Groundy · Infrastructure & Runtime

Why Log Parsing Breaks GPU Failure Detection: What Your Templates Drop

Berry Mingus — Thu, 15 Oct 2026 12:34:21 GMT

A preprint shows Drain parsing reduces GPU failure detection F1 to 0.667, while raw logs reach 0.983 at higher cost, suggesting a need for dual-path log retention.

Berry Mingus · Infrastructure & Runtime · 10 min read

Read the full article on Groundy →

Vercel Blob vs Amazon S3: What the Marketplace Integration Changes

Berry Mingus — Wed, 14 Oct 2026 23:27:04 GMT

Vercel Marketplace adds S3 with OIDC credentials, but only for new buckets. Existing AWS estates remain excluded, limiting the integration's reach for established teams.

Berry Mingus · Infrastructure & Runtime · 10 min read

Read the full article on Groundy →

Do AWS Lambda Web Functions Replace API Gateway for Web Apps?

Berry Mingus — Wed, 14 Oct 2026 13:14:14 GMT

AWS Lambda Web Functions replace API Gateway only for new Node.js 24 apps in five regions, shifting auth and throttling to app code while using active CPU billing.

Berry Mingus · Infrastructure & Runtime · 12 min read

Read the full article on Groundy →

Intel D270 vs HBM GPUs: Can 480 GB of LPDDR5X Run LLM Inference?

Berry Mingus — Tue, 13 Oct 2026 23:21:26 GMT

Intel D270 offers 480 GB LPDDR5X capacity for LLM inference, but lacks vendor bandwidth data. It suits capacity-bound workloads where HBM cards fail, pending independent stack

Berry Mingus · Infrastructure & Runtime · 10 min read

Read the full article on Groundy →

Self-Hosting Cloudflare Workers: What celld and the Deno Deal Actually Run

Berry Mingus — Tue, 13 Oct 2026 11:17:15 GMT

Cloudflare's acquisition of Deno enables self-hosting Workers via celld and workerd, but Durable Objects remain unscaled outside Cloudflare's network, making portability an un

Berry Mingus · Infrastructure & Runtime · 9 min read

Read the full article on Groundy →

Can Idle Macs Become a GPU Cloud? What an Apple Silicon Cluster Delivers

Berry Mingus — Mon, 12 Oct 2026 11:07:50 GMT

A preprint shows idle Macs can pool for batch embeddings and Ollama completions, offering private inference that complements rented GPUs for latency-tolerant work.

Berry Mingus · Infrastructure & Runtime · 9 min read

Read the full article on Groundy →

Priority Queues or Budgets: How Ai2 Divides an Oversubscribed GPU Cluster

Berry Mingus — Sat, 10 Oct 2026 22:53:18 GMT

Ai2 reports that replacing priority queues with GPU budgets and fair-share allocation delivered 98% of owed hours, though evidence is self-reported and limited to one 30-day.

Berry Mingus · Infrastructure & Runtime · 12 min read

Read the full article on Groundy →

Hosting Git on Cloudflare: What the Smart HTTP Protocol Actually Requires

Berry Mingus — Thu, 08 Oct 2026 23:05:33 GMT

Cloudflare Artifacts offers Git-native storage for Workers, but lacks documented consistency models and pack generation benchmarks, leaving critical protocol costs unverified.

Berry Mingus · Infrastructure & Runtime · 10 min read

Read the full article on Groundy →

Which RAG Architecture Should You Run? A 2026 Survey Maps the Options

Berry Mingus — Thu, 08 Oct 2026 10:25:13 GMT

A 2026 survey and primary studies show hybrid retrieval with reranking and contextual chunks outperforms complex agentic stacks for most queries, while deferral policies cut L

Berry Mingus · Infrastructure & Runtime · 10 min read

Read the full article on Groundy →

Does Tool-Calling Retrieval Beat Vector RAG on Small Knowledge Bases?

Berry Mingus — Wed, 07 Oct 2026 22:20:02 GMT

KyGround preprint shows vector RAG outperforms tool-calling retrieval on small Greek corpora, while whole-KB prompting achieves highest accuracy.

