weekend ai reads for 2026-04-03

Instead of licensing real soul records from the 60s or 70s or hiring studio musicians, producers are using AI to generate fictional retro samples. Producer Young Guru, Jay-Z’s longtime sound engineer, estimates that “more than half” of sample-based hip-hop is now made this way.

AI capabilities keep increasing, and using it to draft, brainstorm or accelerate you will be increasingly useful. However, using AI should not make your productivity someone else’s burden. New tools require new manners.

Getting the people in charge to make explicit decisions and commit to them long enough to find out if they're right... that’s the hard part.

I’m grateful for what it gave. I’m honest about what it took. And I’m done performing either gratitude or grievance about it.

For years now, Silicon Valley has largely failed to produce something that most people want, or are even comfortable having in their lives; it has failed to make the case for AI to a public that mostly fears for their jobs, their energy bills, their children’s safety and future.

AI slop ensures that no one forgets this. No wonder OpenAI wants to pivot to focusing on enterprise AI, where no one has to look at the technology’s visual exports unless they are forced to by their boss.

And this is why I’ve come to see AI as a potential savior. Something that will break down how we build product into its constituent parts so we can build it back up again. Better.

Generative AI vegetarianism, simply put, is avoiding generative AI tools as much as you can in your day-to-day life. For me, that means:

• Turning off all of the optional AI settings I can find

• Not using any of the built-in AI features that I can’t turn off

• Not consuming or re-sharing articles, photos, music, or videos that other people have produced with generative AI.

• Choosing software products that don’t have AI features

I think I understand the sentiment behind it: You’re not a designer because you know all the Figma shortcuts. You’re not a perfect typewriter away from The Next Great American Novel. Mastery of a tool is not mastery of the subject matter.

But I also disagree. Good tools do make you a better designer.

To meet our new minimum bar, candidates will need to clearly show:

• AI embedded into their core work

• Repeatable systems, not one-off prompts

• Clear impact on quality, efficiency, or related outcomes

If someone isn’t meaningfully improving their work with the support of AI, they don’t meet the bar.

Here are a few concrete examples of what that bar looks like, broken down by department.

CIOs have long struggled to balance enterprise tech budgets—and big investments in AI are compounding the problem. Our research shows how they can reallocate expenditures to generate maximum growth.

As I noted in a later post, the answer depends on the college-major pair. A finance degree from Wharton and a psychology degree from a commuter college are different.

…

Some top universities will adapt well, while others will not, often for reasons that are hard to predict in advance: leadership, governance, institutional culture. Among less selective institutions, some will move toward value propositions AI does not threaten (adult education, community, credentialing in regulated fields), while others will simply disappear.

If I want my classroom to be the kind of place where students encounter sentences that then become part of their soul, if I want those bricks to become castles, I need to create space for that slow growth. Carrera has reminded me of what I once thought education could be, and he’s showing me, even in this age of AI, that I don’t have to surrender that belief. So I try slowing things down, making space for slow, sustained, deep engagement.

Open models like GLM-5 and MiniMax M2.7 now match closed frontier models on core agent tasks — file operations, tool use, and instruction following — at a fraction of the cost and latency. Here's what our evals show and how to start using them in Deep Agents.