Blog

Engineering, AI, and craft.

Our teams share field experience, architectural decisions, and lessons learned in production. By seniors, for seniors.

AI12 min
Local-first + CRDT: the new 2026 SaaS architecture
IndexedDB, Yjs, server sync, Postgres materialisation: local-first becomes the 2026 SaaS architecture for write-heavy apps.
July 14, 2026Read
Engineering11 min
Turbopack stable + React 19 Compiler: the end of webpack?
Turbopack stable for Next.js 16, React 19 Compiler, HMR under 100 ms, dev server under 2 s: what the Rust wave really changes for DX in 2026.
June 9, 2026Read
Performance11 min
Hunting the 50 KB ruining your LCP: a project post-mortem
50 KB too many on mobile 4G means 200 ms of LCP gone. Here is how we tracked them down and removed them, step by step.
May 12, 2026Read
Security12 min
Enterprise SSO: 3 months to find the right stack
A project post-mortem: choosing a B2B SSO stack in 2026. Auth0, WorkOS, Clerk, Supabase, and Keycloak compared — and why WorkOS won for our context.
April 14, 2026Read
Strategy11 min
Sovereign AI: GDPR-compliant AI in Europe
EU AI Act, GDPR, Mistral, OVHcloud, SecNumCloud: building a sovereign AI stack in 2026, from model choice to the SMB adoption roadmap.
March 10, 2026Read
Design11 min
Design tokens 2026: a friction-free Figma → code pipeline
Figma Variables, Tokens Studio, Style Dictionary, Tailwind v4 @theme: the automated pipeline that keeps design and code in sync without drift, with WCAG validation in CI.
February 9, 2026Read
AI13 min
RAG in production: 18 months of field lessons
Sequel to our Feb. 2025 article: twelve months later, multimodal RAG, LLM-as-judge in CI, prompt caching, and hybrid search have changed the game.
January 13, 2026Read
Engineering12 min
Edge computing + Durable Objects: the edge-first architecture
Workers + Durable Objects: the promise of stateful edge at 50 ms for 95% of the internet, plus the edge vs region decision framework.
December 9, 2025Read
AI12 min
MCP (Model Context Protocol): the missing AI piece
The open standard that finally unifies LLMs and business tools. Architecture, the 2025 ecosystem, and production patterns.
November 11, 2025Read
Security11 min
Application-level Zero Trust: done with passwords in 2026
Passkeys, SSO, short-lived tokens, and push-free MFA: the 18-month roadmap for getting B2B off passwords. OpenID, SAML, and SCIM covered.
October 7, 2025Read
Strategy10 min
Build vs Buy 2026: the tech SMB decision framework
SaaS fatigue, AI lowering the cost of build, EU sovereignty: 2026 changes the equation. Our 5-criteria framework to decide build, buy, or hybrid.
September 15, 2025Read
Performance11 min
WebGPU in the browser: real B2B use cases
WebGPU has landed in Chrome, Edge, and Safari 26. Concrete B2B use cases, compute shaders, on-device AI, and a production decision framework.
August 12, 2025Read
AI12 min
Autonomous LLM agents: from prototype to production
From AutoGPT demos to durable pipelines: what actually changes to industrialise LLM agents in 2025, and what's taking shape for 2026.
July 8, 2025Read
Engineering11 min
Server Components: 12 months of production hindsight
One year after migrating our Next.js 15 apps to RSC: what broke, what paid off, and the mistakes we won't make again.
June 10, 2025Read
Strategy8 min
Tech strategy for SMBs in 2025: invest, simplify, harden
A French SMB in 2025: where should the tech budget go? Our argued priority order — AI at #1, cybersecurity at #2, simplification at #3.
May 20, 2025Read
Security13 min
OWASP LLM Top 10: securing generative AI applications
Prompt injection, data poisoning, SSRF via tool use: the OWASP LLM Top 10 applied to a production RAG/agent system. Concrete antidotes.
April 28, 2025Read
Design11 min
Tailwind v4 design system: tokens, primitives, composites
How to build a coherent design system with Tailwind v4: CSS-first tokens, typed primitives, accessible composites. The VALRY LABS experience report.
April 2, 2025Read
Performance10 min
Core Web Vitals 2025: INP, LCP, CLS — the field guide
INP replaces FID. LCP gets stricter. CLS stays stable. The field guide for getting every site into the green without wrecking the UX.
March 10, 2025Read
AI12 min
RAG in production: 7 lessons after 18 months of deployments
Beyond the POC: 7 real pitfalls and their antidotes for shipping a RAG system that doesn't hallucinate, doesn't cost a fortune, and doesn't let users down.
February 4, 2025Read
Engineering9 min
Next.js 15 in production: what changes with React 19
Server Components by default, partial prerendering, new caching API: field notes from migrating a Next.js 14 app to 15.
January 14, 2025Read

Engineering, AI, and craft.

Local-first + CRDT: the new 2026 SaaS architecture

Turbopack stable + React 19 Compiler: the end of webpack?

Hunting the 50 KB ruining your LCP: a project post-mortem

Enterprise SSO: 3 months to find the right stack

Sovereign AI: GDPR-compliant AI in Europe

Design tokens 2026: a friction-free Figma → code pipeline

RAG in production: 18 months of field lessons

Edge computing + Durable Objects: the edge-first architecture

MCP (Model Context Protocol): the missing AI piece

Application-level Zero Trust: done with passwords in 2026

Build vs Buy 2026: the tech SMB decision framework

WebGPU in the browser: real B2B use cases

Autonomous LLM agents: from prototype to production

Server Components: 12 months of production hindsight

Tech strategy for SMBs in 2025: invest, simplify, harden

OWASP LLM Top 10: securing generative AI applications

Tailwind v4 design system: tokens, primitives, composites

Core Web Vitals 2025: INP, LCP, CLS — the field guide

RAG in production: 7 lessons after 18 months of deployments

Next.js 15 in production: what changes with React 19