I know a lot of people talk about GH outages, but I personally haven't encountered it even tho as you can see form my profile (github.com/crisdosaygo) I'm on there everyday. Maybe my workflows don't hit the weak spots, idk.
But the reason I created this was because Actions always worked so well for me, and it seemed to have so many possibilities to build things on it.
Regarding the Global Free Tier, every GH account comes with Actions Minutes and this is way to have a nice CLI to put them to use toward your building, and maybe have a spot for agents to do some work you don't want locally.
Bigger picture, I feel GH led the way providing this idea of "compute as utility" (free compute for even free GH accounts, was amazing - but I really think that the future is shaped like that). I'm serious about that: AI will eventually become cheaper to train and infer, and the oversupply of compute will be a background layer we will have access to much cheaper. Just one of the trends. So the idea of the Global Free Tier or Background Compute as a universal utility, is something I think si really real.
Also probably important to note that the reliability issues GH seems to have faced, are more of a recent uptick, but Actions has had this free tier for ages.
Given the consistent outages everywhere on GitHub, they actually should put GitHub Actions under a paid tier only if they want their platform to be sustainable and stable. Period.
It's quite irresponsible of them to have almost all the core features free and a paid tier would significantly reduce abuse of it (and especially GitHub Actions) like this.
If they don't, then don't be surprised to see more outages on their platform.
Tell me more about your workflow! For us, the workflow is, we'd assign the ticket to a bot user we create (broccoli in this case), and broccoli will go spin up a sandbox and do the execution. Do you trigger the task execution from Codex by giving it a linear ID? That was Broccoli v0 but of course still requires you to setup Codex with all the right keys.
Oh got it! In this case, the main difference is that we go through a flow from design to implement using our own prompts, and uses both Codex and Claude Code so they can improve off of each other.
There are many providers popping up every day offering sandboxes, I think Cloudflare is ahead of the game for pricing and performance, that being said it would be super nice to see a huge competitor analysis:
Cloudflare vs e2b vs daytona vs freestyle vs whatever else
It's Arch based, with (iirc) Hyperland as it's "DE", so really not much memory I'd guess.
My desktop runs Arch with Sway (so quite close), three monitors, and uses ~400MB ram after boot. Most of it are the framebuffers. All the rest is eaten by Firefox, rust-analyzer and qemu.
reply