The Elements of Slop

On this week's episode, Kris and Matt are joined by Fred Hebert and Reid Savage from Honeycomb to talk discuss building products with tiny teams, why code reviews are a useful bottleneck that we should keep around, and how AI mandates make a mistake in risk management. Fred discusses how his team uses vibe-coding to create false work that gets a new product standing on its own. Then the panel discusses having forgiving environments, why production isn't great place to learn, ensuring that worst case scenarios aren't avoided, and why the best use of AI is for personal tooling, not the stuff your team uses. Do you hear that? Yes! That is the sound of supporter content! This week that includes Fred's admission that AI has made his life as an engineer worse, Reid's argument for why managers are definitely sticking around, why we're still going to be buying technology instead of just vibe coding it, and Fred and Reid's must-read book list. Not a supporter yet? Become one today by going to https://fallthrough.fm/subscribe, you'll get extra content in Fallthrough, bonus discussions in Continue, and higher quality audio-only editions! If you prefer to watch this episode, you can view it on YouTube: https://youtu.be/m52ef550Fpg. No episode of the aftershow this week. We'll have more aftershow episodes soon! In the meantime, catch up on previous episodes at https://break.show. Thanks for tuning in and happy listening!

Change is challenging but quite often the tools that we used to solve yesterday's problems are unfit for tomorrow's problems. In this episode, Kris and Matt are joined by Rain Paharia, one of Matt's colleagues at Oxide who is also the creator of cargo-nextest. The trio discuss version control, the most difficult problem of Rain's career, and the challenge of rigor in an age when it's becoming cheaper than ever. Rain brings their knowledge from a decade at Meta to our ongoing discussion of jj, git, and why version control systems are so much better at the big tech companies. The trio discusses their gripes with GitHub, including some fresh perspectives from Rain on GitHub's attempts to fix issues (including some awkwardness with how Stacked PRs actually work). Then the conversation shifts to discussing how the team at Oxide managed to create an incremental updating system that works on air-gapped racks with hundreds of components. Along the way the trio also discuss API versioning and property based testing. You might be wondering where Kris' voice is in this episode? Well, he can certainly be heard in this weeks' supporter content! This week the expanded edition includes deep conversations about why LLMs got better at code but worse at prose, why Rain doesn't allow LLMs to write comments, LLMs as cliche machines, and some discussion of our favorite topic, Go modules. Are you not a supporter yet? You should fix that! Head on over to https://fallthrough.fm/subscribe and sign up today. You'll get expanded content in Fallthrough, bonus discussions in Continue, and higher quality audio-only editions. If you prefer to watch this episode, you can view it on YouTube: https://youtu.be/C-xkusj-uvc. Thanks for tuning in and happy listening!

How should we spend our time now that writing code is cheap? In this episode, Kris and Matt are joined by Doug Soo, a staff engineer working on the storage systems at Honeycomb. They discuss his blog post Spend More Time Talking to Humans, how AI is changing software development, and how that affects the way we communicate. The trio go from hedging the claim that we trust the output of compilers so we should trust the output of agents, how correcting LLMs can lead to exhaustion with little benefit, the ways in which decision making has become the core job of software engineers, how change can be actively bad, and the difference between optimism and hope. For our wonderful supporters we have an expanded edition! In this week's expanded content we have advice on how to reconcile AI with your career, regardless of whether you're just getting started or if you've been doing this for 30 years; an MIT class that shaped the way Doug thinks; how to tell if you actually enjoy implementation and which implementation skills matter; and Kris' prediction that the era of agents is coming to an end. Not a supporter yet? You can fix that right now by going to https://fallthrough.fm/subscribe and signing up today. You'll get expanded editions of Fallthrough, bonus content in our other podcasts, and higher quality audio-only editions. Sign up today! If you prefer to watch this episode, you can view it on YouTube: https://youtu.be/kMo8aazKPCc. No episode of the aftershow this week. We'll have more aftershow episodes soon! In the meantime, catch up on previous episodes at https://break.show. Thanks for tuning in and happy listening!

Steve is back! He joins Kris and Matt to discuss the drama around DHH's latest keynote at Rails World and Steve's latest views on AI. Interestingly, Steve believes that he'll sit in opposition to how Kris and Matt see the current situation. We start with Steve giving us the long backstory of Rails and DHH. The trio then dives into a conversation about purity tests, whether you can learn from people you dislike, and why making Linux a viable third option for desktop environments is important. From there they dive into whether the agents that carried out hacks were actually told to do that, whether alignment is about intent or obedience, and whether it's better to be worried about Skynet now or later. Hilariously, during this conversation, Steve's AI personal assistant texts him during the recording, bringing up a deeper conversation about AI personal assistants and how much information we should actually give them. Supporter only content? But of course! This week the panel digs into whether AI agents are basically hammers, why agents make native apps economically viable, Kris' argument that the literacy crisis poses a much larger and more existential risk than AI, the ways in which the pronunciation trend on TikTok is weird classism, and what happens when Jev played the trolley problem. Not a supporter yet? You can fix that today by heading over to https://fallthrough.fm/subscribe where you can sign up to gain access to extra content in our podcasts and higher quality audio-only editions. If you prefer to watch this episode, you can view it on YouTube: https://youtu.be/0Rdd1RZx-t4. Thanks for tuning in and happy listening!

Another week, another duo episode, but this time around it's Dylan who's joining Kris! He brings a fresh perspective to the AI conversation that Kris and Matt have been engaging in. The big question: is AI going to kill us all? Dylan thinks the odds are "astronomically small" and neither Kris nor him are going to lose sleep over the fear driven hype. The two are asking bigger questions: whether what language models are doing is thinking, why those who build these models in AI labs aren't actually the people who understand them the best, and whether TypeSafe AI's new Jev model is an LLM and what one can do with it. They touch on token maxxing, the increasing flood of AI slop, whether we can truly look at the code we ship, and as always why Kris is unbothered by nearly all of this. We've got supporter content, of course! In this week's content we include a discussion of what one can actually build with TypeSafe AI's Jev, why good and evil aren't useful categories, a candid conversation about policing from two somewhat opposing vantage points, why filling potholes isn't an ideology, and more. Not a supporter yet? Fix that today by heading over to https://fallthrough.fm/subscribe where you'll get not only extra content but also higher quality audio. Sign up today! If you prefer to watch this episode, you can view it on YouTube: https://youtu.be/xCtImPOmEzg. No episode of the aftershow this week. We'll have more aftershow episodes soon! In the meantime, catch up on previous episodes at https://break.show. Thanks for tuning in and happy listening!