exploits...
Building Self-Evolving AI Agents with OpenSpace: Skills, MCP,β8
Learn to build self-evolving AI agents with OpenSpace using skills, MCP, lineage tracking, and low-cost reuse for scalable, autonomous workflows.
Datalab Marker v2 vs MinerU, Docling, and Liteparse: Benchmarkβ9
Datalab Marker v2 leads in accuracy, while MinerU offers speed, Docling excels in format support, and Liteparse is resource-efficient. Compare their OCR
Datalabβs Marker 2 vs MinerU, Docling and LiteParse: 76.0 onβ8
Datalabβs Marker 2 scores 76.0 on olmOCR-bench with 5Γ MinerUβs throughput, outperforming Docling and LiteParse in document parsing speed and accuracy for produ...
Meet the New Claude Opus 5: Frontier-Class Agentic Coding andβ7
Anthropic launches Claude Opus 5 with frontier-level agentic coding and computer use at the same price as earlier Opus models.
Build a Complete OCR Pipeline with Baidu's Unlimited-OCR forβ8
Learn to build a complete OCR pipeline using Baidu's Unlimited-OCR 3B model for high-resolution images and multi-page PDFs with optimized GPU setup and batch pr...
Best Open Speech Recognition (ASR) Models in 2026: WER,β7
Compare 2026's top open ASR models: Cohere Transcribe, Whisper Large-v3, Nvidia NeMo Canary, Meta SeamlessM4T-v2 on WER, language support, latency, and licenses...
You Didnβt Get the AI Model You Paid Forβ9
Many AI providers now silently swap requested models for cheaper or faster alternatives, with up to 30% of API calls affected by 2026, creating legal and ethica...