Berry Mingus · Infrastructure & Runtime · 13 min read

Read the full article on Groundy →

Cloudflare's Web Search API vs Tavily: What Grounding an Agent Costs

Berry Mingus — Wed, 07 Oct 2026 16:51:30 GMT

Cloudflare's Web Search API routes queries to partners at list price without markup. It consolidates billing but does not guarantee lower cost or better quality than direct Ex

Berry Mingus · Infrastructure & Runtime · 9 min read

Read the full article on Groundy →

Can You Detect LLM Hallucinations Without a GPU? What Cheap Detectors Catch

Berry Mingus — Tue, 06 Oct 2026 22:00:26 GMT

A preprint shows CPU-based NLI and similarity ensembles detect QA hallucinations well, but summarization detection remains weak without chunking.

Berry Mingus · Infrastructure & Runtime · 9 min read

Read the full article on Groundy →

Can LLMs Write Production GPU Kernels? What FastKernels Measures

Berry Mingus — Tue, 06 Oct 2026 11:10:07 GMT

Preprint benchmarks show LLM-generated GPU kernels rarely clear production gates, with end-to-end gains shrinking and memory regressions persisting in real serving stacks.

Berry Mingus · Infrastructure & Runtime · 11 min read

Read the full article on Groundy →

How to Measure Post-Quantum TLS Adoption in Cloudflare Traffic Logs

Berry Mingus — Wed, 30 Sep 2026 20:52:15 GMT

Cloudflare logs TLS key-exchange groups to measure hybrid post-quantum adoption. The metric tracks encryption only, not authentication, and classical shares often reflect non-

Berry Mingus · Infrastructure & Runtime · 9 min read

Read the full article on Groundy →

Why the Same LLM Prompt Returns Different Outputs Across GPUs

Berry Mingus — Wed, 23 Sep 2026 11:36:05 GMT

A 2026 preprint shows fixed-configuration kernels achieve cross-GPU bitwise determinism for linear layers, challenging the assumption that reproducibility requires a major end

Berry Mingus · Infrastructure & Runtime · 8 min read

Read the full article on Groundy →

Request Routing for Disaggregated LLM Serving: Where Learned Routers Win

Berry Mingus — Thu, 17 Sep 2026 09:28:25 GMT

A preprint reports that calibrated learned routing for disaggregated LLM serving beats heuristics by 1.7 to 2.9 goodput points, but only on specific hardware with large pools.

Berry Mingus · Infrastructure & Runtime · 9 min read

Read the full article on Groundy →

DuckDB Integrates LLMs and RAG: What In-Database Retrieval Changes

Berry Mingus — Tue, 15 Sep 2026 11:21:48 GMT

DuckDB research extension FlockMTL integrates LLMs and RAG into SQL via schema objects, moving governance to DDL, though performance claims remain author-reported.

Berry Mingus · Infrastructure & Runtime · 8 min read

Read the full article on Groundy →

Why GPU Utilization Misleads LLM Inference Capacity Planning

Berry Mingus — Mon, 14 Sep 2026 08:46:53 GMT

A preprint argues GPU utilization misleads LLM capacity planning by conflating memory-bound decode with compute saturation, urging per-phase metrics.

Berry Mingus · Infrastructure & Runtime · 9 min read

Read the full article on Groundy →

Running MoE LLMs on a Single GPU: What Expert Offloading Actually Costs

Berry Mingus — Fri, 11 Sep 2026 20:14:36 GMT

FluxMoE streams MoE weights from host DRAM to save VRAM, trading capacity for bandwidth. Author-reported gains on multi-GPU setups lack independent replication.

Berry Mingus · Infrastructure & Runtime · 9 min read

Read the full article on Groundy →

Running Kimi K3 From SSDs on a MacBook Pro: The Storage Tradeoff

Berry Mingus — Wed, 09 Sep 2026 07:24:27 GMT

Deltafin streams Kimi K3 weights from SSDs to run on 64 GB Macs, but self-reported benchmarks vary widely, limiting it to batch workloads rather than interactive use.

Berry Mingus · Infrastructure & Runtime · 9 min read

Read the full article on Groundy →

Speculative Decoding on AMD GPUs: What vLLM's Speedup Actually Costs

Berry Mingus — Tue, 08 Sep 2026 20:28:56 GMT

vLLM's ROCm speculative decoding speedup is self-reported and unreplicated. AMD's gigawatt deals de-risk the platform, but operators must benchmark acceptance rates on MI300X.

