Skip to content
Sign in
Artwork for Fallthrough
Technology

Fallthrough

Fallthrough Media

A deep and nuanced conversational podcast focused on technology, software, and computing.

Play
Open in
  • 23 episodes
  • weekly
  • Avg 1 hr 6 min
  • English

Episodes 23

NewestOldestLongest
All20Unplayed20In progress0Played0
Counted on this page — what you have heard stays on this device, so it is not something the list can be paged by.
  • #86
    Yesterday · 57 min

    Be Kind, Rewind

    Matthew takes the host chair this week with Kris as an unusually quiet producer. They're joined by Ron Efroni and James Bayer from Flox for a long conversation about Nix. Ron co-founded Flox and is also the president of the NixOS Foundation. James gives a live demo, going from flox search to flox activate with no commit SHA anywhere in sight, and Matthew finally gets an answer to a flake.lock code review problem: what exactly am I approving when the diff is one hash for another? Ron makes the cancel worthy claim that Nix is not hard, it just lacks opinionation by design, which turns out to be the entire reason Flox exists. Then the conversation turns to agents, why skills are software, and the company that uses Flox to turn back time on a model and replay the exact environment a bad answer came out of. We've got supporter content, of course! This week that includes Kelsey Hightower's hand in the containerize project, Ron on where the open source ends and the invoice begins, the customer who compared Flox to really good Wi-Fi, what actually keeps Ron up at night, Kris talking about the truly absurd amount of hardware humming away in his apartment, and a closing detour through AOL discs, Blockbuster late fees, and dedicated tape rewinders. If you prefer to watch this episode, you can view it on YouTube: https://youtu.be/eWjJVaDKRWU 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! Table of Contents: Chapter 1: Matthew Takes the Host Chair (00:00:58) Chapter 2: What Is Flox, and Who Is Actually Using It (00:02:34) Chapter 3: Do You Still Need Nix? Escape Hatches and SHA Pinning (00:09:43) Chapter 4: Live Demo: The Flox CLI and the Catalog (00:12:00) Chapter 5: What Goes In Your Repo, and the Nix Store (00:19:52) Chapter 6: Why Not Just Use Nix? Scale, Enterprise, Multiplayer (00:23:02) Chapter 7: Nix Is Not Hard, It Just Has No Opinions (00:24:59) Chapter 9: Packages, Environments, and Building the Artifact (00:30:59) Chapter 11: Does Local Dev Even Matter? (00:36:15) Chapter 12: Skills Are Software: Agent Stacks and Sovereign AI (00:39:03) Chapter 13: Turning Back Time on a Model, and the CVE Curve (00:45:45) Chapter 16: Don't Miss the Shot, and Why Nix Pilled (00:50:04) Epilogue (00:56:15) Host: Matthew Sanabria Co-Host: Kris Brandow Guests: Ron Efroni and James Bayer Producers: Matthew Sanabria and Kris Brandow Beats: Breakmaster Cylinder Socials: Website Bluesky Threads X/Twitter LinkedIn Instagram TikTok Changelog Zulip Gophers Slack (00:00) - Prologue (00:58) - Chapter 1: Matthew Takes the Host Chair (02:34) - Chapter 2: What Is Flox, and Who Is Actually Using It (09:43) - Chapter 3: Do You Still Need Nix? Escape Hatches and SHA Pinning (12:00) - Chapter 4: Live Demo: The Flox CLI and the Catalog (19:52) - Chapter 5: What Goes In Your Repo, and the Nix Store (23:02) - Chapter 6: Why Not Just Use Nix? Scale, Enterprise, Multiplayer (24:59) - Chapter 7: Nix Is Not Hard, It Just Has No Opinions (30:59) - Chapter 9: Packages, Environments, and Building the Artifact (36:15) - Chapter 11: Does Local Dev Even Matter? (39:04) - Chapter 12: Skills Are Software: Agent Stacks and Sovereign AI (45:45) - Chapter 13: Turning Back Time on a Model, and the CVE Curve (50:04) - Chapter 16: Don't Miss the Shot, and Why Nix Pilled (56:15) - Epilogue

