Blog
Technical notes & production insights.
Methodology, software architecture, AI in production: our hands-on feedback, without unnecessary jargon.
Jun 15, 2026 · AI · 6 min
Open WebUI: the self-hosted interface for your LLMs
The open-source, self-hosted interface that brings the ChatGPT experience to your own models—local or cloud, with built-in RAG, under your control.
Read →
Jun 11, 2026 · Web development · 5 min
KYC and AML for a marketplace: how to avoid getting blocked
A marketplace handles third-party funds: KYC and AML are mandatory. The practical building blocks and the pitfall of addressing them too late.
Read →
Jun 11, 2026 · AI · 6 min
LangChain, LangGraph, LlamaIndex: Which Framework for Your AI?
LangChain for quick assembly, LangGraph for stateful agents, LlamaIndex for document RAG—or none at all. How to decide.
Read →
Jun 11, 2026 · Web development · 5 min
Algolia or Meilisearch: the search that converts
Slow or imprecise search means lost sales. Algolia is managed and feature-rich, Meilisearch is open source and sovereign: how to choose.
Read →
Jun 11, 2026 · Web development · 5 min
React Native: a mobile app without doubling the teams
One codebase for iOS and Android. When React Native is the right choice, and when to prefer native or a PWA.
Read →
Jun 11, 2026 · AI · 5 min
MCP (Model Context Protocol): Connect Your Tools to AI, Cleanly
MCP standardizes how an LLM accesses your data and tools—controlled, reusable, auditable. Here’s what changes.
Read →
Jun 11, 2026 · Web development · 5 min
Auth0 or Keycloak: Which SSO Solution for Your Application?
Auth0 (managed and fast), Keycloak (open source and sovereign), Supabase Auth (built-in). How to choose your authentication layer.
Read →
Jun 11, 2026 · AI · 5 min
Hosting an open source LLM: vLLM, Ollama, and sovereignty
When sovereignty or scale demands it, self-hosting an open source model is the way to go. Ollama for getting started, vLLM for production.
Read →
Jun 11, 2026 · AI · 5 min
pgvector or Qdrant: Which vector database for your RAG?
pgvector stores your embeddings in PostgreSQL, while Qdrant is a dedicated vector database. When one suffices, and when the other is essential.
Read →
Newsletter
A letter, only when we truly have something to say.
Production insights, AI updates, and behind-the-scenes of our projects — straight to your inbox.