Stop measuring AI adoption. Start measuring AI impact. | LinearB’s APEX framework
How to measure AI in the critical path.
Are your AI coding tools actually making your team faster, or are they just creating downstream chaos? This week, Ben Lloyd Pearson and Dan Lines introduce APEX, LinearB’s new engineering leadership framework built explicitly to measure and manage software delivery in the AI era. Moving beyond traditional frameworks like DORA and SPACE, APEX balances AI Leverage, Predictability, Efficiency, and Developer experience to ensure upstream code generation translates into actual business value. Tune in to learn how to break past the illusion of coding speed, prevent AI slop from clogging your review pipelines, and discover which pillar of the APEX framework your team needs to tackle first.
1. What happens when a PR becomes a billboard?
GitHub found itself in hot water this week after Copilot inserted an advertisement directly into a user’s pull request. While GitHub reversed the behavior and claimed it was just a programming logic issue, it gave us a bizarre glimpse into a future where our tools might actively market to us while we work. It raises serious questions about authorship and trust. By the way, if you are not on an enterprise plan, you only have until April 24th to explicitly opt out of your private data being used to train Copilot. Go uncheck that box so your private code stays private.
Read: GitHub backs down, kills Copilot pull-request ads after backlash
Source Zach Manson
2. Surviving the whiplash of “dead” AI trends
Bryan Bischof, Head of AI at Theory Ventures, wrote a highly entertaining, satirical piece tracking how quickly the tech community declares AI technologies dead. From RAG to prompt engineering, things go from essential to obsolete and back again in the blink of an eye. It is a fun take on the whiplash we all feel trying to keep up with the changing fundamentals of our industry. Of course, the title itself is a clever play on the actual ripgrep utility, which is one piece of software that is certainly not dying anytime soon.
Read: Software’s Next Epoch: Our Investment in R.I.P. grep
3. Zero tests, 500k leaked lines, and virtual terminal pets
Anthropic accidentally exposed over half a million lines of Claude Code’s TypeScript source through a misconfigured build. The leak revealed unreleased features like a background daemon and coordinator mode, but the most shocking part was the architecture itself: zero tests and a massive three thousand line function. The absolute best discovery was “Buddy mode.” If you update your terminal and use the /buddy command, you can hatch your very own virtual ASCII pet. I got a legendary robot named Trixel that comments snark on my coding sessions now. Look out, Grok!
Read: Anthropic accidentally leaked Claude Code’s entire source. Here’s what 512,000 lines reveal
Watch: Kshetrajna Raghavan’s presentation at the Bay Area DSPy Meetup
4. Swapping giant models for highly tuned smaller ones
Shopify recently cut their inference costs 75x by switching a specific data extraction task from GPT-5 to a self-hosted Qwen 3 model. By using tools like DSPy (I ❤️ DSPy) to fine tune the smaller model and deploying a multi-agent architecture, they actually managed to double their output quality. It is a massive lesson in efficiency, proving that you do not always need the most expensive foundation model to get exceptional results.
Read: Shopify Cuts Inference Cost by 75x Using Qwen 3 for Merchant Data Extraction
5. Faster coding doesn’t mean faster delivery
AI tools might be accelerating your coding phase, but if your review queues are suddenly clogged then the bottleneck has merely shifted. To prevent speed from creating chaos, engineering leaders need a way to measure AI in the critical path.
That’s why LinearB published the APEX framework. It’s a complete operating model designed to connect AI activity to real delivery outcomes, ensuring that productivity gains don’t come at the cost of predictability or developer burnout. Stop guessing about AI ROI and start measuring what matters.
6. No more CSS in your browser?
Cheng Lou, an engineer at Midjourney, just dropped a fascinating new library called Pretext. It allows you to measure and render text in a browser using pure JavaScript, completely bypassing CSS and the traditional DOM thrash. Not only does this offer incredible rendering speeds, but it also allows agents to headlessly render and understand website layouts. Tucked away in the project’s documentation is a steering thought that doubles as a manifesto and perfectly captures this era: “The cost of any verifiable software will trend towards zero.”
Read: Pretext Does What CSS Can’t — Measuring Text Before the DOM Even Exists
7. Managing open source in the age of agentic pull requests
Steve Yegge is back with a look at what it takes to manage an open source project when you are handling 50 AI generated pull requests a day. Instead of outright rejecting AI contributions or demanding perfection from human submitters, he flips the script. He uses his own agents to bring the contributed code up to his strict architectural standards. It is a brilliant, consensus driven approach to development. Maintainers who refuse to adapt risk fracturing their communities, while those who embrace this model can steer their projects faster than ever before.
Read: Vibe Maintainer
