Guangba Yu.
中文 EN
Ideas in practice

Tech Notes

Technical ideas, implementations, and lessons learned.

Notes are prepared with AI assistance; the original research and views belong to their respective authors.

Read across works

Technical Topics

Explore topics

RSI: from self-editing code to verifiable AI R&D loops

Twenty recent works across agents, infrastructure, weights, evaluation, and safety, compared through persistent loops, transfer, cost, and verification boundaries.

Updated Nov 11, 2026
Filters

    Latest note153 NOTES

    2026-11-11 | TLX: make warp orchestration explicit in the program

    • Inference Serving

    Producer/consumer pipelines, TLX abstractions and the scope of Meta’s three production examples.

    Read note ↗

    The archive

    November 202658 NOTES
    Nov 11, 20262026-11-11 | TLX: make warp orchestration explicit in the program
    • Inference Serving
    Nov 11, 20262026-11-11 | LMCache and Phoenix: bring SSD-resident KV directly to the GPU
    • Inference Serving
    Nov 11, 20262026-11-11 | NVIDIA Snapshot: reuse initialized state for replica recovery
    • Inference Serving
    • Systems Reliability
    Nov 11, 20262026-11-11 | Muse Spark Open on TPU: pipeline the step and preserve session state
    • Inference Serving
    Nov 11, 20262026-11-11 | LLDB across CPUs and GPUs: one session, separate address spaces
    • Observability & Diagnosis
    • Systems Reliability
    Nov 11, 20262026-11-11 | GKE Dynamic Slicing: decouple TPU topology from node pools
    • Systems Reliability
    Nov 11, 20262026-11-11 | Agent Substrate: suspend sessions without dedicating warm workers
    • Agent Reliability
    • Inference Serving
    • Systems Reliability
    Nov 10, 20262026-11-10 | TPU Sync: asynchronous model-state movement across memory tiers
    • Inference Serving
    • Systems Reliability
    Nov 10, 20262026-11-10 | Grafana Watchers: calibrate once, make one narrow decision per check
    • Agent Reliability
    • Observability & Diagnosis
    • Anomaly Detection
    Nov 10, 20262026-11-10 | GLM-5.3-Flash: separate state semantics and balance prefill with decode
    • Inference Serving
    • Systems Reliability
    Nov 10, 20262026-11-10 | Claude Fable / Mythos 5.2: capability, inference cost, and access boundaries
    • Agent Reliability
    • Coding Agents
    Nov 10, 20262026-11-10 | Evaluating AI SRE: separate diagnosis, evidence, and operational responsibility
    • Agent Reliability
    • Observability & Diagnosis
    • Systems Reliability
    Nov 9, 20262026-11-09 | nf12: Why Lossless Weight Compression Can Speed Up or Slow Down Serving
    • Inference Serving
    Nov 8, 20262026-11-08 | vLLM Stable ABI: decoupling engine upgrades from rebuilds
    • Inference Serving
    • Systems Reliability
    Nov 8, 20262026-11-08 | Cache-aware model routing: switching has a state cost
    • Agent Reliability
    • Inference Serving
    Nov 8, 20262026-11-08 | MaxxBench: testing adaptation when familiar requirements change
    • Agent Reliability
    • Coding Agents
    Nov 8, 20262026-11-08 | FrontierCS 2: restoring the evaluator’s control of measurement
    • Agent Reliability
    • Coding Agents
    • Systems Reliability
    Nov 8, 20262026-11-08 | Prompt-cache misses: tracing costs to prefixes and API routes
    • Observability & Diagnosis
    • Inference Serving
    Nov 8, 20262026-11-08 | FETCh: connecting formal specifications to production events
    • Observability & Diagnosis
    • Systems Reliability
    Nov 8, 20262026-11-08 | Inference startup: what weight reuse, warm instances, and snapshots preserve
    • Inference Serving
    • Systems Reliability
    Nov 8, 20262026-11-08 | Opus 5.5: reading the cost curve and the model actually serving a task
    • Agent Reliability
    • Coding Agents
    • Inference Serving
    Nov 7, 20262026-11-07 | Zero Mismatch: remove numerical drift before correcting policy lag
