This Year’s Most Influential Open Source Projects (My Honest Picks)
Every December, the open source community publishes “year in review” lists. The lists are useful but predictable. The most popular projects on GitHub get the most attention. The projects that changed my workflow rarely make those lists. This is my honest list of the open source projects that mattered most in the last year — not the most starred, the most impactful for actual work.
The Methodology
I have been tracking which open source projects I actually use. Not the ones I read about. Not the ones that show up on Hacker News. The ones I open daily or weekly. The list below is the result. I excluded projects I’ve covered in detail in other articles (the ones that are obvious). I included projects I have not covered in detail. The right test for “influential” is whether the project changed my workflow. The right test for you is whether the project changes your workflow.
Project 1: Zed (The Editor That Made Me Switch)
Zed is a code editor written in Rust. The performance is unmatched (it opens a 100MB log file in under a second). The interface is minimal. The language server support is the same as VS Code. The extensions are fewer than VS Code but the ones that exist are good. The right test for an editor is whether you stop fighting it. I stopped fighting Zed within a week. I switched from VS Code. The trade-off: the extension ecosystem is smaller. The themes are fewer. The integrations with non-Microsoft tools are weaker. The right answer for most developers is to try Zed. The right answer for those who depend on a specific VS Code extension is to wait. Zed has changed my workflow. The performance is the killer feature. The right test is whether you can deal with the smaller extension ecosystem. For most developers, the answer is yes.
Project 2: Ollama (The Local AI Runtime)
Ollama is the easiest way to run a local LLM. The installation is a single command. The model management is one command per model. The API is OpenAI-compatible. The trade-off: the model selection is smaller than the OpenAI/Anthropic APIs. The performance is limited by your hardware. The right answer for local AI is Ollama. The right answer for serious production AI is the cloud. Ollama has changed my workflow. The local AI is good enough for 80% of my work. The right test is whether the model you need is available. The right test for you is whether the local model is good enough for your use case.
Project 3: Bun (The JavaScript Runtime That Replaced Node)
Bun is a JavaScript runtime written in Zig. The performance is 2-4x Node.js. The TypeScript support is built-in. The package manager is faster than npm/yarn/pnpm. The trade-off: the ecosystem is not as mature as Node.js. Some npm packages do not work. The right answer for new projects is Bun. The right answer for legacy projects is Node.js (until the Bun ecosystem matures). Bun has changed my workflow. The startup time is 10x faster. The install time for dependencies is 5x faster. The right test is whether your dependencies work. The right test for new projects is yes.
Project 4: Litestream (The SQLite Backup Tool)
Litestream is a tool that streams SQLite changes to S3-compatible object storage. The setup is a single binary. The recovery is one command. The cost is the S3 storage. The right answer for a small service that uses SQLite is Litestream. The right answer for a large service that uses Postgres is the Postgres backup tools. Litestream has changed my workflow. The “what if the disk dies” question is answered. The backups are continuous. The recovery is fast. The right test is whether your service uses SQLite. If yes, Litestream is the answer.
Project 5: Caddy (The Web Server That Just Works)
Caddy is a web server written in Go. The HTTPS is automatic. The configuration is the simplest of any web server. The performance is good. The trade-off: the module ecosystem is smaller than nginx. The right answer for a new web project is Caddy. The right answer for a complex nginx config is to stay on nginx. Caddy has changed my workflow. The HTTPS setup is one line. The reverse proxy is one line. The wildcard cert is one line. The right test is whether your use case fits Caddy’s model. For most use cases, it does.
Project 6: Biome (The Linter/Formatter That Replaced ESLint + Prettier)
Biome is a linter and formatter for JavaScript and TypeScript. The performance is 10-100x ESLint + Prettier. The configuration is simpler. The output is the same as Prettier. The trade-off: the rule set is smaller than ESLint. The plugin ecosystem is smaller. The right answer for new projects is Biome. The right answer for legacy projects with custom ESLint configs is to stay on ESLint. Biome has changed my workflow. The lint step is now instant. The right test is whether your project fits Biome’s rule set. For most projects, it does.
Project 7: Dify (The LLM App Builder)
Dify is a no-code/low-code platform for building LLM applications. The visual workflow editor is good. The integrations with major LLMs are built in. The deployment is one click. The trade-off: the no-code part has limits. The right answer for prototyping is Dify. The right answer for production is custom code. Dify has changed my workflow. The prototyping time is 10x faster. The right test is whether your use case fits Dify’s model. For most prototyping, it does.
Project 8: Bruno (The API Testing Tool I Already Covered)
Bruno is the open-source Postman alternative. The files are Git-trackable. The CLI is real. The right answer for API testing is Bruno. (Already covered in detail elsewhere.)
What This List Is Not
This list is not the most popular projects on GitHub. The most popular projects are React, Vue, TensorFlow, PyTorch, Kubernetes, etc. The right answer for those is to use them. The right answer for this list is to highlight the projects that are changing workflows but are not the obvious picks. The right test is whether the project is on your team’s radar. The right test for these 8 is “probably not, but should be.” The right test for the obvious picks is “you already use them.” The right answer for the long term is to track both lists. The right answer for the workflow-changers is the list above.
Final Thoughts
The open source projects that matter in 2026 are not the ones with the most stars. The ones that matter are the ones that change workflows. Zed, Ollama, Bun, Litestream, Caddy, Biome, Dify, Bruno — these are the 8 that changed my workflow this year. The right test is whether the project changed your workflow. The right test for you is whether any of these 8 change your workflow. The right answer for the long term is to track workflow-changers. The right answer for the short term is to try the ones that apply. The result is a workflow that gets better year over year. The result is worth the time.