CI for agent-scale dev. 80% cheaper than GH, 70% cheaper than Blacksmith. On dedicated, fast hardware.
Joined October 2026
- Tweets41
- Following24
- Followers241
- Likes270
Pinned Tweet
whoa who made this
Introducing Jog. CI is the bottleneck for shipping code at agent speed. But it shouldn’t be.
@njaremko and I were spending up to $50,000 per month on CI when building our last product. I was running three macbooks as mini GH runners to save costs. Recently a friend was hit with a $20,000 Blacksmith bill.
Anyway, we decided to fix the problem and crush costs down as far as we could in a managed offering on really fast, dedicated hardware. We think it’s pretty neat and we’re saving people a boat load already.
Have fun, ship more and worry about costs a lot less.
all CI eventually flows to jog
Currently I am using a ton of OVH vps with ephemeral runners and I have a overflow queue that dumps into @usejog
This alone cut my monthly CI from $1700
To sub $450
There is definitely the need of much more intervention from day to day but since I’m currently a solo dev it’s not a problem.
Jog retweeted
In recent years, GitHub actions have often been in an unavailable state, with long queuing times leading to CI failures. We can consider using third-party CI services as an alternative. Jog @usejog is a good choice, you can earn 5000 credits through my invitation link. usejog.com/join/handsomezhuz…
Jog retweeted
CI is now #2 on my list of costs, right below tokens.
Our GitHub Actions usage went from $341 in August to $3,096 in October, that's insane!
Had a chat with @mattgapp about some options to save money without slowing everything to a halt!
1) AWS Spot + @RunsOnRunsOn
TIL, Spot is discounted spare computing capacity from AWS. You can rent it cheaper, but AWS can reclaim it when it needs the capacity back.
RunsOn connects GitHub Actions to runners in your own AWS account, so you don’t have to build all the plumbing yourself.
Works even better if you have both GitHub and AWS credits!
2) Rent servers from cheaper providers like @OVHcloud or @Hetzner_Online.
Run your own GitHub Actions runners there, or use @buildkite to manage the pipelines. More stuff to maintain, but worth looking at.
3) Use a managed runner provider like @usejog or @useblacksmith.
Both are pitching cheaper and faster GitHub Actions runners. From my explorations, Blacksmith seems more focused on speed and caching, while Jog is really pushing on price. Both claim savings, so we’ll see what happens on our actual builds.
Probably worth trying them on our actual builds rather than comparing landing-page numbers.
We're going to see a lot more of these popping up! @boatdotdev is another interesting one on the VM/CI side. Will check them out!
Jog retweeted
Had a fun chat with @0xZoZoZo who is shipping like crazy. My recommendation was 1) if you have free money (cloud credits) and you don't have something better to spend them on, then use them for CI or 2) if you don't have free money, then you should consider alternatives like Blacksmith or @usejog. We're very focused on speed and caching btw - check out Jog Max.
CI is now #2 on my list of costs, right below tokens.
Our GitHub Actions usage went from $341 in August to $3,096 in October, that's insane!
Had a chat with @mattgapp about some options to save money without slowing everything to a halt!
1) AWS Spot + @RunsOnRunsOn
TIL, Spot is discounted spare computing capacity from AWS. You can rent it cheaper, but AWS can reclaim it when it needs the capacity back.
RunsOn connects GitHub Actions to runners in your own AWS account, so you don’t have to build all the plumbing yourself.
Works even better if you have both GitHub and AWS credits!
2) Rent servers from cheaper providers like @OVHcloud or @Hetzner_Online.
Run your own GitHub Actions runners there, or use @buildkite to manage the pipelines. More stuff to maintain, but worth looking at.
3) Use a managed runner provider like @usejog or @useblacksmith.
Both are pitching cheaper and faster GitHub Actions runners. From my explorations, Blacksmith seems more focused on speed and caching, while Jog is really pushing on price. Both claim savings, so we’ll see what happens on our actual builds.
Probably worth trying them on our actual builds rather than comparing landing-page numbers.
We're going to see a lot more of these popping up! @boatdotdev is another interesting one on the VM/CI side. Will check them out!
cha ching that’s cash money back in your pocket
you’re going to make me blush
@usejog 's support team deserves a shout, they are always around to help, getting CI set up was easy, website looks stunning 🔥your paid credit carries over and never expires.
Paying accounts with credit also get 5k free minutes each month (read docs).
usejog.com/join/olinkluzz1v
maybe jog is too straightforward? should we add limits or a subscription?
Ok, this post is for every single person/company considering what looks like cheaper CI bc I've been getting a bunch of q's about this. Please listen up! Some of the CI providers you’ve seen pop up are not charging you for compute - they’re charging you for (surprise!) *not* using their compute. Startup fees, limits, plans, subscriptions, credits that expire, less RAM per vCPU etc. You can’t actually use the compute they advertise (at least not for non-trivial CI workloads).
This is funny to me but it hurts my midwestern sensibilities. To me this is obvious because 1) I run a CI company that is powering millions of CI jobs with enough servers to fill a DC, serving petabytes of cache for real companies (Series A, B+), and 2) I come from finance - was an IB finance nerd at J.P. Morgan - and the math on CI compute is simple. You buy compute for X at some longer duration (a month, a year, forever) and rent it at a much shorter duration - per minute or per second. You take a margin on the compute liquidity. It's not rocket science. But you can only charge so little before you start losing money. You can try to over-provision and CPU steal, but that only slows down your customers’ CI and they end up paying more for the same CI job anyway. It’s not complicated math, so trust me: when providers advertise CI that is 10x+ cheaper than Blacksmith, it’s just fake. They’re either not profitable or the real price is in the fine print. Or maybe they run on hopes and dreams. We do not, we run on AMD.
I backtested one of these providers - who I’m not going to name - on our prod data and they’d end up costing 4-9x more than us for the same jobs when you account for startup fees or compute limits or RAM limits.
Blacksmith and the other big players are actually straight shooters when it comes to this, and I respect them for that. If you’re buying compute, buy from someone who’s in the business of making money from compute.
hi Matt
Hi X,
I’m Matt.
@njaremko and I are building the best way to run CI after dealing with $50k/mo CI bills ourselves. @usejog is faster + 80% cheaper than GitHub-hosted runners.
Since launching just three weeks ago, we’ve run millions of CI jobs & served petabytes of cache. We’re now saving hundreds of companies millions of dollars a year.
we sell d̶o̶l̶l̶a̶r̶s̶ CI for less
Today we're launching Dollars 💰
Most AI startups spend so much time and effort building an AI product with negative gross margins. We skipped straight to the end.
We sell US dollar bills for $0.80.
TAM: $1T+ in circulation. We’re going after all of it.
Let’s scale to the moon together.
dollars for less dot com
