What is 4DA in one sentence?
4DA reads the internet for developers — privately, locally — and gets sharper every day. It's a Tauri 2.0 desktop app that scans 20+ sources, scores everything against your actual codebase, and rejects what isn't for you — 92% of incoming content, measured across 9 developer personas.
How does it get sharper?
It learns from how you engage with what it shows you. Items you complete, save, or share strengthen related signals; items you bounce or dismiss weaken them. Yesterday's noise becomes tomorrow's signal. All learning happens locally on your machine — nothing leaves your computer.
How is this different from daily.dev or Feedly?
daily.dev personalizes by engagement (what you click). Feedly aggregates by subscription (what feeds you add). 4DA scores by relevance to your actual codebase -- your Cargo.toml, package.json, and Git activity define the filter. It's categorically different: project-aware scoring instead of behavioral or topical filtering.
Does 4DA read my source code?
4DA reads manifest files (Cargo.toml, package.json, go.mod, etc.) and Git metadata to build your technology profile. It does not read your actual source code files. The profile is built and stored locally -- nothing leaves your machine.
Can I use it without any API keys?
Yes. The free tier works entirely without API keys for scoring and content surfacing. AI-powered features like daily briefings are also free -- just add an LLM provider. Use Ollama for fully local inference or provide your own API key (BYOK) for Anthropic or OpenAI.
What happens if I never pay?
Nothing changes. The free tier includes all 20+ sources, the full 5-axis scoring engine, AI daily briefings (BYOK), behavior learning, the MCP server with 14 tools, and the CLI. The free tier never expires. Signal features include a 14-day trial so you can evaluate before subscribing. Most individual developers will never need Signal.
What's the confirmation gate?
The confirmation gate requires 2+ independent scoring axes to agree before an item enters your feed. A single keyword match isn't enough. A single interest match isn't enough. Independent confirmation from multiple relevance dimensions is required. This is what makes the high rejection rate possible.
How big is the download?
About 30 MB on Windows and macOS; the Linux AppImage is larger (~112 MB) because it bundles its own runtime. 4DA is built with Tauri 2.0 (Rust), not Electron. Pre-built binaries with auto-updates for Windows, macOS, and Linux.
Is it open source?
Source available under FSL-1.1-Apache-2.0. You can read, audit, and self-host. The license converts to Apache 2.0 (full open source) after 3 years. The MCP server is Apache-2.0 licensed.