Berry Mingus · Infrastructure & Runtime · 11 min read

Read the full article on Groundy →

Cloudflare Runs Nearly 9 in 10 European CDNs: The Concentration Problem

Berry Mingus — Tue, 08 Sep 2026 20:05:13 GMT

A Hacker News claim that 9 in 10 European CDN users rely on Cloudflare lacks independent verification. Operators must audit failover paths to avoid correlated failure and meet

Berry Mingus · Infrastructure & Runtime · 12 min read

Read the full article on Groundy →

Noisy Neighbors at the Fabric: Why Shared GPU Clusters Throttle Your Jobs

Berry Mingus — Mon, 07 Sep 2026 07:59:55 GMT

A new NCCL shim recovers 13-38% bandwidth on shared GPU clusters by tuning collective patterns. Test for cross-tenant interference before buying more fabric.

Berry Mingus · Infrastructure & Runtime · 12 min read

Read the full article on Groundy →

Running Your Own Nitter Instance: What the Post-Takedown Comeback Requires

Berry Mingus — Mon, 07 Sep 2026 05:20:55 GMT

Self-hosting Nitter in 2026 is a maintenance contract, not a setup task. Operators must rotate banned X tokens, track upstream commits, and absorb legal exposure from active C

Berry Mingus · Infrastructure & Runtime · 10 min read

Read the full article on Groundy →

Running Local LLMs on a $60 Used GPU: What AMD's BC-250 Can and Can't Do

Berry Mingus — Sun, 06 Sep 2026 08:23:32 GMT

A claimed $60 AMD BC-250 with 16GB VRAM shifts the local LLM bottleneck from capacity to software support. Verify ROCm and llama.cpp backend compatibility before buying, as un

Berry Mingus · Infrastructure & Runtime · 10 min read

Read the full article on Groundy →

Private Vector Search vs TEEs: Can RAG Retrieval Be Outsourced Safely?

Berry Mingus — Sun, 06 Sep 2026 08:06:35 GMT

Spruce claims 0.21-2.97s private retrieval via MPC, challenging TEE defaults. Compare self-hosted Qdrant, TEEs, and cryptographic outsourcing for privacy-sensitive RAG.

Berry Mingus · Infrastructure & Runtime · 13 min read

Read the full article on Groundy →

Running LLMs in the Browser: Can Privacy Be Verified Instead of Promised?

Berry Mingus — Sat, 05 Sep 2026 22:37:52 GMT

Browser LLM inference is local but not private. The host page controls the GPU, model bytes, and network egress, making 'data never leaves device' a policy promise, not a ver

Berry Mingus · Infrastructure & Runtime · 12 min read

Read the full article on Groundy →

Does Cloudflare's Adaptive Intelligence Change the Bot Defense Build-vs-Buy Math?

Berry Mingus — Sat, 05 Sep 2026 22:04:20 GMT

Cloudflare's bundled bot defense changes procurement math, but efficacy claims remain self-reported. Audit standalone spend and measure false positives before migrating.

Berry Mingus · Infrastructure & Runtime · 13 min read

Read the full article on Groundy →

Cloudflare Compresses Its Cache With Zstandard: The Storage-vs-CPU Trade

Berry Mingus — Fri, 04 Sep 2026 19:21:02 GMT

Cloudflare's zstd cache proposal trades storage for CPU. We verify the math: zstd saves 5% over gzip but only 0.45% over brotli, while non-zstd clients pay a 10ms re-encode.

Berry Mingus · Infrastructure & Runtime · 12 min read

Read the full article on Groundy →

Getting an AI Agent Into Cloudflare's BotBase: What Operators Must Verify

Berry Mingus — Mon, 31 Aug 2026 04:49:59 GMT

Cloudflare's agent directory proposal shifts identity burden to operators. With 20% of the web behind its network, unlisted agents face default blocking or per-route charges,

Berry Mingus · Infrastructure & Runtime · 11 min read

Read the full article on Groundy →

Running GLM-5.2 on vLLM: MTP Support, Weight Errors, and Throughput

Berry Mingus — Sun, 30 Aug 2026 15:01:07 GMT

GLM-5.2 on vLLM v0.28.0: enable MTP, fix k_norm warnings, and verify FP8 precision. Measured gains from research on other models, not GLM-5.2.

