The Great Software Quality Collapse  ## Metadata - Author: [[The Changelog: Software Development, Open Source]] - Full Title: The Great Software Quality Collapse - Category: #podcasts - URL: https://share.snipd.com/episode/0c1e85df-7fee-4532-974c-88860010f27b ## Document Notes ## Highlights - **Abstraction Tax Compounds Into Collapse** - Modern software stacks compound small overheads across many abstraction layers into large inefficiencies. - Denis Stetskov argues these compounded costs normalize catastrophic resource use and failures. Transcript: Jerod Santo Quote, we've normalized software catastrophes to the point where a calculator leaking 32 gigabytes of RAM barely makes the news. This isn't about AI. The quality crisis started years before ChatGPT existed. AI just weaponized existing incompetence. End quote. Dennis has been tracking software quality metrics for three years. In this post, he cites nine plus examples from across the industry where memory consumption has lost all meaning and system level failures have become routine. What exactly is to blame? Not one single factor, but Dennis says the abstraction tax compounds exponentially. Like this, quote, Modern software is built on towers of abstractions, each one making development easier while adding overhead. Today's chain is React, to Electron, to Chromium, to Docker, to Kubernetes, to VM, to Managed DB, to API gateways, and each layer adds only 20 to 30%, but compound a handful and you're At 2 to 6x overhead for the same behavior. That's how a calculator ends up leaking 32 gigabytes, not because someone wanted it to, but because nobody noticed the cumulative cost until users started complaining. End quote. Dennis says we need to ask ourselves some hard questions. When did we accept that a calculator leaking 32 gigabytes is normal? How many abstraction layers are actually necessary? And what happens when we can't buy our way out anymore? ([Time 0:00:54](https://share.snipd.com/snip/060e7785-f7d7-40a2-9a69-d0b9dad75ff0)) - **Question Your Abstraction Choices** - Reassess which abstraction layers are truly necessary and measure cumulative costs. - Denis Stetskov urges asking hard questions before we "can't buy our way out anymore." Transcript: Jerod Santo End quote. Dennis says we need to ask ourselves some hard questions. When did we accept that a calculator leaking 32 gigabytes is normal? How many abstraction layers are actually necessary? And what happens when we can't buy our way out anymore? Click through for his proposed path forward and the bottom line. ([Time 0:02:07](https://share.snipd.com/snip/588f5372-ee61-47a7-858c-850d40fc4850)) - **React Moves To A Foundation** - Meta is transferring React and React Native stewardship into a Linux Foundation-hosted React Foundation. - Meta commits five years, $3M+, and engineering support while retaining involvement for now. Transcript: Jerod Santo The new home for React and React Native. Meta is officially handing React and React Native, as well as supporting projects like JSX, over to a foundation. The React Foundation's mission is to help the React community and its members. The React Foundation will maintain React's infrastructure, organize React Conf, and create initiatives to support the React ecosystem. The React Foundation will be part of the Linux Foundation, which has long fostered a vendor-neutral environment for open-source projects. End quote. Meta isn't abandoning the projects, at least not yet. They've committed to a five-year partnership with the newly formed foundation, including $3-plus million in funding and dedicated engineering support. ([Time 0:02:27](https://share.snipd.com/snip/2ab854a4-6285-4d62-817f-8588434436e6)) - **GitHub Puts Azure Migration First** - GitHub is prioritizing migrating its infrastructure to Azure, delaying feature work. - Internal docs suggest migrations stalled and required pulling engineers off other projects. Transcript: Jerod Santo GitHub prioritizes Azure migration over features. Here's Frederick Lardenoise, apologize on the pronunciation, reporting for the new stack. Quote, with GitHub CEO Thomas Domke leaving the company this August and GitHub being folded more deeply into Microsoft's organizational structure, GitHub lost that independence. Now, according to internal GitHub documents, the new stack has seen the next step of this deeper integration into the Microsoft structure is moving all of GitHub's infrastructure To Azure, even at the cost of delaying work on new features. End