Insights
Insights
What we're learning, writing, and occasionally arguing about. Engineering posts, opinion pieces, and case notes from our work.
Data
When ClickHouse beats Postgres for analytics — and when it doesn't
A concrete decision framework for picking an analytical store. Benchmarks, tradeoffs, and the cases where Postgres still wins.
Jan 15, 2026 · Anthra AI Team · 4 min read
Read article
AI
The LLM eval harness we wish we'd built sooner
A walkthrough of the minimal evaluation pipeline we run on every LLM project — code, structure, and the lessons that made us build it.
Anthra AI Team · Mar 19, 2026 · 3 min read
AI
Fine-tuning LLMs in 2026: when it's worth the effort
Fine-tuning vs prompting vs RAG — the decision framework, economics, and pitfalls. When does custom training pay off?
Anthra AI Team · Mar 12, 2026 · 4 min read
Product Analytics
Building an internal analytics platform: the 14-week playbook
How to replace your Mixpanel/Amplitude bill with a production-grade ClickHouse-based analytics stack — week by week.
Anthra AI Team · Mar 5, 2026 · 5 min read
Infra
Edge or origin? A decision framework for latency-sensitive features
When to move compute to the edge, when to stay at origin, and the hybrid patterns that work in practice.
Anthra AI Team · Feb 26, 2026 · 8 min read
Product Analytics
Event schema design: what every product team gets wrong
Naming, versioning, properties, and ownership. The event-taxonomy decisions that determine whether analytics are trustworthy six months from now.
Anthra AI Team · Feb 19, 2026 · 6 min read
AI
Choosing a vector database in 2026: a practical comparison
pgvector, Pinecone, Qdrant, Weaviate, Vespa — compared across performance, cost, operational complexity, and feature maturity.
Anthra AI Team · Feb 12, 2026 · 9 min read
Infra
Cloud cost: a checklist before your next AWS bill surprise
A concrete audit checklist to cut AWS spend without re-architecting. Works through the top 12 cost categories in order of impact.
Anthra AI Team · Feb 5, 2026 · 7 min read
Data
Kafka topic design: the 5 mistakes we see most often
Partitioning keys, retention, compaction, naming, and schema evolution. The design decisions that become expensive when wrong.
Anthra AI Team · Jan 29, 2026 · 6 min read
AI
RAG evaluation: the tests we run before shipping any LLM feature
A production-grade evaluation harness for retrieval-augmented generation. Golden datasets, LLM-as-judge, retrieval metrics, and regression gates.
Anthra AI Team · Jan 22, 2026 · 5 min read
Newsletter
Newsletter
Get new engineering essays and practical AI notes in your inbox.