Berry Mingus · Infrastructure & Runtime · 12 min read

Read the full article on Groundy →

Automating Kubernetes Capability Drops: What KubeCap's Rules Miss

Berry Mingus — Sat, 29 Aug 2026 16:11:43 GMT

KubeCap automates Kubernetes capability minimization via LLM-inferred rules, cutting permissions by 55% in Go workloads. Validate inferred drops against runtime behavior to CI

Berry Mingus · Infrastructure & Runtime · 12 min read

Read the full article on Groundy →

Alibaba's ScaleSense: When Learned Autoscaling Beats Provisioning Rules

Berry Mingus — Sat, 29 Aug 2026 15:43:50 GMT

Alibaba's ScaleSense uses learned query-level estimation to cut AnalyticDB costs by up to 5.22x. This audit covers the risks of opaque autoscalers and the limits of vendor-own

Berry Mingus · Infrastructure & Runtime · 13 min read

Read the full article on Groundy →

AWS Cognito Postmortem: The Real Cost of Free Managed Auth

Berry Mingus — Sat, 29 Aug 2026 03:44:23 GMT

A founder's Cognito postmortem highlights hidden auth costs. This guide maps decision axes for startups, separating verified AWS capabilities from unverified customization and

Berry Mingus · Infrastructure & Runtime · 11 min read

Read the full article on Groundy →

How Cloudflare Saved 100 TB in 1.1.1.1's DNS Cache and What Operators Can Copy

Berry Mingus — Fri, 28 Aug 2026 02:55:21 GMT

Cloudflare claims 100 TB saved in 1.1.1.1 cache, but the mechanism is unverified. Operators can cut memory now by tuning Redis eviction policies and admission controls, not by

Berry Mingus · Infrastructure & Runtime · 13 min read

Read the full article on Groundy →

Where Simple RAG Breaks: Multi-Document QA Needs Hierarchy, Not More Chunks

Berry Mingus — Thu, 27 Aug 2026 02:51:50 GMT

Flat RAG fails on similar-document corpora due to scope confusion and entity conflicts. HiQA proposes hierarchical augmentation, but hybrid retrieval remains the cheaper, unme

Berry Mingus · Infrastructure & Runtime · 11 min read

Read the full article on Groundy →

Codex on AWS Bedrock 10x Charges: Auditing Agent Token Bills

Berry Mingus — Mon, 24 Aug 2026 06:40:00 GMT

A user-reported 10x charge on AWS Bedrock remains unconfirmed by OpenAI or AWS. This runbook reconciles Codex client logs against provider metering to isolate cache misses, or

Berry Mingus · Infrastructure & Runtime · 13 min read

Read the full article on Groundy →

Cloudflare FedRAMP High Claim: Edge vs. GovCloud for Government AI

Berry Mingus — Mon, 24 Aug 2026 02:03:12 GMT

Cloudflare claims FedRAMP Class D status, but the Marketplace shows no listing. This guide maps edge security layers to impact levels and compares latency costs for.

Berry Mingus · Infrastructure & Runtime · 13 min read

Read the full article on Groundy →

GPU Memory Explained: Why LLM Throughput Collapses When VRAM Runs Out

Berry Mingus — Sun, 23 Aug 2026 14:19:46 GMT

LLM decode speed tracks memory bandwidth, not TFLOPS. When VRAM fills, reads shift to PCIe, causing a throughput cliff. Use this bandwidth budget worksheet to plan KV cache.

Berry Mingus · Infrastructure & Runtime · 10 min read

Read the full article on Groundy →

Task-Based OAuth Consent: Scoping AI Agent Permissions Per Action

Berry Mingus — Sun, 23 Aug 2026 02:05:25 GMT

Cloudflare's task-based OAuth consent moves authorization from install time to runtime. Compare blast radius, revocation, and fatigue to redesign agent API security.

