The fastest way to run your GitHub Actions. Trusted by @supabase, @trychroma, @clerk and 10,000+ others.
San Francisco, CA
Joined January 2024
- Tweets819
- Following11
- Followers8.5K
- Likes945
Pinned Tweet
@sequoia is leading our Series C at a $1B valuation.
Since January, we’ve gone from running 5M CI jobs a week to over 90M. Agents are writing an order of magnitude more code, and all of it still needs to be built and tested before it ships.
Heading to KubeCon, November 9-12? Find Blacksmith at booth 272.
Run GitHub Actions jobs up to 2x faster than GitHub-hosted runners.
Change one line of YAML, keep your workflows, and leave autoscaling, caching, and infrastructure to us.
On November 5, 2026, our security team detected and contained unauthorized access to the system that stores the console output (job logs) of CI jobs run on Blacksmith runners. The access was revoked the same day, and a later attempt by the same party to regain access was blocked. Our investigation is ongoing. At this time there is no evidence that customer exposure goes beyond the job-log store.
Based on what we know at this time, the incident impacted job-log data belonging to approximately 2.6% of customer accounts. We have contacted each of these customers directly with the specifics for their organization, and we continue to work with them. We do not believe any other Blacksmith customer needs to take action at this time. If our understanding changes, we will update this notice and notify any affected customers directly.
We are continuing to actively investigate and to harden our systems, and we will publish a full report once that work is complete. Please reach out to security@blacksmith.sh if you have any further questions.
You can't have fast CI with a concurrency cap.
So Blacksmith doesn't have one, and we go to great lengths to keep it that way.
Managed CI shouldn't mean managing limits. When your agents open 100 PRs at once, none of them should wait in a queue.
To restore a cache, Blacksmith runners pull from many storage servers at once, and under load those bursts kept stalling.
We retuned the network path, and the median restore for caches over 256 MiB fell from 5.5 s to 2.8 s.
The slowest 1% went from 35 s to 17 s.
Turborepo caching is now on by default for free on Blacksmith.
400+ teams are already using it. One monorepo that runs over a thousand Turborepo tasks a day now sees cached runs finish 97% faster.
Large downloads from GitHub are now 22% faster on Blacksmith runners (p90: 54 → 65 MB/s).
We shipped a network optimization and measured it on production traffic, before and after. More coming soon.
ALT Line chart showing GitHub download speeds on Blacksmith runners improving 22%, from 54 MB/s to 65 MB/s, after a network optimization.
Our CEO @adijayaprakash is live on @etnshow at 9:30am ET to talk about our Series C and what's next for code validation.
GM. Friday. A lot going on. Here's the line up:
- @rosstaylor90 (General Reasoning)
- @max_buchan (Valarian)
- @8eanmurray (Fuse Energy)
- @alirhaddad (XRlabs)
- @gluk64 (Matter Labs)
- Richard Starson (Stellios Technologies)
- @adijayaprakash (Blacksmith)
See you on the show.
Blacksmith now supports GitHub Enterprise Server (GHES) and GitHub Enterprise Cloud with data residency.
Companies who require their own GitHub instance or a fixed data region no longer have to pick between security and faster CI.
Learn how to get set up in the comments.
How to set up Blacksmith on your GitHub Enterprise instance: docs.blacksmith.sh/introduct…
Blacksmith is in the Disruptors 60 Class of 2026, curated by @GreenfieldVC at @TechCrunch Disrupt.
Writing code isn’t the slow part anymore. Every generated change still has to be built and tested before it merges. 10,000+ companies run that step on Blacksmith.
Two days at @WeAreDevs talking about what happens when agents write more code than ever and CI becomes the bottleneck.
Our CEO @adijayaprakash covered the shift, Piotr Bejda the infrastructure behind it, and Abbe Piels showed engineers how to move CI to Blacksmith with one line of YAML.
More highlights soon.
We schedule more than 10 million CI jobs a day, and CI is one of the spikiest workloads that exists. When all your developers are pushing code, we routinely see a customer go from 0 to tens of thousands of CPUs for a few minutes and then settle back down. Extrapolate that across 7,000 customers of different sizes and bin packing becomes a genuinely hard technical problem.
No one better than @ajwerner to think about this and solve it. Before designing anything, he built a simulator on top of our production data so he could replay real traffic against candidate designs and see how each one held up.
The simulator is in the post, and you can play with it yourself! Link in comments.
Anthropic reported 25x growth in CI job volume over six months.
For context, that lands above the 99th percentile in our data. They’re part of a small group experiencing explosive growth, but they’re not the only ones.
Across more than 1,200 organizations from March to August, half grew by more than 1.87x. But the fastest-growing 10% grew 5.95x, and the fastest-growing 1% grew 19.24x!
The companies pushing hardest into agentic coding are giving us a preview of what's coming. It’s a simple formula really: more code = more changes = more tests = a lot more CI.
It’s great that software companies can pump out more code than they ever imagined, but someone still has to validate it. As AI code-gen scales, CI becomes the backbone of the modern org, where it needs to absorb an entirely new scale and pattern of work that didn’t exist before.
If you’re curious how the fastest moving companies are navigating code validation in this new era, then stay tuned for our upcoming Blacksmith CI index; we’ve got a lot more to share.
Anthropic source in first comment.
How do you know when a scheduling approach stops working as you scale?
We built simulations to find out, and used them to validate our scheduler designs as we built them.
Andrew Werner (@ajwerner) ran them live at infra.nyc last week.
Thanks to Megan for having us, and to Chelsea Dole and the other speakers for a great evening.
Interactive version + technical deep dive coming this week 👀
Before it became an anvil head, [code]smith almost ended up as a hammer with eyes, a literal blacksmith, or a clanker robot.
We tried them all.
Here’s the story behind how we designed the character, brought it to life, and made room for some fun personality in a product that takes performance really seriously.
Read more here: blacksmith.sh/blog/does-a-co…
Container caching is now on by default for every Blacksmith org. And it’s free.
Every Docker-based CI job pulls the same images over and over: Postgres, Redis, Compose stacks, base images.
On Blacksmith, those pulls disappear. We persist each org’s Docker image store on a sticky disk, so the images are already local when the next runner starts.
No config. No cache keys. No workflow changes.
Across six real CI jobs, this made them 25–89% faster end to end.
More details in the blog soon.
ALT Line chart of six CI jobs showing total job duration dropping 25% to 89% after container caching was enabled.
Heading to LDX3 New York, September 15-16? Come find Blacksmith at booth 102 👋
Waiting on GitHub-hosted runners? Still self-hosting CI?
Blacksmith is fast CI for GitHub Actions, running your jobs up to 2x faster on high-performance bare metal.
Change one line of YAML and keep your existing Actions and workflows. We’ll handle the autoscaling, caching, and infrastructure behind them.
Stop by booth 102 and tell us what your builds look like.
#LDX3NewYork #LeadDev #Blacksmith #DeveloperExperience #CICD
5
1,215