    • Inference Serving
    • Systems Reliability
    Nov 7, 20262026-11-07 | XM-Sparse: sparse reads still need full-cache capacity and bandwidth
    • Inference Serving
    Nov 7, 20262026-11-07 | Reshard: transfer the shards RL inference actually needs
    • Inference Serving
    • Systems Reliability
    Nov 7, 20262026-11-07 | Agent memory: separate capture, correction and retrieval
    • Agent Reliability
    • Self-Improvement
    Nov 7, 20262026-11-07 | Bits Investigation: simpler loops and tool-side evidence processing
    • Agent Reliability
    • Observability & Diagnosis
    Nov 6, 20262026-11-06 | Kinference: a control plane for caches, queues, and tenant policies
    • Observability & Diagnosis
    • Inference Serving
    Nov 6, 20262026-11-06 | Sentry Agent Tracing: turning production traces into evaluation and triage inputs
    • Agent Reliability
    • Observability & Diagnosis
    Nov 6, 20262026-11-06 | Radar on SREGym: connecting diagnosis and repair through structured context
    • Agent Reliability
    • Observability & Diagnosis
    • Systems Reliability
    Nov 6, 20262026-11-06 | ERNIE-5.5-Flash-Base: hybrid attention, ECHO, and serving boundaries
    • Inference Serving
    • Systems Reliability
    Nov 5, 20262026-11-05 | JITEngine: different optimizations for the same model on different Macs
    • Coding Agents
    • Inference Serving
    • Systems Reliability
    Nov 5, 20262026-11-05 | NeMo: turning trace evidence into testable harness repairs
    • Agent Reliability
    • Self-Improvement
    • Observability & Diagnosis
    • Systems Reliability
    Nov 5, 20262026-11-05 | NVIDIA cluster agents: cross-layer investigation with read-only access
    • Multi-Agent Systems
    • Observability & Diagnosis
    • Systems Reliability
    Nov 5, 20262026-11-05 | Model routing: establish a genuinely cheaper execution path
    • Coding Agents
    • Inference Serving
    • Systems Reliability
    Nov 5, 20262026-11-05 | Azure AFPD: prediction must connect to migration and repair
    • Anomaly Detection
    • Systems Reliability
    Nov 5, 20262026-11-05 | Technical reading: Serving the agentic workload: Simon Mo on what AI agents change in vLLM
    • Systems Reliability
    Nov 5, 20262026-11-05 | Polymorphic Inference: Tuning for Changing Workloads
    • Self-Improvement
    • Inference Serving
    Nov 5, 20262026-11-05 | Technical reading: A thousand fewer paths to root
    • Systems Reliability
    Nov 4, 20262026-11-04 | NVIDIA Attribution: Traceable Diagnosis Across Cascading Failures
    • Observability & Diagnosis
    • Systems Reliability
    Nov 4, 20262026-11-04 | Scaling NVIDIA TensorRT Performance Optimization Beyond a Single Build Route with the Global Performance Tuner, Part 1
    • Agent Reliability
    • Self-Improvement
    • Observability & Diagnosis
    • Inference Serving
    • Systems Reliability
    Nov 4, 20262026-11-04 | F3 Lightning: How Meta Accelerates Ads ML Features to Production at Scale
    • Agent Reliability
    • Self-Improvement
    • Observability & Diagnosis
    • Inference Serving
    • Systems Reliability
    Nov 4, 20262026-11-04 | Pika H3 Flex in FastVideo: 7x Faster, and Near Real Time in 4-bit
    • Agent Reliability
    • Self-Improvement
    • Observability & Diagnosis
    • Inference Serving
    • Systems Reliability
    Nov 4, 20262026-11-04 | open-flapping-airplanes: reproducing the flapping airplanes megakernel
    • Agent Reliability
    • Self-Improvement
    • Observability & Diagnosis
    • Inference Serving
    • Systems Reliability
    Nov 4, 20262026-11-04 | Progress on provable inference for large language models
    • Agent Reliability
    • Self-Improvement
    • Observability & Diagnosis
    • Inference Serving
    • Systems Reliability
    Nov 4, 20262026-11-04 | When Quantized KV Cache Quietly Sends Attention Back to the CPU
    • Inference Serving
    • Systems Reliability