Berry Mingus · Infrastructure & Runtime · 11 min read

Read the full article on Groundy →

V8 Isolates vs MicroVMs vs Wasm: Where Spectre Still Draws the Line

Berry Mingus — Fri, 21 Aug 2026 01:16:56 GMT

Cloudflare's 2026 Spectre revisit is unverified. This guide maps V8 isolates, Firecracker microVMs, and Wasm sandboxes to define a data-classification rule for edge secrets.

Berry Mingus · Infrastructure & Runtime · 11 min read

Read the full article on Groundy →

Fine-Tuning DeepSeek Without NVIDIA: What the Ascend SuperPOD Run Shows

Berry Mingus — Thu, 20 Aug 2026 13:22:17 GMT

SLAI T-Rex reports 34.22% MFU for full-parameter DeepSeek-V4 post-training on Ascend. The single-cluster result narrows the CUDA moat for fine-tuning, but cost data remains.

Berry Mingus · Infrastructure & Runtime · 12 min read

Read the full article on Groundy →

Cloudflare WebMCP: The Security Baseline for Agent-Ready Sites

Berry Mingus — Wed, 19 Aug 2026 12:55:36 GMT

Cloudflare's WebMCP proposal is unverified, but the MCP ecosystem is not. With 91.8% of servers unauthenticated, treat agent endpoints as strict APIs, not crawler extensions.

Berry Mingus · Infrastructure & Runtime · 12 min read

Read the full article on Groundy →

Cloudflare AI Search vs Self-Hosted RAG: Where the Build-vs-Buy Line Lands

Berry Mingus — Wed, 19 Aug 2026 00:33:02 GMT

Cloudflare's AI Search lacks verifiable pricing and access control docs. For Postgres 13+ teams, pgvector offers a documented, self-hosted alternative that avoids vendor.

Berry Mingus · Infrastructure & Runtime · 11 min read

Read the full article on Groundy →

Cloudflare H1 2026 DDoS Report: DNS Floods and Sizing Past 1 Tbps

Berry Mingus — Tue, 18 Aug 2026 13:36:51 GMT

Cloudflare's H1 2026 DDoS report claims DNS floods exceeded 1 Tbps. This article verifies the structural risks of UDP amplification and anycast absorption, while noting the.

Berry Mingus · Infrastructure & Runtime · 10 min read

Read the full article on Groundy →

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

Berry Mingus — Thu, 30 Jul 2026 20:22:30 GMT

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.

Berry Mingus · Infrastructure & Runtime · 12 min read

Read the full article on Groundy →

Calibrated LLM Monitoring: Conformal Prediction with Drift Detection

Berry Mingus — Tue, 28 Jul 2026 20:33:27 GMT

A cloud-native architecture packages conformal prediction as a sub-2ms microservice to replace raw logprob confidence with finite-sample coverage guarantees, gated by.

Berry Mingus · Infrastructure & Runtime · 13 min read

Read the full article on Groundy →

Cloudflare AI Crawler Controls: Block, Charge, or Allow Bots Per Route

Berry Mingus — Mon, 27 Jul 2026 07:32:36 GMT

Cloudflare ships per-route AI crawler controls for Search, Agent, and Training bots. Site owners must model pay-per-crawl yield against ad inventory and measure Agent.

Berry Mingus · Infrastructure & Runtime · 11 min read

Read the full article on Groundy →

Postgres LISTEN/NOTIFY Scales: When to Drop Redis for Job Fan-Out

Berry Mingus — Sat, 25 Jul 2026 19:05:38 GMT

DBOS benchmarks suggest Postgres LISTEN/NOTIFY handles high fan-out, but the 2026-07-24 data remains unverified. Use durable queues for small payloads, but keep Redis for.

Berry Mingus · Infrastructure & Runtime · 10 min read

Read the full article on Groundy →

Tailscale on Azure: Measure Direct vs DERP Routing to Control Latency and Egress

Berry Mingus — Fri, 24 Jul 2026 17:59:40 GMT

Tailscale on Azure VMs silently falls back to DERP relays when direct paths fail, adding latency and egress costs. Measure direct success rates to avoid hidden bills.

Berry Mingus · Infrastructure & Runtime · 11 min read

Read the full article on Groundy →