Studio playbook.
Notes on shipping software across surfaces, hiring a product studio, and the engineering decisions behind AI products that actually go to production. Written by the people doing the work.
First-time founder's guide to building a software product
You have an idea but no technical background. Here's how to scope it, how much it should cost, what an MVP actually is in 2026, and the seven mistakes first-time founders make before talking to a studio.
How to hire a product studio in 2026: a buyer's checklist
Most "find an agency" guides are written by agencies. This one is from inside a studio. The seven questions that separate a real partner from a pitch deck — and the red flags that should stop the conversation.
Build vs buy: a CTO's framework for AI products
A decision matrix for the question every product CTO is being asked in 2026. When custom AI is genuinely worth the cost, when off-the-shelf is the right answer, and when neither — and you should wait six months.
RAG vs fine-tuning in 2026: a decision matrix
After shipping three production RAG systems and one fine-tune, here's the framework we use. When retrieval beats training, when training beats retrieval, and when you actually need both. With the cost math.
Building AI agents with guardrails for regulated industries
From shipping voice agents to a regulated SaaS company. The seven controls a risk officer actually wants to see, the patterns that keep agents in scope, and why most demo-ware fails the first compliance review.
What does it cost to build a SaaS, AI agent, or multi-surface app?
Real numbers from real engagements, in 2026 dollars. Three brackets: the lean MVP, the production-ready build, and the multi-surface ship. With the line items most studios won't put in a proposal.
One spine, many surfaces: how to ship a product to web, mobile, desktop and AI without losing the thread
Our internal methodology, in writing. The fixed structure that lets a small team ship across six surfaces without translation drift. With the token-audit ritual that keeps it honest.
Fractional CTO vs product studio vs in-house team: which model when?
Three engagement models, three failure modes, three sweet spots. A side-by-side that answers the question every founder asks in their first hire conversation.
MCP servers in production: lessons from shipping three connectors
Anthropic's Model Context Protocol is the new integration layer. After three production deployments, the patterns that work, the auth model that doesn't, and the one mistake we keep seeing.
Tauri vs Electron vs PWA: choosing a desktop runtime in 2026
The three real options for shipping a desktop app in 2026, with the trade-offs that actually matter — bundle size, native APIs, update story, security model. From a studio that ships against all three runtimes.
AI for financial services: what compliance actually approves
From shipping AI to a commodity trading desk. The architectural patterns that pass risk review, the ones that don't, and the language that gets you through procurement faster.
Legal-tech AI: why citation beats fluency every time
After shipping a grounded legal-research surface with a zero-hallucination policy. Why the retriever is the product, the model is replaceable, and what "citation chain" actually means in practice.
Reducing alert fatigue in clinical software: design patterns that reduce noise
From a 240-bed ICU deployment that cut dismissed alerts by 41%. The five design decisions that turn alarm noise into signal — without adding new sounds to a clinician's already-overloaded ear.