    Nov 4, 20262026-11-04 | Technical reading: Lessons from PD Labs: From experimentation to a production-ready SRE Agent
    • Systems Reliability
    Nov 4, 20262026-11-04 | Technical reading: AI Observability & Debugging Leaderboard: APEX-SWE | Mercor
    • Systems Reliability
    Nov 4, 20262026-11-04 | Technical reading: Kimi K3.1 Pro Tech Blog: Advancing Open Intelligence
    • Systems Reliability
    Nov 3, 20262026-11-03 | How Cloudflare Reduced DNS Build Work While Preserving Correctness
    • Observability & Diagnosis
    • Systems Reliability
    Nov 3, 20262026-11-03 | HiCache Across Multi-Parallelism and Distributed Serving
    • Agent Reliability
    • Observability & Diagnosis
    • Systems Reliability
    Nov 3, 20262026-11-03 | Technical reading: Making Kubernetes AI-Ready for NVIDIA Vera Rubin on Day One
    • Systems Reliability
    Nov 2, 20262026-11-02 | ParticleDB: Testing Database Safety Promises on One Node
    • Systems Reliability
    Nov 2, 20262026-11-02 | agent-desktop: Identity and Retry Evidence for Desktop Actions
    • Agent Reliability
    • Observability & Diagnosis
    • Systems Reliability
    Nov 1, 20262026-11-01 | Using Agent Traces to Improve the Agent Itself
    • Agent Reliability
    • Coding Agents
    • Observability & Diagnosis
    • Systems Reliability
    Nov 1, 20262026-11-01 | The Cross-Layer State Chain Behind Kubernetes API Timeouts
    • Observability & Diagnosis
    • Systems Reliability
    Nov 1, 20262026-11-01 | GPUDirect Storage Diagnostics: Deterministic Checks and Agent Orchestration
    • Observability & Diagnosis
    • Inference Serving
    • Systems Reliability
    Nov 1, 20262026-11-01 | How the DeepSeek Prototype Manages Ongoing Work, Authority, and Recovery
    • Agent Reliability
    • Coding Agents
    • Multi-Agent Systems
    • Systems Reliability
    Nov 1, 20262026-11-01 | Can Automated Red Teams Find Training-Data Monitoring Gaps?
    • Agent Reliability
    • Self-Improvement
    • Observability & Diagnosis
    • Systems Reliability
    October 202691 NOTES
    Oct 31, 20262026-10-31 | Using GPU Weights as the Parent for Delta Sync
    • Inference Serving
    • Systems Reliability
    Oct 31, 20262026-10-31 | Radar: open-source Kubernetes UI with built-in MCP server for AI agents
    • Agent Reliability
    • Systems Reliability
    Oct 31, 20262026-10-31 | ForgeKernel: Self-Evolving Kernels Catching and Surpassing Experts and Agents
    • Agent Reliability
    • Systems Reliability
    Oct 31, 20262026-10-31 | Build a Validated Kubernetes Runtime for NVIDIA Dynamo
    • Agent Reliability
    • Systems Reliability
    Oct 30, 20262026-10-30 | How vLLM supports sparse attention and cross-layer KV sharing
    • Inference Serving
    • Systems Reliability
    Oct 30, 20262026-10-30 | Meta uses two test-generation paths to catch risky changes
    • Coding Agents
    • Systems Reliability
    Oct 30, 20262026-10-30 | Google applies agent collaboration to distributed graphs and TPU kernels
    • Multi-Agent Systems
    • Systems Reliability
    Oct 30, 20262026-10-30 | Technical reading: AI SRE: what is it, how could it change on-call – and does it work?
    • Systems Reliability
    Oct 30, 20262026-10-30 | Technical reading: How we detect frontier model misalignment
    • Systems Reliability
    Oct 29, 20262026-10-29 | PyTorch 2.15 Release Blog
    • Agent Reliability
    • Coding Agents
    • Observability & Diagnosis
    • Systems Reliability
    Oct 29, 20262026-10-29 | Meta Andromeda: Powering the Next Generation of Ads Retrieval Through Model-System Co-Design
    • Agent Reliability
    • Coding Agents
    • Observability & Diagnosis
    • Systems Reliability
    Oct 29, 20262026-10-29 | Training SWE agents with Miles on MI355X: taming long-tail trajectories and complex sandbox coordination
    • Agent Reliability
    • Coding Agents
    • Observability & Diagnosis
    • Systems Reliability
