groundy

Groundy — independent coverage of developer tools, infrastructure, and platforms

models

Operator-Level Triage for Silent Mixed-Precision Instability

arXiv 2607.25494 introduces a single-pass tool to localize silent numerical instability in bf16 and fp16 training. Treat it as a triage layer for unexplained loss spikes, not.

6min·11:42·jul 31
models

BeyondUncertainty: Weak Confidence Signal for RAG Routing, Not Calibration

arXiv:2607.25600 shows verbalized confidence is a weak but real routing signal for RAG. It saves 20.4% retrieval calls for 28.2% token overhead. The probe is poorly.

6min·09:37·jul 31
policy

Public Sector AI Procurement Must Shift from Model Certification to Task Authorization

Public sector AI governance breaks under general-purpose models. One-time procurement certification fails because model behavior is not fixed. Teams must shift to task-level.

infra

daVinci-kernel shifts the RL kernel bottleneck from reward shaping to skill libraries

The daVinci-kernel preprint argues RL-tuned CUDA kernels plateau because skill libraries are mis-architected, not because rewards are mis-shaped. It co-evolves skill.

agents

Cloudflare Precursor: Behavioral Detection for AI Agents

Cloudflare Precursor reportedly shifts AI agent detection from spoofable headers to continuous behavioral signals. This article verifies the claims and outlines operator.

devtools

Provenance as a CI Gate: Attributing Agent-Authorship in Code

Git credits the merger, not the model. F(AI)2R proposes a PROV-O provenance graph gated by CI to record agent authorship and human verification, preventing blame collapse in.

devtools

OHTTP CLI: Stateless Privacy for Agents vs VPN and Tor

Cloudflare's OHTTP protocol splits request identity from content across two parties. This CLI approach offers stateless privacy for agents, shifting key management burden to.

models

Kimi K3 on M1 Max: Bandwidth, Not Capacity, Limits Local MoE Inference

Running Kimi K3 on an M1 Max proves local MoE feasibility via expert offloading, but unified memory bandwidth caps throughput. Treat this as a prototyping tool, not a serving.

featurededitor's picks
most popularlast 6 days · by traffic
tophighest rated
feedall 653 articles · reverse chronological
  1. jul 31modelsOperator-Level Triage for Silent Mixed-Precision Instability
  2. jul 31modelsBeyondUncertainty: Weak Confidence Signal for RAG Routing, Not Calibration
  3. jul 31policyPublic Sector AI Procurement Must Shift from Model Certification to Task Authorization
  4. jul 30infradaVinci-kernel shifts the RL kernel bottleneck from reward shaping to skill libraries
  5. jul 30agentsCloudflare Precursor: Behavioral Detection for AI Agents
  6. jul 30devtoolsProvenance as a CI Gate: Attributing Agent-Authorship in Code
  7. jul 30devtoolsOHTTP CLI: Stateless Privacy for Agents vs VPN and Tor
  8. jul 30modelsKimi K3 on M1 Max: Bandwidth, Not Capacity, Limits Local MoE Inference
  9. jul 30policyWhy Written AI Policies Fail to Control Agent Behavior
  10. jul 29agentsWhy Multi-Agent LLM Delegation Concentrates Risk
  11. jul 29modelsKimi Linear Cuts KV Cache 75% but Recall Remains the Binding Constraint
  12. jul 29policyWhy Vendor Model Cards Fail Clinical Ethics Procurement
  13. jul 29agentsHarness vs Scaffold: Why Claude Code and LangGraph Are Not Interchangeable
  14. jul 28devtoolsFine-Tuning vs RAG for Internal APIs: StarCoder2 Constraints
  15. jul 28agentsMCP Tool Discovery Moves From Hardcoded Config to Runtime Agent Search
  16. jul 28infraCalibrated LLM Monitoring: Conformal Prediction with Drift Detection
  17. jul 28devtoolsMellum2 Unverified: Why MoE Active Parameters Matter More Than Total Size
  18. jul 28agentsCode-as-Action Agents Beat GAIA But Require Runtime Sandboxing
  19. jul 28policyTRIDENT Benchmark: LLM Safety Gaps in Finance, Medicine, and Law
  20. jul 28modelsWhy Chat Leaderboards Do Not Predict Image Quality
  21. jul 27devtoolsPyPI Wheel Reproducibility: 15% Byte-Identical, 79% Source-Equivalent
  22. jul 27modelsKimi K3 Procurement: Governance Review Over Phantom Government Assessments
  23. jul 27policyEU AI Act Traceability: Why ML Pipelines Fail Conformity Assessment
  24. jul 27infraCloudflare AI Crawler Controls: Block, Charge, or Allow Bots Per Route
  25. jul 27agentsWhy Agent Security Tests Must Audit Full Trajectories, Not Single Turns
  26. jul 26agentsx402 Per-Call Payments: Agent Wallet Custody and Replay Risks
  27. jul 26modelsDeepSeek Compute Leak: Why Open-Weight Routing Needs a Swap Path
  28. jul 26modelsContext Ordering Beats Window Size for Long-Context Agents
  29. jul 26devtoolsCHRONO-RESOLUTION: npm, PyPI, and crates.io lockfile drift measured at release points
  30. jul 25infraPostgres LISTEN/NOTIFY Scales: When to Drop Redis for Job Fan-Out
  31. jul 25devtoolsCLI-Tool-Bench: Why Patch Leaderboards Fail for 0-to-1 Code Generation
  32. jul 25policyImplicit Bias in LLMs Passes NYC and EU Audits
  33. jul 25agentsCodeRabbit Review Study: 56% Rejection Rate Demands Targeted Scoping
  34. jul 24modelsDiffusion LLMs: Training Cost, Not Parallel Decoding, Drives Deployment
  35. jul 24infraTailscale on Azure: Measure Direct vs DERP Routing to Control Latency and Egress
  36. jul 24infraAccelerate vs Megatron Core: The Model Size Curve for Distributed Training
  37. jul 24agentsLLM Agents Ignore Mid-Flight Halt Signals: 0 of 40 Trials Stopped
  38. jul 24modelsOpen-Weight Routers vs Fable 5: The Routing Math That Actually Matters
  39. jul 24agentsAgent-First CLIs: Why GitHub, npm, and PyPI Must Publish Machine-Readable Contracts
  40. jul 23policyEU Driver Monitoring: GDPR Compliance Without Consent
  41. jul 23infraWhy cgroups, not permission prompts, bound AI agent CPU and memory
  42. jul 23modelsDeepSeek-V4 1M Context vs RAG: Why Retrieval Stays
  43. jul 23modelsQwen-Image-3.0 Does Not Exist: Why Self-Hosting Image Models Is Premature
  44. jul 22policyEU AI Act bans emotion AI in schools, but permits it where models fail
  45. jul 22agentsMCP and AGENTS.md Standardize Context, Not Agent Coordination
  46. jul 22industryAI Search's One-Answer Rule: When Better Content Makes Search Worse
  47. jul 22agentsCursor's Swarm Math: When Cheap Agents Save Money and When They Fail
  48. jul 21agentsRuntime monitoring beats alignment for agent-to-agent coercion
  49. jul 21infravLLM Configs Shift Energy, Latency, and Accuracy: A 9,000-Run Study
  50. jul 21policyHuggingFace vs GitHub Models vs Replicate: Policy Compliance for Uploaders
load older →