Solomon Hykes

@shykes.bsky.social

Making dagger.io. Before that: founded Docker. "No is temporary, yes is forever".

It's here! Dagger 0.18 can run AI agents - the Dagger Way. Not a monolith: composable system primitives. Cross-language & model-agnostic Not a silo: LLMs are just another node in the DAG Not a black box: deep tracing OOTB Not a blank check: you control the AI's environment

Dagger@dagger.io · last yr.

Build AI agents into your software delivery workflows — with full control, observability, and portability. LLMs are now a core building block in Dagger. utm.guru/uijSj

Our friends at Earthly🌎 announced the end of their original open-source project, to focus on a new product: Lunar. Their founder @vladaionescu.com, being a real mensch, wants to minimize pain to his users. So we are collaborating to offer Earthly users a soft landing to planet Dagger. 🧵

Wait 🤔 you're telling me none of you have typed composition in your MCP servers? What about *object* composition: call a function, get an object back with more functions, pass that to another function, and so on? MCP returning more MCP? You're telling me only @dagger.io has this?

OMG!!! you do not want to miss this one! I will be going live the Solomon Hykes (Founder of Dagger and co founder of Docker) and Dagger team with this exclusive workshop on Dagger and AI agents. This will be fun so be sure to sign up and ask questions live!

Bild

What if the Unix shell stole the best ideas from Docker, Make and Nix? Native supports for containers and secrets; typed objects; declarative execution, sandboxed and cached by default... What if those were just standard shell features? 🤩 Well, we built it! dagger.io/blog/dagger-...

A Shell for the Container Age: Introducing Dagger Shell - Dagger

Powerful, programmable CI/CD engine that runs your pipelines in containers — pre-push on your local machine and/or post-push in CI

dagger.io

MCP is great for extending AI apps, but less useful for extending AI code. It forces a rigid separation between "code called by llm" and "code called by code", when in practice you want to keep that boundary flexible. Right now that nuance is getting lost in the hype.

has anyone tried plugging a llm into an interactive terminal emulator, and watching it interact character by character, at full resolution? It would be fun to watch a model type on a terminal at superhuman speed as it explores its environment.

🚀 Dagger v0.14 is here! 🚀 Enhanced support for Git integration, flexible networking, and powerful new APIs to supercharge your CI/CD pipelines. Dive into all the new features in our latest blog post! dagger.io/blog/dagger-...

Bild