GitHub in 2026: How AI Agents Are Changing How Developers Work

**Mario Rodriguez** (0:00)
I already consider you a developer. You don't want your bank application to be vive-coded. Lower the floor, raise the ceiling, because you could formalize yourself to no value. So I think the new GitHub is going to be all about.

**SPEAKER_2** (0:20)
Hi Mario, and welcome to Inference show by Turing Post. Thank you for joining me.

**Mario Rodriguez** (0:25)
Thanks for having me. It's a beautiful day outside. Yesterday was a little bit cloudy, but today the sun came out, so I'm really happy.

**SPEAKER_2** (0:31)
It's a beautiful day. But let's get to GitHub. Yes. According to similar web, GitHub has over 630 million monthly visitors. And that's an enormous surface area of development, like developing and experimentation. So since late 2025, when the agent workflows really started working, what did change at GitHub? What did you notice?

**Mario Rodriguez** (0:57)
You know, it's an interesting thing. Like you're saying, around December, one of the key things that we ended up noticing is that the model capabilities took a jump up.
And before, if you were going to, what I call a macro delegate to an agent, you constantly had to go and correct the agent. You'd be like, no, you took this path, that you should not have taken that path, or you do this other thing, you should have done that thing. You know, it's kind of like, imagine if you have a toddler, you're constantly going like, no, no, don't go that way. Like go this way, be safe on this other way. And what we notice is that in the December timeframe, that now you could actually go and say, no, go ahead and play, it's safe, and you will get an output of very high quality.
So in my opinion, there's two things that that unlock. In the developer machine, meaning kind of in the workflow, they could macro delegate significantly more, and then kind of micro steer when they need it. And the micro steer did not feel frustrating to them. It didn't feel like, oh my God, I just wasted a bunch of tokens, and like now I need to go and tell you everything you did wrong. It actually was a little bit more like, oh, you did that. Now let me go on a loop with you to create it better. And it felt like you actually are doing this iteratively in a creation process instead of a correction process. So we noticed that at the developer level. And then if you think about it at the agents are running a little bit more autonomous in automation, it meant that they could actually go longer, and it meant you could give it better and better tasks, meaning the ROI of the task also went up.
As this started catching up in the industry, everyone comes back from break in January. It takes a second from after the holidays for people to get settled. But then both things started happening. More and more of the individual developers started using these new models with new agent capabilities, and then also more and more automation started happening. And if you think about us, we can spend the entirety of that development life cycle, so it's not just about the repo and getting the code. We also have issues, we also have pull requests, we also have actions for you to build that, and we also have security. And they all intersect and build upon each other, so they compound on each other. So if you get more commits, you're going to have probably more PRs. You get more PRs, you're going to have more action runs. If you get more action runs, more security scans that you need to do. So like everything is compounded. So we have published some of these numbers. I think we saw 17 million in March alone of PR agents. We could get into maybe other stats, but everything shut off like this.
You probably heard Jensen in the keynote as well reference it. We're all feeling this overall. So what we saw change is that it's more and more people are coming into the platform at an accelerated rate. More because there's a low floor now or a lower floor of entry, and we could talk about that later. So I think this is a very exciting thing from a human perspective. Because of that, the more people are creating commits, more people are creating PRs, more actions end up running. So all of our services ended up seeing a record acceleration.
And if you, I'll make up some numbers here, but if you think about, we were expecting 5% growth as an example, now you're seeing 3x growth. And that's amazing. And really proves that, hey, we're starting to see value now out of all of these agentic workloads, which is why we're here.