Engineering ·
AgentSysBench examines serving for agents
Published on August 21, 2026, the paper presents a benchmark covering 10 agentic applications and reports that model inference is often not the main serving bottleneck.
Radar
Daily bookmarks: relevant news, facts and discoveries.
Articles 1219
Engineering ·
FIBER separates GPU registers from execution
Published on August 21, 2026, the article presents FIBER, a GPU architecture that decouples register ownership from parallel execution instances. It describes ISA extensions, lightweight microarchitectural changes, and a programming model compatible with CUDA.
Engineering ·
Marin details materials on model training
A publication dated August 21, 2026 describes visualizations of pretraining mixtures by domain, sampled documents, live training loss, configurations, and scaling laws.
Engineering ·
Slides on post-training and real-world adaptation
Introductory lecture slides for economists cover post-training methods and research questions about agents in changing environments.
Engineering ·
A year of LLM serving workloads in production
A paper examining a year of production LLM traffic reports workload changes, short-lived prefix reuse, and tradeoffs between KV-cache locality and load balancing.
Engineering ·
A diary documents pretraining a MoE model
Published on August 20, 2026, the free book recounts an attempt to pretrain a 1.02-billion-parameter MoE model on an H200.
Engineering ·
dstack announces Presets for inference optimization
In an announcement dated August 20, 2026, dstack presents Presets, an open-source toolkit that uses agents to simplify inference optimization and provides a portable format for its results.
Engineering ·
CuTe: layout representation and algebra for GPUs
An article covers CuTe layout representation and algebra, a relevant topic for engineers working with layout abstractions in GPU programming.
Engineering ·
Movfuscator compiles C using only MOV
Movfuscator turns C into Turing-complete assembly using only the MOV instruction, showing how restricting a compiler’s output can complicate code analysis.
Engineering ·
Report points to a shell prompt mismatch in DSH
A post dated August 17, 2026 reports that different prompts in two DSH tools triggered a 3.5-second timeout; after a constant was changed, one command reportedly fell from 3,600 ms to 158 ms.
Engineering ·
Study reports code overlap in filtering systems
Authors of a USENIX Security paper report overlap between leaked Geedge Networks source code and China’s Great Firewall, including DNS and RST injection behavior.
Engineering ·
RL replay buffer with top-k in SQL
A post dated August 16, 2026 describes weighted top-k SQL sampling for a reinforcement-learning replay buffer and reports selecting 8,192 rows from 50 million in about 1.1 seconds on a laptop.
Engineering ·
datatrove 0.10.0 expands pipeline execution options
Version 0.10.0 introduces an executor for pipelines on a jobs service, support for storage buckets as data folders, and preservation of reasoning outputs in inference results.
Engineering ·
Tracing an LDG Instruction on the RTX 4090
An article published on August 13, 2026, follows an LDG.E SASS instruction through the RTX 4090’s hardware units and memory hierarchy, including reverse engineering of undocumented details.
Engineering ·
Swizzling and bank conflicts in CUDA shared memory
A technical article published on August 13, 2026, discusses using swizzling to address bank conflicts in CUDA shared memory, a memory-access issue that can affect CUDA kernel performance.
Engineering ·
IO Factory simulates AI-driven influence campaigns
An article dated August 12, 2026 presents an AI framework for simulating coordinated, adaptive influence campaigns.
Engineering ·
A CUDA challenge path for GPU programming practice
Published on August 12, 2026, the bookmark recommends LeetGPU challenges for CUDA practice and outlines a progression from naive SIMD kernels to more advanced techniques and tools.
Engineering ·
Inference hooks control prompts
Published on August 11, 2026, the text describes hooks that send prompts for security checks before inference, allowing an organization to permit or deny each request.
Engineering ·
ExploitGym evaluates AI agents on vulnerability exploitation
Research published on August 10, 2026 describes an evaluation of AI agents’ ability to turn vulnerabilities into concrete impacts, such as unauthorized file access or code execution.
Engineering ·
ResidencyRL trains clinical agents in simulated consultations
Research explores reinforcement learning in simulated clinical consultations, involving patient history, diagnostic hypotheses, and decisions under uncertainty.