Articles, one at a time.
Every piece cites primary sources and adds something you can act on: a worked example, a tested command, a documented limitation. Drafts that only summarize what already exists stay unpublished.
Claude Agent Skills in Production: Packaging, Permissions, Caching
We package and run a Claude Agent Skill through the Agent SDK, then isolate the discovery, permission, and cache failures that can derail production.
Read →
Stop Averaging p95: A Python Lab for API Latency Histograms
A reproducible Python lab shows why averaging worker p95 fails, what merged histograms preserve, and where bucket interpolation can mislead API dashboards.
Read →
Weaviate 1.30 BlockMax WAND: Hybrid Search vs Qdrant and Pinecone
A hands-on benchmark of Weaviate 1.30 BlockMax WAND hybrid search, measuring latency, recall, and cost against Qdrant and Pinecone on RAG workloads.
Read →
Agent Reliability: Retries, Loop Detection, and Timeout Budgets
Hands-on tests of exponential-backoff retries, step-count loop detection, timeout budgets, and human escalation triggers for production LLM agents.
Read →
LangGraph in Production: Postgres Checkpoints and Re-Execution Traps
We ran LangGraph with a Postgres checkpointer, reproduced silent re-execution on long tool calls, and documented the gotchas, workarounds, and cost math.
Read →
Qdrant Hybrid Search in Production: Dense, Sparse, and RRF Tested
We ran Qdrant dense+sparse hybrid search with FastEmbed and RRF fusion in one Docker container, hit the documented failure modes, and applied the fixes.
Read →
Google Antigravity 2.0: Browser-Driving Agents on Real Repo Work
We put Google Antigravity 2.0's Agent Manager and browser-driving agents through scripted repo tasks, including a browser E2E test: what worked, what broke.
Read →
vLLM on Kubernetes: From Notebook Demo to a Production SLA
Moving vLLM from a single-GPU notebook demo to Kubernetes, with measured p50/p95 latency under concurrency, grounded in the vLLM production-stack release.
Read →
Give AI Agents Their Own OAuth Identity with Scoped Credentials
Why sharing users' OAuth tokens with AI agents breaks audit and least privilege, and how on-behalf-of token exchange gives agents scoped, revocable access.
Read →
Sonnet 5 Pricing: Why a Cheaper Token Can Cost You More
Sonnet 5 costs a third less per token and counts your text differently. The break-even math, plus the caching and routing architecture that decides it.
Read →
Your LLM Bill Is an Architecture Problem, Not a Prompt Problem
Provider invoices bill credentials, not customers. Three LLM cost attribution patterns, a list-price model, and which control to build first.
Read →
Semantic Caching vs. Prompt Caching: Where Break-Even Sits
Two caching layers on one workload, priced from Anthropic's official cache rates. Where each pays, and the measurement that tells you which fits.
Read →