Focused on frontier AI technologies, empowering AI researchers' progress.
DeepSeek V4.1 Flash introduces a new asymmetric Causal Encoder-Decoder architecture with 552B MoE parameters, reducing KV cache HBM demand by 75% and SSD demand by 87.5% while adding native multimodal vision and speculative decoding.
Researchers discover S-Space, a stable low-dimensional spatial workspace in multimodal models that encodes object positions, but models confuse coordinate systems and fail to reliably rotate spatial representations, revealing a gap between having spatial representations and using them for reasoning.
PolyWorkBench introduces 67 cross-lingual long-horizon workflow tasks across 5 domains and 10 languages, revealing that top models like Claude Opus 4.8 show 22.5% performance variance across agent harnesses and significant drops on commerce tasks and low-resource languages due to language understanding and cross-lingual coordination errors.
Anthropic's 20-person Labs team uses two-week evaluation cycles to incubate products like Claude Code and MCP, accepting a 20-30% success rate while graduating viable projects to independent teams once they exceed four members, creating a Bell Labs-style engine for turning frontier model capabilities into commercial tools.
GPT-6 Astra achieves near-perfect scores on the ARC-AGI-3 benchmark using a symbolic world model that internalizes reasoning and tool creation, but the $360-per-task compute cost and reliance on external harnesses raise questions about whether this represents genuine AGI progress or expensive brute-force engineering.
OpenAI publishes internal data showing AI agents now contribute 3.1 workdays per human researcher day, with median researchers spending $600 daily on inference, while acknowledging complex tasks still require human intervention and safety restrictions caused GPU usage shifts.
OpenAI Codex lead Tibo reveals why next-gen AI agents will make skills and memory management disappear, how Ultra Fast mode restores real-time flow, why ChatGPT and Codex are merging into a personalized AGI, and how recursive self-improvement now extends from model training to CUDA kernels and infrastructure.
Om AI's VLX-VR model achieves 78.8% accuracy on Google DeepMind's MINERVA long-video reasoning benchmark, outperforming Gemini 3.5 and GPT-4.1, and uniquely improves accuracy on videos over 15 minutes, reversing the typical performance drop with longer videos.
Westlake University's Code World Model separates world evolution from visual rendering: a coding agent maintains executable state and rules via code, while a video model generates high-fidelity observations guided by a lightweight Proxy representation, enabling long-horizon, controllable open-world simulation.
Microsoft research shows sliding window attention with four attention sinks achieves 99% performance recovery on knowledge tasks without any post-training, matching linear attention methods requiring hundreds of millions of tokens while outperforming them 2-10x on long-context benchmarks.
