From the studio
Notes from a small studio shipping vertical SaaS
Honest write-ups on what we ship, what breaks, and what we would do differently. No growth-hack think pieces. No influencer threads.
The Case for Self-Hosting Your Own Analytics in 2026
Self-hosting analytics 2026 is not just a privacy stance. It is a cost, control, and reliability decision. Here is how we picked Umami and where we kept hosted tools.
Read article
Claude Code vs Cursor vs Copilot for Serious Dev Work in 2026
Claude Code vs Cursor vs Copilot compared on real production workloads: context, agent loops, pricing, and the tasks each one wins. The choice we ship on.
Read article
Why We Use Claude Code for Everything (and What We Tried First)
Claude Code coding agent runs almost all our engineering across four SaaS. Here is why we switched off Cursor, what changed, and the specific workflow that compounds.
Read article
Umami vs Plausible vs Matomo in 2026: Which Self-Hosted Analytics?
Umami vs Plausible vs Matomo compared on real production criteria: RAM, query depth, ownership, and the use cases each was built for. The choice we ship on.
Read article
Customer Development With the Mom Test: Lessons From Carriva Interviews
Mom Test customer interviews SaaS founders run badly more often than they admit. Here is how 20 calls with retirement advisors changed Carriva, with the exact questions we now ask.
Read article
Coolify vs Dokku vs CapRover: Self-Hosted PaaS for Solo Founders
Coolify vs Dokku vs CapRover: an honest comparison for solo founders. The dealbreakers, operational overhead, and which one we ship on across our SaaS portfolio.
Read article
Solo Dev DevOps: My Proxmox + Coolify Homelab Setup
Solo developer homelab Proxmox Coolify is the unfashionable answer to running production SaaS on a budget. Here is the setup, the routines, and the failure modes.
Read article
Niche SaaS in Saturated Markets: How to Win a Sub-Niche
Niche SaaS strategy in saturated markets is not about being smaller. It is about being more specific. Here is the wedge framework we use across five products.
Read article
From Idea to Production in 6 Weeks: How PrepareMesCours Was Born
Ship SaaS in 6 weeks is the story we tell new founders, and PrepareMesCours is the proof. Here is the week-by-week build, the cuts we made, and what we kept.
Read article
Customer Interview Questions That Do Not Lead the Witness
Customer interview questions framed wrong tell you what you wanted to hear. Here is the script we use, the followups, and the cues that signal you are leading.
Read article
The Cost of AI-Generated Content for SaaS Marketing (Real Numbers)
AI content marketing cost is rarely shared honestly. Here are our real per-article numbers across 4 SaaS, the human-edit overhead, and the SEO outcomes.
Read article
B2B SaaS Landing Page Patterns That Actually Convert in 2026
B2B SaaS landing page patterns that actually move conversion: the structure templates, the sections that pull weight, and the ones to cut. Tested across five products.
Read article
Building a Localized SaaS for 10 LATAM Countries (Creaclases)
Localizing SaaS multiple countries is harder than it looks when curriculum, currency, and slang differ. Here is how Creaclases handles 10 LATAM countries from one monorepo.
Read article
Picking and Locking the Right Node.js Version for Production in 2026
The Node.js version 2026 picture: which line to run, which to skip, and how to lock it across CI, Docker, and your editor without painting yourself into a corner.
Read article
Why We Ditched PostHog for Umami (and When We Would Switch Back)
PostHog vs Umami self-hosted: an honest cost and feature breakdown after running both. Why we moved most products to Umami, and why Carriva still pays for PostHog Cloud.
Read article
Building a Typed Prompt Library in TypeScript for Production LLMs
A typed prompt library TypeScript pattern: schemas at both ends, versioned templates, provider routing. The setup we use across two production LLM apps.
Read article
RAG for Regulated Industries: Lessons From Carriva
RAG regulated industries the right way: how we built Carriva's retrieval layer for retirement-advisory, the failure modes we caught, and the citation discipline that keeps the LLM honest.
Read article
SaaS Pricing Models in 2026: Per-Seat, Usage, Hybrid, and Credits
SaaS pricing models 2026: per-seat, usage-based, hybrid, and credits compared. The decision framework, the migration patterns, and what AI features break in each.
Read article
Pricing AI Features Without Scaring SMBs Away
Pricing AI features for SMB customers is the part most founders get wrong. Here is the math, the customer reactions, and the simple model that worked across four products.
Read article
Vector Databases Compared in 2026: pgvector, Qdrant, Pinecone, Weaviate
Vector database comparison 2026: pgvector, Qdrant, Pinecone, and Weaviate side by side on cost, latency, ops overhead, and the failure modes nobody puts on the marketing page.
Read article
Shipping 4 SaaS in Parallel as a 1-Person Studio (and What Broke)
Solo founder shipping multiple SaaS sounds like a flex. It is mostly a discipline problem. Here is what worked, what broke, and the parts we are still embarrassed about.
Read article
TypeScript Strict Mode Flags That Actually Catch Bugs (and the Noise)
TypeScript strict mode flags pay off unevenly. Some catch real bugs. Some create noise that drowns the signal. Here is the ranked list we apply across five SaaS apps.
Read article
Self-Hosting Postgres in 2026: When It Pays Off, When It Does Not
Self-hosting Postgres saves real money, but only if you actually do the work. Here is the math, the failure modes, and the line we draw between owned and managed.
Read article
Postgres jsonb vs Typed Columns: When to Pick Which (Real Examples)
Postgres jsonb vs columns is not a religious choice. It is a query-pattern question. Here is the decision rule we use across multi-tenant SaaS, with real schema snippets.
Read article
Multi-Tenant From Day 1: A Small Studio's Playbook
Multi-tenant SaaS architecture sounds heavy until you do it on day one. Here is the small-studio playbook we use across Carriva, PrepareMesCours, and Creaclases.
Read article
Claude vs GPT vs Gemini: Choosing for Production Workloads in 2026
Claude vs GPT vs Gemini is not about benchmarks. It is about which model holds up under your actual production traffic. Here is what we measured across five SaaS apps.
Read article
The Case for Vertical AI SaaS in 2026 (vs Horizontal Copilots)
Horizontal AI copilots are commoditized. Vertical AI SaaS still has moats. Here is the founder math we use to choose niches, and why we think 2026 sharpens the gap.
Read article
RAG vs Fine-Tuning in 2026: A Decision Framework That Works
RAG vs fine-tuning is not religion. It is a budget and update-frequency question. Here is the decision framework we use across regulated and consumer apps, with real numbers.
Read article
Why We Built Carriva: 18 Months Shipping a SaaS for Retirement Advisors
Eighteen months ago we picked a vertical most engineers run from: French retirement-advisory firms. Here is what we got right, what we got wrong, and why we still bet on it.
Read article
Next.js 15 Standalone Docker Output: A Production Guide That Works
Most Next.js 15 standalone Docker images break the first time because the trace misses one file. Here is the production setup we run on five SaaS, with image sizes and real numbers.
Read article