Here’s Leah Tharin with a good reminder of what it means to ship, and how AI can (and cannot) help. In short, building is only one part of creating valuable products. Shipping involves:

• Ideation: There’s an idea
• Development: You build the idea
• Validation: You validate whether what you think the idea does is actually happening

Yes, vibe coding tools like Lovable et al. help you to ship things faster, but only as long as these ideas struggle with the “development” part and don’t need Ideation and Validation.

Source: AI’s “Just Ship it.” problem →

This is some really interesting data.

In a new paper, several Stanford economists studied payroll data from the private company ADP, which covers millions of workers, through mid-2025. They found that young workers aged 22-25 in “highly AI-exposed” jobs, such as software developers and customer service agents, experienced a 13 percent decline in employment since the advent of ChatGPT. Notably, the economists found that older workers and less-exposed jobs, such as home health aides, saw steady or rising employment. “There’s a clear, evident change when you specifically look at young workers who are highly exposed to AI,” Stanford economist Erik Brynjolfsson, who wrote the paper with Bharat Chandar and Ruyu Chen, told the Wall Street Journal.

Source: The Evidence That AI Is Destroying Jobs For Young People Just Got Stronger →

I almost skipped this Michael Lopp piece, but I’m glad I didn’t. It’s one of his best in a long time, especially if you’re a manager. Reframing the act of working with AI as “making the robots dance” is so good. But there’s more to it than that. Just read it, ok?

Robots don’t lie. Lying requires intent to deceive, and when a robot provides you with plausible-sounding, but incorrect statements, it’s either following its programming or making an error. Or both. Humans lie. They boast, they are tragically optimistic, they exaggerate, they forget, I could go on for a long, long while. It’s a list of foibles that make them familiar… that makes them human. What do I do as a leader to work with these troublesome humans? Well, here’s a short, essential list:

• I speak clearly and specifically, so my intent is clear.
• I frame conversations with context so everyone understands my ideas.
• I understand errors are part of the process and work to build tools to prevent them.
• I debate and plan big ideas before I begin.

Source: Every Single Human. Like. Always. →

This is some fascinating research.

Researchers at Oxford University and King’s College London tested LLMs using game theory, giving LLMs from OpenAI, Google, and Anthropic prompts that mimicked the setup of the classic Prisoner’s Dilemma.

They found that Google’s Gemini is “strategically ruthless,” while OpenAI is collaborative to a “catastrophic” degree. Their paper, published on the preprint repository Arxiv (and not yet peer reviewed), claims that this is due to OpenAI model’s fatal disinterest in a key factor: how much time there is left to play the game.

Source: Gemini Is ‘Strict and Punitive’ While ChatGPT Is ‘Catastrophically’ Cooperative, Researchers Say →

Don’t let the clickbait title put you off. Related to my link about AI killing jobs in tech, here Cal Newport produces some compelling “both sides” receipts about how AI is helping + hurting software development. His conclusions are solid:

My advice, for the moment:

1. Tune out both the most heated and the most dismissive rhetoric.
2. Focus on tangible changes in areas that you care about that really do seem connected to AI—read widely and ask people you trust about what they’re seeing.
3. Beyond that, however, follow AI news with a large grain of salt. All of this is too new for anyone to really understand what they’re saying.

AI is important. But we don’t yet fully know why.

Source: No One Knows Anything About AI →

I think we’ve all seen the internal Shopify memo on requiring teams to use AI. This is a great article on what happened next. I especially love the internal tools Shopify built to make adoption easier:

Employees can use the LLM proxy to build the workflows they need. They can select from different models, which are updated with the latest versions as soon as they’re released. There’s a collection of MCPs, and all it takes is asking the proxy (or another tool like Cursor) to access them. There’s even a stable of agents already created by other people for anyone to use. It’s a one-stop shop for everything someone needs to use AI.

Source: From Memo to Movement: Shopify’s Cultural Adoption of AI →

There are a lot of these “how to beat the AI cookie monster” posts out there right now, but this one by Benjamin Todd is well-researched and articulated, with lots of practical examples on how to do the one thing that we all need to do anyway: keep learning.

I break this down into four key categories of skills likely to increase in value:

1. Hard for AI: data poor, messy, long-horizon tasks where a person-in-the-loop is wanted
2. Needed for deploying AI: the skills of organising and auditing AI systems, as well as those used in complementary industries such as data centre construction
3. Used to make things the world could use far more of: skills that contribute to improved healthcare, housing, research, luxury goods, etc. – things which people want more of as they get better and cheaper
4. Hard for others to learn: rare expertise that matches your unique strengths

Source: How not to lose your job to AI →

You know how those of us who read The Lord of the Rings before the movies came out got all weirdly and annoyingly upset about all the “new fans” and how they should have “read the books years ago”? That’s how I feel about the em dash and its AI takeover.

The real issue isn’t me—it’s you. You simply don’t read enough. If you did, you’d know I’ve been here for centuries. I’m in Austen. I’m in Baldwin. I’ve appeared in Pulitzer-winning prose, viral op-eds, and the final paragraphs of breakup emails that needed “a little more punch.” I am wielded by novelists, bloggers, essayists, and that one friend who types exclusively in lowercase but still demands emotional range.

Source: The Em Dash Responds to the AI Allegations →

This was a very comprehensive survey about everything from AI tools to Terminal app preferences, CI/CD systems, and more. Very much worth the click to skim through the results. Gergely also has an interesting theory on why developers hate Jira so much:

But I wonder if the root problem is really with JIRA itself, or whether any project management tool idolized by managers would encounter the same push back? It is rare to find a dev who loves creating and updating tickets, and writing documentation. Those who do tend to develop into PMs or TPMs (Technical Program Managers), and do more of “higher-level”, organizational work, and less of the coding. Perhaps this in turn makes them biased to something like JIRA?

Source: The Pragmatic Engineer 2025 Survey: What’s in your tech stack? →

Great list of resources here by Pete Steinberger:

These resources will help you master the new paradigm of AI-assisted development, where agents become true collaborators that can handle entire codebases and ship production features. Each piece was chosen for its practical, real-world insights.

I especially appreciate that it’s a combination of articles (yay!) and videos (not for me!), and that he provides a nice overview of each so you can decide if you want to click through or not. Excellent curation, would recommend!

Read Essential Reading for Agentic Engineers ↗