Cedric Charly (@cedricgc) profile picture

systems eng

California, USA
Joined April 2011
Photo by Cedric Charly (@cedricgc)Photo by Cedric Charly (@cedricgc)Photo by Cedric Charly (@cedricgc)Photo by Cedric Charly (@cedricgc)Photo by Cedric Charly (@cedricgc)
Cedric Charly retweeted
Today we’re announcing that OpenAI will acquire @pierrecomputer to accelerate our vision for making Codex a place where teams can build software together with AI, from first idea to the version they ship. Pierre’s open-source developer tools already power parts of Codex today. Together, we’ll build on that foundation for the agentic era. openai.com/index/openai-to-a…
281
292
190
4,638
560,226
the craft and care this team puts into everything we build never ceases to amaze me proud to be part of it. excited for what's next
an announcement from the pierre computer company
1
14
388
it’s called we do a little coding
Signups are now open on Code[dot]Storage. Today, more new repos are created per second on Code Storage than any other platform. Lovable, Bolt, Amp, Poke, and many more of the largest agentic workloads store their code and agent memory at scale using Code.Storage.
10
2,042
Cedric Charly retweeted
For the past few months, @pierrecomputer has been working on a brand new primitive, Trees. Just like Diffs, it’s an incredibly fast, modern, and beautiful library for rendering lists of files and folders.
46
107
30
2,067
171,256
I think often how different our software dev process would be if the popular VCS could easily support checking in all our build tools
1
393
My power grows when I start trending on HN I can feel it
2
7
533
Would love a video on debugging a Nix flake
engineering.fb.com/2022/11/1… Monorepo and stacked diffs support is what I wanted for so long. Source control is at the root of software development, can't shift left more than this. I expect Sapling to have knock on effects on how we build tools and develop software going forward.
1
2
We had our first CUE community call focused on documentation: youtu.be/tuidWYIpD7w Highly recommend watching. I think what we discussed will have wide effects on how we build software in a few years
1
2
5
Also check out diataxis.fr/ I think I lacked a theory of mind on how to approach documentation and the diataxis system cleared it up immediately. CUE will adopt this methodology going forward for docs, which I know people have wanted to see for awhile😄
1
The natural endpoint for a configuration language is an eventual configuration database Truly massive configurations can flip all the tradeoffs on its head Details like storage of keys, sharding the structure, and reconciling new data will be done first party in the language
1
3
I like the framing determinate.systems has for Nix. Cross-language dependencies is a massive pain point no matter the stack, so this is a strong place for Nix to make inroads for the average developer. h/t @grhmc
1
16
I will need to dig into Riff to understand the strategy, but Riff is willing to integrate with a language ecosystem and interface with Nix, instead of the Bazel/Nix approach of modeling the world in its own system as a first step. determinate.systems/posts/in…
1
7
Cedric Charly retweeted
matt-rickard.com/the-unreaso… Make is lindy for a reason. It's funny that powerful build systems like Bazel exist but they haven't reached the ubiquity of our elder build systems. I think something like @dagger_io fits the bill as a modern make equivalent.
1
7
19
Breaking out and scaling up the K8s API server is key for long term ubiquity We are going to need *many* CRDs to model all the infra out there
Hitting that limit on custom resources? 😰Upbound can help you scale! @internegz talks you through how to scale to THOUSANDS of CR’s in his latest blog ➡️ buff.ly/3dmlTc9
2
Cedric Charly retweeted
Announcement 🔈: the CUE project would like to officially welcome @carmatrocity, @rogpeppe, @mvdan_, and @ddm_tweets. Thread 👇
5
9
9
86
usenix.org/publications/logi… @palats Look at similar systems Observe when they make different tradeoffs
1
2
- Clean resource naming (K8s apiVersion/kind/id fits) -> Recover state from prod, obsolete state files - Save config languages for the parts you want humans to manage -> generate to same resources as SOT inputs - Annealing models what a more sophisticated `tf apply` is
1
2
- Take versioned snapshots of resources as an immutable base for consistency and introspection - Extra work on the generators makes backwards compatibility easier and easier to integrate existing ways of doing things to standard model
2