John Coghlan

For community, DevRel, and tech content follow me on LinkedIn: linkedin.com/in/johnwcoghlan

I’d like to announce the public beta launch of GitLab Orbit - the Context Graph for the SDLC. Previously known as the GitLab Knowledge Graph, GitLab Orbit aims to make all of your AI agents better, faster, and cheaper. More in the 🧵

If you’ve ever had to modernize a system where “just don’t touch it” was the safest strategy ... At @WeAreDevs in July, I will share my honest take: where AI agents can actually help with legacy modernization, and where human judgment still matters. ⬇️

The closest thing to magic you can find right now: lower agentic cost, improve outcomes, speed up reasoning - GitLab Orbit.

@silverfighter @gitlab You were an awesome guest. Thanks for joining!

Great day in London at @gitlab Transcend. Was fun joining the @gitlab Developer Show as a guest speaker today.

I expected Duo Agent Platform to beat Cursor, Claude, Copilot and Devin on price with our $0.25/review fixed price, I did not expect us to win all of the above on precision and recall, but check out this site to learn more, including how we rank against all the code review solutions in market. duo-review-bench-6f7260.gitlab.io This is the power of repo-side agents, and it’s just the beginning of more powerful, higher quality and lower cost agentic engineering. This is our structural advantage in action.

Anthropic models have been the Duo default since inception and we’re happy to share Opus 4.8 is now available, same day release. about.gitlab.com/blog/claude-op…

AI coding tools are great at writing code, but shipping it securely is a different problem. This tutorial shows how Claude Code + GitLab Duo Agent Platform closes the gap, from bug fixes to production, with CI/CD, security scanning, and code review built in. 1️⃣ Fix a bug with Claude Code and let GitLab CI/CD, security scanning, and GitLab Duo Code Review do the rest. 2️⃣ Add GitLab MCP context so Claude works from the context saved within the GitLab issue, and not just local files. 3️⃣ Use a Claude-powered external agent in GitLab Duo Agent Platform to address review feedback directly in the MR. See how they work together in our blog: about.gitlab.com/blog/claude-co…

New from @dnsmichi: learn how to leverage Claude Code and GitLab to ship things faster. At the end of the day, it's about delivering value to customers and this tutorial shows you a few different ways to do it. Enjoy. about.gitlab.com/blog/claude-co…

We’ve heard from teams who want to understand the path for moving from GitHub to GitLab, so my team put together a simple migration guide. It’s an early version, and we’ll continue polishing it, but hopefully it gives folks more clarity on the process and what to expect. Feedback welcome: migrate-to-gitlab-e6e296.gitlab.io

@mitchellh Would love to discuss how GitLab can be part of Ghostty’s future.

Ghostty is leaving GitHub. I'm GitHub user 1299, joined Feb 2008. I've visited GitHub almost every single day for over 18 years. It's never been a question for me where I'd put my projects: always GitHub. I'm super sad to say this, but its time to go. mitchellh.com/writing/ghostt…

@RyanRodemoyer2 @bstaples Got it. If your instance is self hosted, you'll need to be on version 18.1 or later and enable the feature flag. Details on how to turn on the flag are on the docs page I linked to in my last message.

@john_cogs @bstaples Whatever instance we have at work .. Rapid Diffs is not available or it's disabled - not sure.

Tired of the pain yet? Come to GitLab and take back control of your destiny. I’ll even throw in the first year free for anyone switching from GitHub who signs a new three year agreement. DM me

@RyanRodemoyer2 @bstaples Have you tried Rapid Diffs, Ryan? In beta and may be of interest. - Docs: #rapid-diffs" target="_blank" rel="nofollow noopener">docs.gitlab.com/user/project/m… - Issue: gitlab.com/gitlab-org/git…

@bstaples Can the team put some love into the merge request review screens? Code diffs are impossible to read. The integrated code coverage is cool tho.

@rawkode @bstaples OSI-licensed projects can get Ultimate via the GitLab for Open Source program: about.gitlab.com/solutions/open…

@bstaples Bring back Ultimate for OSS

Devs teams: I noticed other providers are hitting walls, @gitlab is happy to serve you. Start free, receive $24/month in free promo AI credits with every Ultimate user, enjoy CLI/IDE/UX access to built in full SDLC agents, Claude and Codex, and your own custom agents, and use your choice of models and model providers so you never get stuck without service.

The truth about enterprise AI: The best AI experiences require context. And the richest context lives inside the systems where teams plan, write, build, ship, and support software. If your vendors are using your data to train shared AI systems, your data could be making the tools your competitors use smarter too. AI governance is no longer just a security concern. It’s a competitive strategy.

Big day on the announcements front. For those following at home, here's what GitLab shipped today. GitLab 18.11 dropped today. This marks 174 consecutive months with a new release. The highlights are: • Agentic SAST Vulnerability Resolution is now GA • Budget guardrails for GitLab Credits are now available • Two new foundational agents in DAP: CI Expert and Data Analyst We continue to build out AI capabilities for the entire SDLC with the governance our customers need and expect. Agentic SAST Vulnerability Resolution reaching GA is yuge. Because if AI is helping teams create more code, then the pressure on security and remediation only goes up. Enabling teams to move faster from finding to remediation is where things start to get operationally useful. I also think the budget guardrails matter more than they may seem at first glance. They’re not flashy but EVERYONE wants this. For AI adoption to expand inside engineering organizations, you need predictability and control. And now you have it with DAP. In addition to the release, we also added Claude Opus 4.7 to DAP. Part of the growing list of models we support. And of course, you can bring your own if you prefer. Oh, and one final note for the GitLab heads: release posts can now be found in our docs.

Yes, when subsidized tools push toward profitability, costs will rise. That's why context is more critical than ever. Agents with better context (pipeline logs, repo histories, security scans) waste fewer tokens. They get to the right answer faster with less exploration and fewer retries. As you build your AI stack, think carefully about where and how you'll use context in your workflow. That's what will keep you delivering results even as the price of inference goes up.

I feel this but I know I can be guilty of this too. I do sense improvement in my use of AI over time. When I'm using AI for work will be shared with my team, a few steps I've taken to make things better for everyone: 1. be more critical of the output, the first pass from AI (like a human's first draft) is usually not very good 2. ensure that it represents my thinking absolutely, a close enough approximation is actually detrimental to everyone involved 3. put in the effort to make the output as succinct as possible. this saves time and improves retention. It's also okay to ignore the AI's suggestions when you're happy with something. You don't have to do everything it says...yet.

Thinking back to the early days of AI-assisted coding. Latency and suggestion acceptance rates were considered key metrics at the time. Amazing how quickly things have progressed in just a few short years. What metrics are you tracking now to measure the impact of agents inside your team? How long will they last?