in Database / Backend on Postgresql, Database, Pgvector, Timescaledb, Vector database, Graph database
There’s a meme in the database community: “Just use Postgres.” It started as a half-joke. Now it’s serious architectural advice. In 2026, the question isn’t whether PostgreSQL can handle your use case — it’s whether the operational simplicity of “one database to rule them all” outweighs the theoretical performance advantages of a specialist.
in Ai / Architecture on Mcp, Model context protocol, Ai agents, Anthropic, Tool use, Interoperability
If you’ve worked with AI agents in 2025, you probably felt the pain: every LLM had its own proprietary way of connecting to tools, every platform required custom integrations, and the combinatorial explosion of “Model A × Tool B” glue code was becoming a full-time job. Then Model Context Protocol (MCP) showed up and changed the conversation entirely.
in Architecture / Distributed systems on Durable execution, Temporal, Restate, Distributed systems, Workflows, Reliability
Here’s a scenario every backend engineer has lived through: you write a workflow that calls three external services, processes some data, and updates a database. It works great in staging. In production, it fails on the second service call at 2 AM, leaves the system in a half-finished state, your on-call engineer spends two hours figuring out what happened, and you spend the next week writing idempotency logic, retry handlers, and compensating transactions you should never have needed.
in Devops / Security on Container security, Kubernetes, Devsecops, Docker, Shift left, Supply chain security
Container security has a reputation problem. Most teams treat it as a checkbox — run a scanner, review a report, ignore 90% of the findings because there are too many to fix, ship the container anyway. This approach provides the feeling of security without the substance of it.
in Ai / Software engineering on Llm, Ai engineering, Structured outputs, Production ai, Prompt engineering, Ai architecture
Building an LLM demo takes an afternoon. Building an LLM-powered product takes a completely different mindset. The gap between “I got it working in a Jupyter notebook” and “it’s running reliably in production serving 10,000 users” is enormous — and most of the difficulty isn’t in the AI part. It’s in applying software engineering discipline to non-deterministic systems.