Welcome back. For years, GitHub was a no-brainer for code hosting. But Cursor and SpaceXAI are flipping the script. They just dropped a GitHub competitor named Origin, and in a twist, you couldn't script, GitHub faced a massive outage just hours later. Read details below.

Also: A guide to get started with a new GitHub alternative, Amazon VP explains how engineers get promoted, and a Google engineer’s guide to save tokens in Claude Code.

Tell us what you like and what you don't about this newsletter. Hit reply on this email and we'll respond to every single one.

Cursor takes on GitHub with agent-native code hosting: The AI startup under the leadership of Elon Musk just dropped Origin, a GitHub alternative for devs. It has every feature you expect from GitHub, like repos, PRs, and code browsing, but with AI agents living directly inside your repos. You can use Cursor to ask questions, ship changes, or push branches instantly. Plus, you can start syncing your GitHub repos right away to keep your existing workflow intact. In a stroke of pure irony, the launch coincided with a major GitHub outage, and devs on the internet had a field day with it. Use this step-by-step guide to get started with Origin.

Claude Code ships design command and lighter CPU load: Anthropic just shipped two powerful updates to its coding harness. You can now type /design in the CLI or desktop app and generate shareable art boards that match your team's existing code. This lets non-designers prototype ideas before handing them off to engineers. You can run it in the terminal or on the desktop app. The company also made a fix to their Claude Code CLI to ensure it uses about half the CPU compute than before.

OpenCode counters DeepSeek's price hike: The AI startup just launched "Operation Cheepseek" after the Chinese lab significantly hiked its API rates. The company claims that this price hike directly affects your agents that lean on long and repetitive prompts. The team saw some early success with their project, and OpenCode Go subscribers can get $30 worth of usage for just $10, while the team tests out new hosting options to bring costs back down. 

You spend hours on things that aren't code. PR descriptions. Slack threads explaining why you made that architecture call. Linear tickets with enough context so your teammate doesn't ping you at 11pm. Docs you keep pushing to next sprint.

Wispr Flow handles all of it. Speak naturally, and it outputs clean text anywhere you type. Syntax-aware, so your variable names and file paths stay intact.

It won't write your code. But it'll clear out everything blocking you from writing it. Works across Mac, Windows, iPhone, and Android.

Teams at Vercel, Clay, and Rivian already use it daily.

Free to start

Beyond review. Once agents write the majority of code, the number of people contributing goes up, and reviewing every diff gets hard. A Monaco engineer recently broke down how their team lets non-engineers ship production code without breaking everything.

Blur the roles. Writing code is no longer just for specialists. Previously, a backend engineer needing a frontend fix had to pull in another developer and wait. Now, they can often do it themselves. Designers can also implement polish directly instead of waiting for feedback rounds. Cutting out handoffs makes shipping faster. The goal is to have machines handle as much of the review process as possible before a human even looks at the PR. 

Here is how Monaco’s engineering team improved their code reviews: 

The broader shift. Engineering teams are moving toward a new reality where reading every single line of code just doesn't scale anymore. The goal is to let automation handle the grunt work of checking for errors. That way, engineers can focus their energy on high-level design and figuring out if a change even makes sense in the first place. 

PS: If your engineering team is looking to automate more of your review process, try this out.

GitHub Actions was fine at 10 engineers, but at a 50+ headcount, 20-minute build queues and infra outages are tanking your deployment velocity.

Top-tier AI labs like OpenAI, Anthropic, Mistral, and xAI solved this early by migrating their CI/CD pipelines to Buildkite.

When Cursor built Origin to challenge GitHub, they chose Buildkite for their CI infrastructure. While Origin manages the source, Buildkite handles the heavy lifting of the build process before shipping to Vercel. It's one unified path from commit to production.

If your CI/CD is your biggest bottleneck, it's time to upgrade to the high-performance infrastructure the frontier labs already trust. 

Start your 30-day all-access trial →

How to make your entire development cycle AI native: Learn how to build and deploy modern, full-stack web applications using Next.js, React, and AI APIs. This tutorial will help you create production-ready SaaS projects, mastering user authentication, database integration, and seamless cloud deployment strategies.

Amphetamine: You can now keep your Mac awake with the lid closed, so local agents, servers, builds, tests, and other long-running tasks can keep working uninterrupted.

Why Agent Plugins are the future of Agent Skills (by Google): This guide shows developers how to package Agent Skills and their MCP dependencies into standardized Agent Plugins, using a simple manifest to turn them into portable, reusable folders that work across compatible AI clients without rebuilding the setup each time.

Every new Codex chat starts blind, forcing you to re-paste context from earlier sessions. An OpenAI DX engineer shared that the Codex app now handles this natively.

Codex pulls the referenced chat's full context and can message the other chat on its own.

P.S. Get 50+ AI coding hacks for Claude Code, Cursor, and Codex here.

Check out how Anthropic watermarks Claude text but spares most code.

Grow customers & revenue: Join companies like Google, IBM, and Datadog. Showcase your product to our 300K+ engineers and 150K+ followers on socials. Get in touch.

You can also reply directly to this email if you have suggestions, feedback, or questions.

Until next time — The Code team