Our notes on software engineering, quality assurance, and AI-augmented delivery
Everything here comes out of delivery work rather than a content calendar. We write when something is worth writing down — a testing approach that survived contact with a real release, a front-end decision that turned out to matter later, an honest account of where AI tooling helps and where it quietly costs you time. If a post makes a claim about automation coverage or regression time, it is a number from a project we ran.
We build, ship, and test software end to end. Tell us what you're working on and we'll suggest next steps.
Our AI engineering work is backed by credentials issued by the model vendor itself, not a self-assessment.
Claude Code in Action
Issued by Anthropic · June 2026
Introduction to Model Context Protocol
Issued by Anthropic · June 2026
Introduction to Agent Skills
Issued by Anthropic · June 2026
What Develune does, what we build with, and whether we take on work someone else started
Develune is a full-stack software development company. We take products the whole way — discovery and design, front-end and back-end engineering, mobile, cloud and DevOps, data, AI features, and the automated testing that keeps releases safe. Clients come to us when they want one team accountable for the outcome rather than a build handed between specialists.
TypeScript, React, and Next.js on the front end; Node.js, Ruby on Rails, Python, PostgreSQL, and GraphQL on the back end; React Native and Flutter for mobile; AWS, Google Cloud, Vercel, Docker, and Terraform for infrastructure; Cypress, Playwright, Selenium, and Cucumber for testing; and Claude, Gemini, and the Model Context Protocol for AI work. We choose the stack that fits the product rather than defending a favourite.
Yes, and it is a regular part of our work. We start with a codebase audit covering architecture, dependency risk, security exposure, and test coverage, then give you an honest assessment of what is salvageable and what is not. We modernize incrementally wherever possible and only recommend a rewrite when we can show the reasoning.
Two things. We build AI features into client products — LLM applications, retrieval systems, document processing, and agents — with evaluation suites, guardrails, cost ceilings, and human review on anything consequential. We also use AI tooling internally to move faster, with an engineer reviewing every change before it ships. Our team holds Anthropic certifications in Claude Code, the Model Context Protocol, and Agent Skills.
