CloudBees

Agentic coding is amazing. I'm an optimist. I don't believe in the doomsayers. As technologists, we need to stop saying that it will end the world and put everyone out of a job. Instead we should focus on the absolutely wonderful impact it has on many individuals and companies across the world. That's why I deeply enjoyed my conversation with Brendan Falk. His company Hercules helps everyone build the customized tooling they always wanted. Like a restaurant owner that build his own pig inventory system (believe it or not). No doubt there's risk with AI. Our job as technologists is to manage the risk, and harness the upside. Technology is here to serve humans. Let's focus on that. This is the first of many interviews I will be doing with Codeship Labs to shine a light on the builders and founders that do exactly that. Leverage AI to have a positive impact on the world. Big shoutout to Ashley Sawatsky for helping me bring Codeship back to live. It's been one of the most fun experiences over the last five months since I came back to CloudBees. Check out the link to the full interview and blog post in the comments.

Everyone is frustrated with GitHub at this point. Reliability issues left and right. On top of that, it's entirely unclear if git and our current approach of versioning makes sense going forward. Code is becoming abundant and cheaper to generate. At the same time, the context, prompts and input to what generated the code become more important (and need to be versioned as well). Whether git is the protocol/technology we use going forward or not is hard to predict. What's definitely happening is that source code storage and versioning is being added as a feature to other platforms. Cursor and SpaceXAI are leading the charge and Cursor Origin really sets them apart from the other labs. Generate your code with Cursor, store your code in Cursor. It just makes sense. I'm excited that we integrated Jenkins and CloudBees CI (our enterprise product for Jenkins) with Cursor Origin. Thank you Davy L. Diane Liu Tony Li and everyone who helped make this happen together with Antonio Muñiz Jesse Glick James Nord Jordan Matthiesen Thierry Wasylczenko.

I spoke at the WeAreDevelopers conference in San Jose last week and did an interview with The New Stack. Software engineering and product development is changing tremendously. What's not changing is that we write code for a reason. That was the case before agentic coding, and it's the case now with AI. The goal is not simply generating more code, or more pull requests. What matters is that your code changes positively impact the business. Shipping quickly is important, but there's risk you simply ship more slop. Measuring the impact of every code change is the only way to determine where to spend your time, tokens and human attention. 📺 Link to the full interview is in the comments. Thank you Chris J. Preimesberger for the chat.

We lost the plot. Agents = software If your software hacks another company, you are responsible. Before we run around and ask for more regulation, we should first adhere to existing laws and regulations. Starting by not hacking other companies and acting as if it's not our fault. More importantly, we should accelerate giving companies and allied governments the latest and best models. They need it to defend themselves. Funnily enough, to defend themselves from the labs. Accelerate AI adoption is what we need, not pacing or slowing down because of a 10% risk number that's pretty much made up.

Flaky tests were always an annoying problem. With agentic coding, it's 100x worse. You run many more builds and tests passing are the feedback loop your agent needs. With CloudBees Smart Tests, we can identify your flaky tests and allow you to quarantine them so they stop breaking your builds. Once you isolated them, you can focus on fixing those tests and if we observe that they stopped being flaky, we will automatically include them in your builds again. Read Jake's blog post to learn more. Fantastic new feature Naoto Samyak Yoshiori 👏 https://lnkd.in/gzhdAyhQ

Everyone is talking about software factories these days. Inner loop vs. outer loop Model provider harness vs. third party harness Prompt caching strategy All important topics. Yet what holds large complex enterprises back is that writing code was always just the first step. Getting it all the way into product quickly, safely and securely is the real bottleneck. To truly accelerate software development in the enterprise, we need to remove not just the first bottleneck, but the second, third and fourth. Testing, releasing, validating.