No posts found matching your search.
Introducing Mulu 1.5: Smarter Agents, Better Routing
Mulu 1.5 brings a rewritten agent runtime, intelligent model routing that picks the right AI for each task, and a redesigned deploy pipeline. The biggest update yet.
GPT-5.4 Is Now Available in Mulu
We added GPT-5.4 to Mulu today. OpenAI says it reaches 57.7% on SWE-Bench Pro (Public), 75.0% on OSWorld-Verified, and 82.7% on BrowseComp, with a few current frontier-model comparisons inside.
Team Mode: Different AI Models Working Together
Pick 2–4 AI models, assign them roles like frontend, backend, or reviewer, and watch them collaborate on a shared blackboard. Claude, GPT, Gemini, and more — building your app as a team.
How We Tackled Cold Start Times
Deployed apps were taking too long to spin up. Here's how we approached the problem: lazy module loading, connection pooling, and the edge architecture behind our hosting stack.
5 Prompting Patterns That Ship Better Apps
Small changes in how you describe what you want can dramatically improve the code Mulu generates. These five patterns work across every model.
Mulu 1.5 Lite: Fast and Focused
A lightweight version of Mulu for quick edits and prototyping. Same AI, smaller footprint. Perfect for when you just need to get something done.
How to Build a SaaS Dashboard With Mulu
Walk through building an analytics dashboard with charts, user auth, and a billing page — from a blank project to a live URL using plain English prompts.
Under the Hood: How Multi-Model Routing Works
Mulu supports Claude, GPT-4o, and Gemini. This post covers how we decide which model to use, how fallback chains work, and automatic model selection.
Why We're Building Mulu in Public
Most dev tools are built behind closed doors. We chose the opposite approach. Here's why we share our roadmap, engineering decisions, and even our mistakes openly.
What's New: February 2026 Changelog
Everything that shipped this month: agent improvements, deploy updates, UI polish, and bug fixes.
Why We Use Tree-sitter for Code Intelligence
Tree-sitter gives us real ASTs, incremental reparsing, and cross-language support. A look at how we integrated it into our indexing pipeline.
Getting the Most Out of AI Agents
Break tasks into steps, give context upfront, use checkpoints. Practical advice for getting better results from Mulu's autonomous agents.
Mulu vs. Traditional IDEs: What's Different
An honest look at where Mulu shines compared to VS Code, Cursor, and Windsurf — and where traditional IDEs still have the edge.
One-Click Deploy: The Architecture Behind It
How we built a deploy system that goes from local project to live URL. Cloudflare Workers, edge caching, and the approach that makes it work.
Designing for Non-Technical Users
Mulu is built for people who've never written code. Here's our approach to making AI tools accessible — the UX decisions and the features we cut.
Prompt Engineering Mistakes That Waste Your Time
Vague instructions, missing context, skipping examples — common prompting mistakes that lead to bad output. Here's how to avoid them.
The Future of Building Software
A short note on where we're headed and why we think everyone should be able to build software, not just developers.