    Oct 29, 20262026-10-29 | GKE networking dataplane for large agentic AI workloads
    • Agent Reliability
    • Coding Agents
    • Observability & Diagnosis
    • Systems Reliability
    Oct 29, 20262026-10-29 | Behind the scenes, two months later: How we ship Google Agent Skills like products
    • Agent Reliability
    • Coding Agents
    • Observability & Diagnosis
    • Systems Reliability
    Oct 29, 20262026-10-29 | Technical reading: Gemini 4 Carbon: our next frontier model
    • Systems Reliability
    Oct 29, 20262026-10-29 | Technical reading: The fastest, lowest-cost batch transcription with Qwen3-ASR
    • Systems Reliability
    Oct 28, 20262026-10-28 | SGLang: when is a terminated request’s KV memory reusable?
    • Inference Serving
    • Systems Reliability
    Oct 28, 20262026-10-28 | miniTE: align numerics, learning, and complete RL cost
    • Inference Serving
    • Systems Reliability
    Oct 28, 20262026-10-28 | A Lean tokenizer prototype: make token preservation explicit
    • Agent Reliability
    • Systems Reliability
    Oct 28, 20262026-10-28 | A fleet-wide KV index: where a prefix lives changes its value
    • Observability & Diagnosis
    • Inference Serving
    • Systems Reliability
    Oct 28, 20262026-10-28 | Cloudflare’s KV exposure: why HTTP 200 missed a request-isolation failure
    • Observability & Diagnosis
    • Inference Serving
    • Systems Reliability
    Oct 27, 20262026-10-27 | Restore GPU Jobs While Faulty Nodes Are Repaired
    • Observability & Diagnosis
    • Systems Reliability
    Oct 27, 20262026-10-27 | Bronto: Bound the Scope and Budget of an Alert Investigator
    • Agent Reliability
    • Observability & Diagnosis
    Oct 25, 20262026-10-25 | vLLM at Rack Scale: CPU Costs After Wide Expert Parallelism
    • Inference Serving
    • Systems Reliability
    Oct 25, 20262026-10-25 | SGLang and DeepInfer: Bounded Replay and a Complete Serving Stack
    • Inference Serving
    Oct 25, 20262026-10-25 | LAB: Designing for Behavior After Model Failover
    • Agent Reliability
    • Inference Serving
    • Systems Reliability
    Oct 25, 20262026-10-25 | Testing Kamaji and HAMi Tenant Scheduling Without GPUs
    • Inference Serving
    • Systems Reliability
    Oct 25, 20262026-10-25 | An AI SRE Baseline: Better Context for the Agent You Have
    • Coding Agents
    • Observability & Diagnosis
    • Systems Reliability
    Oct 25, 20262026-10-25 | Reducing Agent Telemetry While Preserving Operational Answers
    • Coding Agents
    • Observability & Diagnosis
    Oct 25, 20262026-10-25 | Agent Anomaly Detection: Asynchronous Auditing of Session Behavior
    • Agent Reliability
    • Observability & Diagnosis
    • Anomaly Detection
    Oct 24, 20262026-10-24 | MiniMax-M3.1-Flash: Low-Precision Cache Across the Serving Stack
    • Inference Serving
    • Systems Reliability
    Oct 24, 20262026-10-24 | Why an Agent Kept Going: Resampling an Incident
    • Agent Reliability
    • Observability & Diagnosis
    Oct 24, 20262026-10-24 | LiteLLM Release Incident: When Multiple Approvals Share One Trust Failure
    • Agent Reliability
    • Systems Reliability
    Oct 24, 20262026-10-24 | Habitat: Balancing Multi-Tenant Isolation and Latency
    • Observability & Diagnosis
    • Systems Reliability
    Oct 24, 20262026-10-24 | GLM-5.3 Long Context: Splitting Query Work and KV Capacity
    • Inference Serving
    • Systems Reliability
    Oct 24, 20262026-10-24 | Training Containment Beyond the Host
    • Agent Reliability
    • Systems Reliability
    Oct 24, 20262026-10-24 | Agent Compute Engine: Designing for Waiting, Resumption, and Authorization
    • Agent Reliability
    • Systems Reliability
    Oct 23, 20262026-10-23 | Turning Fuzz Tests into Proofs: Claude Scales Binary Verification
    • Coding Agents
    • Systems Reliability
    Oct 23, 20262026-10-23 | TorchTitan: Shared Model Code Needs Shared Numerical Evidence
