Sam: Eleven billion dollars, Kai. And the plan to get the samples back involved launching a second rocket from the surface of Mars.
Kai: A rocket they hadn't even designed yet. I'm Kai.
Sam: And I'm Sam.
Kai: And this is Trending Repos for Tuesday, September 22nd, 2026!
Sam: We've got three stories for you today where big ambitions meet even bigger engineering realities.
Kai: We’re starting with NASA officially grounding its ambitious Mars Sample Return mission.
Sam: Then, we're asking: if an AI wrote it, is it even worth reading?
Kai: And we'll look at how AI coding assistants are getting so good, they’re breaking CI pipelines at companies like Linear.
Kai: So, let's talk about Mars. Or, rather, the samples we are not getting from Mars. Science.org is reporting that NASA's Sample Return mission is effectively dead.
Sam: Dead is an understatement. This thing was on life support. The independent review board came back with a price tag of up to $11 billion and a return date of 2040. For a handful of rocks.
Kai: The most important rocks in the solar system, Sam! But yeah, the complexity was staggering. Perseverance caches samples, a second lander shows up with a fetch rover, that launches them into Mars orbit, and then a third spacecraft has to catch them and fly home.
Sam: It sounds like a Rube Goldberg machine designed by a committee with an unlimited budget. Which, it turns out, it was.
Kai: Come on, this is rocket science! It's supposed to be hard! You have to admire the ambition.
Sam: I admire ambition that has a passing familiarity with physics and economics. They basically admitted they had 'unrealistic and unsubstantiated expectations'. That's a quote.
Kai: So what happens now? The samples just sit there?
Sam: For now. NASA is going back to the drawing board, looking for a 'cheaper, faster, and less risky' plan. Which is government-speak for 'this isn't happening in our lifetimes'.
Kai: This feels like a huge opening for private space companies, though. If someone can figure out a leaner way to do this, the scientific payoff is immense.
Sam: The lesson here for developers on Earth is pretty clear: even with the smartest people in the world, if your project's complexity spirals, it's going to fail. Scope creep can kill anything—even a mission to another planet.
Sam: Alright, let's talk about a different kind of authenticity. A post from Colin Breck's blog is getting a lot of attention, and the title says it all: 'I don't want to read what you didn't write'.
Kai: Ooh, a spicy take on AI-generated content. Okay, I get the sentiment, but isn't this a little... gatekeepy? AI is a tool. It helps me organize my thoughts, fix my grammar, get past writer's block. It helps me write.
Sam: Helps you write, or writes for you? Breck's point is that writing is thinking. When you outsource the writing, you're outsourcing the thinking. The result is a hollow echo of human thought, not the real thing.
Kai: But what if the end result is clearer and more helpful? I'd rather read a clear tutorial co-authored with an AI than a confusing one someone struggled to write on their own.
Sam: That's a false choice, Kai. Using it as a proofreader is one thing. Letting it generate whole paragraphs is another. You're not getting clarity, you're getting a statistical summary of the internet. It has no soul, no unique insight.
Kai: [sighs] Okay, so where do you draw the line? Is using a thesaurus outsourcing your vocabulary?
Sam: The line is intent. Are you using a tool to sharpen your own thoughts, or to have thoughts for you? The firehose of AI-generated SEO spam is making real, human-written content more valuable, not less.
Kai: I guess I see that. The signal-to-noise ratio is getting pretty bad out there.
Sam: Exactly. And the takeaway here is that your unique perspective, your struggles, your weird analogies—that's your value as a writer. Don't let a machine sand it down into a smooth, featureless blob.
Kai: Okay, but that same AI is creating new problems elsewhere. Let's talk about when it works a little too well. Linear just published a great deep-dive on how AI coding assistants made their own Continuous Integration, or CI, a huge bottleneck.
Sam: The snake eats its own tail. We built machines to help us write code, and now we need to build bigger machines to handle all the code the first machines are writing.
Kai: But it's a great problem! Their engineers were just becoming so much more productive. The article says they saw PRs per engineer per week just skyrocket. Their already beefy CI setup just couldn't keep up. Queues were getting longer and longer.
Sam: More PRs doesn't automatically mean more productivity. It can also mean more half-baked ideas, more AI-generated boilerplate, more code that needs reviewing and testing. They were drowning in their own 'productivity'.
Kai: Okay, but their solution is super cool. They built a custom orchestration layer on top of their cloud provider that can dynamically spin up thousands of ephemeral, parallel runners. They call it 'Now'. The result? Their average CI wait time went from over 15 minutes down to just 90 seconds.
Sam: So they spent a ton of engineering effort to build a complex distributed system just to be able to run tests on code we're not even sure is good. It feels like we're building a hyper-efficient highway system for a clown car parade.
Kai: [laughs] Maybe! But it's a fast clown car parade. And seriously, this is everyone's future. If your team is leaning into AI code generation, your CI/CD pipeline is the next thing that's going to break. Linear just gave you a peek at the roadmap.
Sam: So you have to be ready to scale your verification and validation right alongside your code generation. Otherwise, you're just making mistakes faster.
Kai: Alright, so NASA's eleven-billion-dollar Mars mission is on ice.
Sam: Meanwhile, human writing is now a premium product, and that same AI is forcing us to re-architect our dev infrastructure just to keep up. A good day for the humans, I think.
Kai: Before we go, Sam, speaking of space, did you see 'terminal-odyssey' on GitHub?
Sam: Don't tell me it's another ASCII space shooter.
Kai: [laughs] No, it's an iTerm2 theme that turns your terminal into the command bridge from 2001: A Space Odyssey. All red text and HAL 9000-style readouts. It's completely useless and I absolutely love it.
Sam: That's our show for today. Find us tomorrow for more Trending Repos.
Kai: See you then! And Sam, I promise I won't use an AI to write my part of the script.
This show is made with AI: the hosts’ voices are synthetic and the scripts are AI-assisted. Every story links to its original source.