Blog
Thoughts on building with AI
Practical insights from shipping AI products. No hype, no fluff, just what actually works.
The honest answer to 'Do I even need a development team anymore?' (It depends on exactly these 4 things).
The question isn't "developers or no developers." Every founder is asking it wrong. The honest answer depends on four things: whether your core risk is generation or judgment, whether your infrastructure is a commodity or a moat, whether failure is recoverable, and whether your competitive advantage
How Google Gemini 2.5 Pro Is Redefining AI Reasoning
Google's Gemini 2.5 Pro introduces hybrid reasoning that combines chain-of-thought with real-time search grounding. Here is what it means for product teams.
The Rise of Multi-Agent Systems in Production
Multi-agent AI architectures are moving from research papers to production codebases. Here is how teams are deploying them and what to watch out for.
RAG in 2025: What Actually Works in Production
Retrieval Augmented Generation has matured significantly. Here is what separates production-grade RAG from demo-quality implementations.
Why We Switched from LangChain to Custom Pipelines
LangChain helped us prototype fast, but production demands pushed us toward custom AI pipelines. Here is why and how we made the switch.
Building AI Features That Users Actually Trust
Most AI features fail not because the model is bad, but because users do not trust the output. Here is how to design AI features that earn confidence.
OpenAI GPT-5: What Changed and What It Means for Developers
GPT-5 landed with native tool use, improved reasoning, and a 1M token context window. Here is what actually matters for building products.
Fine-Tuning vs Prompt Engineering: When Each Makes Sense
The fine-tuning debate continues. Here is a practical framework for deciding when prompt engineering is enough and when fine-tuning is worth the investment.
Anthropic Claude 3.5 Sonnet: The Developer's Favorite Model
Claude 3.5 Sonnet has become the go-to model for code generation and complex reasoning. Here is why developers prefer it and where it falls short.
The Real Cost of Building AI Products in 2025
AI API costs, infrastructure, evaluation, and the hidden expenses nobody talks about. Here is what it actually costs to build and run an AI product.
Supabase pgvector: Building Vector Search Without the Complexity
You do not need a separate vector database. Supabase with pgvector gives you embeddings, similarity search, and your relational data in one place.
AI Code Review: How We Use LLMs to Catch Bugs Before Production
We integrated AI-powered code review into our CI pipeline. Here is what it catches, what it misses, and the setup that works.
Choosing Between React Native and Flutter in 2025
The cross-platform framework debate is not settled. Here is an honest comparison based on shipping products with both in 2025.
Edge Functions vs Traditional APIs: When Serverless Makes Sense
Not everything should be serverless. Here is a practical guide to choosing between edge functions and traditional API servers for product backends.
Why MVP Does Not Mean Low Quality
The biggest misconception about MVPs is that they should be throwaway code. Here is how we build MVPs that scale into real products.
Cursor, Copilot, and Windsurf: AI Coding Tools Compared
We use AI coding tools daily across our team. Here is an honest comparison of Cursor, GitHub Copilot, and Windsurf based on real engineering workflows.
How to Evaluate AI Models for Your Product
Picking the right AI model is not about benchmarks. Here is the evaluation framework we use to choose models for production products.
TypeScript Strict Mode: Why We Enforce It on Every Project
Strict TypeScript catches bugs before they reach production. Here is why we made it mandatory and how it changed our code quality.
AI Prompt Management: Treating Prompts Like Code
Prompts buried in application code create chaos. Here is how we manage AI prompts with version control, testing, and deployment pipelines.
The Non-Technical Founder's Guide to Hiring Engineers
You do not need to understand code to hire great engineers. Here is how non-technical founders can evaluate engineering talent effectively.
AI Hallucination: Detection Strategies That Actually Work
AI hallucinations are the biggest risk in production AI products. Here are the detection and mitigation strategies we use across client projects.