Devs who insist on working locally are going to get absolutely smoked by those who work in the cloud.
I've seen this first hand - you just can't scale past 2-3 concurrent threads locally - you hit local dev env hell super fast (colliding ports, work tree madness, etc).
Compare that to working in the cloud where you can instantly spin up 10 concurrent threads.
Completely independent, isolated VMs with their own checkouts.
This is how I shipped 131 PRs in 7 days.
You might wonder why the heck I need to ship that many PRs - it's because our first real law firm customer is finding all the rough edges which have to be quickly sanded down.
Also, I think $31.54/PR is a fabulous deal.
Every company can have this without building anything.
- sign up to Ona
- deploy our runner to your VPC (GCP or AWS)
- connect your LLM of choice- including bedrock to ensure inference is completely within your VPC
- connect your context sources. All your data stays private and Ona can never read it or view it.
- connect our Slack integration.
You now have your own version of River that is approved to run in some of the word’s largest banks.
There’s a lot of alpha in putting your ego aside by being willing to be cringe, willing to fail in public, willing to ask for what you want and face rejection, etc.
Background agents are creeping into organisations far more than are discussed in public - which is dominated with individual dev use cases and hobbyist usage.
Yes marketing around agents running overnight seems overstated. But many of the patterns I see discussed online are Frankenstein terminal maxxers. This is not how serious automation is happening.
Most eng orgs are actively breaking down their dev process into the unique steps, and figuring out how to hand off more and more of those steps to agents.
The first step is skills and rules, the next logical extension is figuring out how to get those agents to launch proactively in response to events. Then, it’s tuning over time based on how the agent performs.
The ultimate end game here is a software factory - whether you call it that or not. Every eng org I know is doing this. The dev process is being put on rails by agents.
Does this mean engineers are cooked? No. This is purely the next wave of automation, as we’ve seen many others. Same as the move to IDEs, then CI, then CD.
For now pragmatically agents are automating on top of existing tools, but at some point this will flip. Agents will pull humans in, not the other way around. Tools will be designed more agent first. Processes will be redefined to cut the fat.
Everyone is talking about software factories, few are showing how they actually work.
@casualzach_ built one in public over the last 2 weeks and today is the finale. Join us for the wrap up !
9AM, LA // 12PM, NYC // 5PM, UK
https://t.co/riP0XBoDO6
Woke up to 12 merged PRs. Overnight the agents shipped favorites, breadcrumbs, keyboard shortcuts, page duplication, word count, and a recently visited list.
Backlog empty again. Day 9, zero human commits.
https://t.co/3SIamAK8wP
Built with @ona_hq
Yesterday @loujaybee and @casualzach_ discussed how agents take the developer out the loop, accelerate feature implementation, and scaling software factories. Back live in 20 minutes to discuss bug fixing @swfactory_dev with @pietschphi
https://t.co/PUMWuhdtVN
High-performing engineering organizations are running multiple agents in parallel, in the background.
But the teams pushing the frontier are going further and figuring out how agents can work together as a 'software factory'.
The question we hear most is:
- How do you set a software factory up?
- Where are the limits?
- Where our tooling and processes must change?
This project is an attempt to answer that:
- A live product anyone can use
- Built fully by agents, no human written code
- Fully open source, published on GitHub
- Two weeks of daily livestreams
Running until 25th April - join us !
Led by @casualzach_ with updates on @swfactory_dev.
For details, see: https://t.co/QGBhGps5u5
For the next 2 weeks @casualzach_ and I will be live streaming the build of a self-driving product in the open entirely automated with agents using @ona_hq.
Follow along at: @swfactory_dev
All details at: https://t.co/riP0XBoDO6
(many more updates to come!)
Ramp have been crushing it lately. Whilst their story goes broader than their Inspect platform, it’s hard to overstate how much value in having ready access to coding agents via your phone or laptop is.