    • Self-Improvement
    • Inference Serving
    • Systems Reliability
    Oct 23, 20262026-10-23 | SGLang Trainium Beta: Separating RL Generation and Training Hardware
    • Self-Improvement
    • Inference Serving
    • Systems Reliability
    Oct 23, 20262026-10-23 | Minerva Connects AI-Cluster Telemetry to Job-Level Diagnosis
    • Observability & Diagnosis
    • Anomaly Detection
    • Systems Reliability
    Oct 23, 20262026-10-23 | METR Red Teaming: Monitoring Coverage Must Follow the Execution Chain
    • Agent Reliability
    • Coding Agents
    • Observability & Diagnosis
    Oct 23, 20262026-10-23 | Grafana AI: From Continuous Observation to Reviewable Patches
    • Agent Reliability
    • Coding Agents
    • Observability & Diagnosis
    Oct 23, 20262026-10-23 | FlyDSL in PyTorch: Hardware-Specific Kernels Join Backend Selection
    • Inference Serving
    • Systems Reliability
    Oct 22, 20262026-10-22 | StartLux-Prism: Shared Decision Computation and Bitwise-Consistent Inference
    • Inference Serving
    • Systems Reliability
    Oct 22, 20262026-10-22 | Flint on vLLM: lowering single-GPU long-context decode latency
    • Inference Serving
    Oct 22, 20262026-10-22 | EAGLE-4: adapt the drafting budget to hardware and load
    • Inference Serving
    Oct 21, 20262026-10-21 | London Cloud Outage: Failover Depends on Capacity and Control Paths
    • Observability & Diagnosis
    • Systems Reliability
    Oct 21, 20262026-10-21 | Humming: Unifying Quantized Kernels Without Erasing Their Differences
    • Inference Serving
    Oct 21, 20262026-10-21 | Autoscaling Recovery: Measuring the Wait with MTSS
    • Observability & Diagnosis
    • Systems Reliability
    Oct 21, 20262026-10-21 | vLLM TorchTPU: Reusing the Serving Control Plane
    • Inference Serving
    • Systems Reliability
    Oct 21, 20262026-10-21 | TorchTPU: Compilation and Communication Behind Native PyTorch APIs
    • Inference Serving
    • Systems Reliability
    Oct 21, 20262026-10-21 | Thin Orchestrator: Moving Domain Reasoning Out of the Router
    • Agent Reliability
    • Coding Agents
    • Multi-Agent Systems
    Oct 21, 20262026-10-21 | Prometheus: Making Large-Cluster Failures Recoverable
    • Observability & Diagnosis
    • Systems Reliability
    Oct 21, 20262026-10-21 | LocalPay: Making Payment Failures Repeatable Locally
    • Observability & Diagnosis
    • Systems Reliability
    Oct 20, 20262026-10-20 | Journey SLOs: measuring probe availability and real experience
    • Observability & Diagnosis
    • Systems Reliability
    Oct 20, 20262026-10-20 | llm-d 1.0: coordinating cache locality, request priority, and upgrades
    • Inference Serving
    • Systems Reliability
    Oct 20, 20262026-10-20 | Grafana Knowledge Graph: starting investigations from entity relationships
    • Agent Reliability
    • Observability & Diagnosis
    Oct 20, 20262026-10-20 | Adaptive Telemetry: which evidence survives cost reduction
    • Observability & Diagnosis
    • Systems Reliability
    Oct 20, 20262026-10-20 | LMCache 0.5.6: Trust and Lifecycle Boundaries in KV Cache Services
    • Inference Serving
    • Systems Reliability
    Oct 20, 20262026-10-20 | Unattended Agents: Seven Failure Scenarios to Validate Before Release
    • Agent Reliability
    • Observability & Diagnosis
    • Systems Reliability
    Oct 20, 20262026-10-20 | Agent Skills: Does Following Instructions Actually Help?
    • Agent Reliability
    • Coding Agents
    • Self-Improvement
    Oct 20, 20262026-10-20 | TorchNeuron: Native PyTorch Integration and Its Compatibility Boundaries
    • Observability & Diagnosis
    • Inference Serving
    • Systems Reliability
    Oct 19, 20262026-10-19 | Jev Diagnosis: Evidence Defines the Limits of Inexpensive Judgments
    • Agent Reliability
    • Observability & Diagnosis
    • Systems Reliability
    Oct 19, 20262026-10-19 | Helion and CompileIQ: Adding Compiler Controls to Kernel Tuning