quote. I'm not at all surprised by this, but I am certainly disappointed. My first thought, we can walk and chew gum at the same time. Why not do both? Quote, while GitHub had previously started work on migrating parts of its service to Azure, our understanding is that these migrations have been halting and sometimes failed. End quote. Yikes, this is terrible optics for Azure. Even a Microsoft-owned entity struggles to migrate to it and has to pull people off other features to make the transition even happen. I've said it a few times this year, I'll say it again, GitHub is primed for disruption. Where will that disruption come from? I'm not sure, but we'll know it when we see it. ([Time 0:03:11](https://share.snipd.com/snip/14beef6d-4085-4552-bb0e-3de7a921f39d)) - **Prefer Efficient Models For Dev Tasks** - Consider smaller, faster models for many dev tasks instead of defaulting to largest models. - CodeRabbit's head-to-head shows Claude Sonnet 4.5 matches larger-model performance at lower cost. Transcript: Jerod Santo Claude Sonnet 4.5 versus Opus 4.1. CodeRabbit just ran a head-to between Claude Sonnet 4.5 and Opus 4.1, and the results are fascinating. Sonnet 4.5 isn't just faster, it's cheaper, and in many real-world dev tasks, smarter. But here's the paradox. Even with stronger reasoning and better latency, most teams still default to bigger, slower models out of habit. The hidden gem? CodeRabbit's data shows Sonnet 4.5 matches opus-level performance on code reviews, debugging, and refactors at a fraction of the cost. It's a reminder that more tokens don't always mean more value. This opens the door for teams to run cheaper, faster, better AI-assisted reviews, and maybe rethink what top-tier really means. ([Time 0:04:28](https://share.snipd.com/snip/c6e37236-a5b4-4603-a699-b256bac297d2)) - **Python 3.14 And PyPy Performance** - Python 3.14 shows CPython performance gains and benefits from free-threading for multi-threaded apps. - Miguel Grinberg's benchmarks also highlight PyPy's extreme speed compared to CPythons. Transcript: Jerod Santo Python 3.14 is here. How fast is it? Python 3.14, with the free threaded now officially supported, was released on October 8th, 2025. So Miguel Grinberg put it through its paces. But first, a warning. Quote, running these benchmarks is fun, and that's why I do it, but it is really impossible to build an accurate performance profile of something as complex as the Python interpreter Just from running a couple of silly little scripts. Have a look at my benchmark, but consider it just one data point and not the last word on Python performance, end quote. Okay, with that out of the way, what did Miguel find? Clicked through for the individual benchmarks, but his conclusions were, 1. CPython 3.14 appears to be the fastest of all the CPythons. 2. Its JIT interpreter doesn't appear to provide any significant speed gains. 3. Its free threading interpreter is faster for CPU-heavy, multi-threaded apps. And 4. PyPy is insanely fast. That last bullet point was the main thing I noticed when scrolling through the results. While 3.14 beats previous versions, PyPy blows all CPythons out of the water while competing with Node and Rust alternatives. ([Time 0:05:27](https://share.snipd.com/snip/ef42c0ab-5845-4a05-a78a-1ab5548e9800)) - **Knowing Art Is AI Changes Perception** - Reactions to AI-generated art vary: discovery that art is AI-made often changes the viewer's feeling. - Matthew Inman and Jerod Santo note feelings shift from enjoyment to deflation upon learning a piece is AI art. Transcript: Jerod Santo Let's talk about AI art. The Oatmeal's Matthew Inman published a take on AI art that's making the rounds. Maybe you haven't seen it yet. Quote, when I consume art, it evokes a feeling. Good, bad, neutral, whatever. When I consume AI art, it also evokes a feeling. Good, bad, neutral, whatever. Until I find out that it's AI art. Then I feel deflated, grossed out, and maybe a little bit bored. End quote. I'm not so sure about grossed out, but deflated and bored both track with how I felt when realizing a piece of art is actually AI art. But on the other hand, I've also seen some pretty imaginative stuff and had a lot of fun bringing my own imaginations to life too. I've hemmed and I've hawed about AI art, but I'm starting to think the right approach is to adapt my stance on other AI tools, which is use AI to help you think, not to think for you. ([Time 0:06:40](https://share.snipd.com/snip/0901a8d7-b70d-41e1-943f-17687e7cfd60))