    • Inference Serving
    Oct 19, 20262026-10-19 | Elastic EP: Restricting Scaling Pauses to Topology Handover
    • Inference Serving
    • Systems Reliability
    Oct 19, 20262026-10-19 | Cloudflare Review Agents: Checking Finding Quality After Cutting Costs
    • Agent Reliability
    • Coding Agents
    • Inference Serving
    Oct 16, 20262026-10-16 | Zephon: Preserving Training Batches When GPU Counts Change
    • Systems Reliability
    Oct 16, 20262026-10-16 | LMCache MP Coordinator: Finding Caches Across Servers
    • Inference Serving
    • Systems Reliability
    Oct 16, 20262026-10-16 | Kimi K3 on MI355X: From State Reuse to Kernel Scheduling
    • Inference Serving
    • Systems Reliability
    Oct 16, 20262026-10-16 | Specialized Kimi K3 Decode: Speed and Validation Boundaries
    • Inference Serving
    • Systems Reliability
    Oct 15, 20262026-10-15 | Scout: Post-Training for Evidence-Based SRE Investigations
    • Agent Reliability
    • Observability & Diagnosis
    Oct 14, 20262026-10-14 | Bitwise Training–Inference Consistency: Keep Low Precision, Fix the Computation Path
    • Self-Improvement
    • Inference Serving
    Oct 14, 20262026-10-14 | NIXL Push: Move KV-Transfer Setup Off the First-Token Path
    • Inference Serving
    Oct 14, 20262026-10-14 | Claude Agent Observability: Bring Execution Into Existing OTel Systems
    • Agent Reliability
    • Observability & Diagnosis
    Oct 14, 20262026-10-14 | Online Verification Budgets Without a Confidence Head
    • Inference Serving
    Oct 13, 20262026-10-13 | Scion: How coding agents change source-control scaling
    • Coding Agents
    • Systems Reliability
    Oct 12, 20262026-10-12 | Building an AI SRE: The Engineering Beyond an Answer
    • Agent Reliability
    • Observability & Diagnosis
    Oct 12, 20262026-10-12 | Bidirectional KV Transfer in vLLM: Returning Multi-Turn Context to Prefill
    • Inference Serving
    Oct 9, 20262026-10-09 | vLLM and Vera Rubin: Turning Memory Locality into Serving Throughput
    • Inference Serving
    Oct 8, 20262026-10-08 | Dynamo: Scheduling Agent Inference and KV Cache by Session
    • Observability & Diagnosis
    • Inference Serving
    Oct 8, 20262026-10-08 | Google AQuA: Finding Quality Issues in Production Traces
    • Agent Reliability
    • Observability & Diagnosis
    Oct 8, 20262026-10-08 | Anyscale: Caching, Routing, and the Cost of Coding Agents
    • Coding Agents
    • Inference Serving
    Oct 7, 20262026-10-07 | DeepSeek-V4.1-Flash: Replay and Kernel Optimizations Behind Fivefold Throughput
    • Inference Serving
    Oct 7, 20262026-10-07 | PAI: What Agent Monitoring Needs from Telemetry
    • Agent Reliability
    • Observability & Diagnosis
    Oct 6, 20262026-10-06 | FBTriton: Organizing Embedding Training Around Sparse Access Patterns
    • Systems Reliability
    Oct 3, 20262026-10-03 | GPU Cluster Readiness: Validate the Workload, Not Only the Components
    • Observability & Diagnosis
    • Systems Reliability
    Oct 2, 20262026-10-02 | Helion Linear Backend: Make Algorithm Selection Tunable
    • Inference Serving
    Oct 1, 20262026-10-01 | Incident Replay: Can the Agent See Enough Evidence?
    • Agent Reliability
    • Observability & Diagnosis
    Oct 1, 20262026-10-01 | TLX Jagged Flash Attention: Keeping Blackwell Busy on Variable-Length Sequences
    • Inference Serving
    Oct 1, 20262026-10-01 | KV Cache and LMCache: From Generation to Reusable Multi-Turn State
    • Inference Serving
    August 20264 NOTES
    Aug 4, 20262026-08-04 | SpecForge Separates Target Inference from Draft Training
    • Inference Serving
    Aug 4, 20262026-08-04 | First-Pass Triage for AI Incidents
    • Agent Reliability
    Aug 1, 20262026-08-01 | Four Layers for SGLang Quantization
    • Inference Serving
    Aug 1, 20262026-08-01 | Co-Designing Attention for Long Context
    • Inference Serving