    • Chapters
  • #85
    August 22 · 1 hr 58 min

    A Modest Proposal Proposal

    The proposal proposal gets the full Fallthrough treatment. Kris, Matthew, and Jamie dig into Austin's plan for taming the Go project's ever-growing proposal backlog. What starts as a recap becomes a referendum on how Go is governed, who gets heard, and what an open source community owes its members. Jamie brings a maintainer's reality check from Renovate's 1,700-item triage queue, Matthew wants to close the whole backlog and let people prove they care, and Kris argues the real product of a proposal system is discourse, not decisions. Along the way: whether Go is a democracy or a Google project, the joy economy that open source actually runs on, where "yes is forever" went when the UUID package shipped without consensus, and the gloves finally coming off over performative civility in Go's discussion spaces. We've got supporter content, of course! This week that includes Steve getting ratioed by Gen X on TikTok, T9 texting supremacy, why Go should never become a nation state, the IETF as proof that communities of joy do ship, and the biggest lie in open source. 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/CJkw4dqY38k 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! Table of Contents: Prologue (00:00:00) Chapter 1: Ratioed on TikTok: Gen X vs the Millennials [Extended] (00:06:29) Chapter 2: The Proposal Proposal: Recap and First Reactions (00:06:32) Chapter 3: Nobody Wants to Run the Process (00:16:38) Chapter 4: Is Go Even a Democracy? (00:22:01) Chapter 5: Just Close the Backlog (00:28:18) Chapter 6: Let People Talk: Discourse Is the Product (00:38:20) Chapter 7: Discussion vs Commitment (00:47:53) Chapter 8: Missing Lexicographers and Shepherds (00:53:26) Chapter 9: The Joy Economy: Who Gets to Say No? (01:00:46) Chapter 11: Yes Is Forever: UUID, Accountability, and 5,000 Words (01:15:29) Chapter 12: What Could They Actually Do? (01:25:54) Chapter 13: Losing Faith: Modules, Blame, and Bright Spots (01:32:07) Chapter 14: Performative Civility: Speak Better, Never Listen Better (01:43:53) Epilogue (01:54:52) Hosts Matthew Sanabria - Host Kris Brandow - Host Jamie Tanna - Host Socials: Website Bluesky Threads X/Twitter LinkedIn Instagram TikTok Changelog Zulip Gophers Slack (00:00) - Start (00:00) - Prologue (06:32) - Chapter 2: The Proposal Proposal: Recap and First Reactions (16:38) - Chapter 3: Nobody Wants to Run the Process (22:01) - Chapter 4: Is Go Even a Democracy? (28:18) - Chapter 5: Just Close the Backlog (38:20) - Chapter 6: Let People Talk: Discourse Is the Product (47:53) - Chapter 7: Discussion vs Commitment (53:26) - Chapter 8: Missing Lexicographers and Shepherds (01:00:46) - Chapter 9: The Joy Economy: Who Gets to Say No? (01:15:29) - Chapter 11: Yes Is Forever: UUID, Accountability, and 5,000 Words (01:25:54) - Chapter 12: What Could They Actually Do? (01:32:07) - Chapter 13: Losing Faith: Modules, Blame, and Bright Spots (01:43:53) - Chapter 14: Performative Civility: Speak Better, Never Listen Better (01:54:52) - Epilogue

    • Chapters
  • #84
    August 15 · 1 hr 4 min

    Backlog to the Future

    Kris is back from GopherCon, Matt is back from the lawn, and Fallthrough is back with news: a brand new website and a whole new show, Continue: Go Proposals Edition, a weekly look at Go proposals. The topic for this week's Fallthrough? It's a listener-requested deep dive into what the two of them have actually been building with AI agents. In the regular edition, Kris runs a frontier-model pricing bake-off and comes away fascinated by how dirt cheap Luna is, Matt hands eight well-scoped Linear issues to overnight orbs and wakes up to three merged PRs, and the two land on a shared principle: even when the inference happens in a data center, the harness should run on hardware you own. It all wraps with Matt prototyping an integration by voice on a walk in the park. Expanded edition? We've got one of those! This week it includes Matt's deep dive on AMP betting everything on orbs, a breakdown of the proposal to change Go's proposal process, the one line that gives the whole document away, Kris's redesign built on gates, shepherds, and commitments, a hard conversation about civility and power in Go discourse, and the case for a decentralized Go. 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/U4ztdSztCy0 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! Table of Contents: Prologue (00:00:00) Chapter 1: New Website, New Show: Continue (00:02:28) Chapter 2: What Kris Built: PHP, Rhetoric Analysis, and Dirt Cheap Luna (00:12:55) Chapter 3: Matt's Agent Assembly Line: Linear Issues to Merged PRs (00:24:20) Chapter 4: Harness Hopping: Codex, Claude Code, and Agent Teams (00:33:35) Chapter 6: The Economics Flipped: Intelligence Too Cheap to Meter (00:36:39) Chapter 7: Keeping the Harness on Your Own Hardware: A Dollar of Orbs (00:43:35) Chapter 8: Vibe Coding, Skill Atrophy, and Prototyping on a Walk (00:52:42) Chapter 9: The Proposal Proposal [Extended] (01:02:14) Epilogue (01:03:16) Hosts Matthew Sanabria - Host Kris Brandow - Host Socials: Website Bluesky Threads X/Twitter LinkedIn Instagram TikTok Changelog Zulip Gophers Slack (00:00) - Prologue (02:28) - Chapter 1: New Website, New Show: Continue (12:55) - Chapter 2: What Kris Built: PHP, Rhetoric Analysis, and Dirt Cheap Luna (24:20) - Chapter 3: Matt's Agent Assembly Line: Linear Issues to Merged PRs (33:35) - Chapter 4: Harness Hopping: Codex, Claude Code, and Agent Teams (36:39) - Chapter 6: The Economics Flipped: Intelligence Too Cheap to Meter (43:35) - Chapter 7: Keeping the Harness on Your Own Hardware: A Dollar of Orbs (52:42) - Chapter 8: Vibe Coding, Skill Atrophy, and Prototyping on a Walk (01:02:14) - Chapter 9: The Proposal Proposal [Extended] (01:03:16) - Epilogue

    • Chapters
  • #83
    August 8 · 1 hr 9 min

    Vibe and Prejudice

    We've got the classic trio of Kris, Matt, and Steve back to discuss Steve's impending cancellation for putting himself in the middle of some drama! Steve gives us a recount of the current saga between Zig, Rust, Bun, Fil-C, and his own programming language, Rue. The group discuss the problems with the drama, especially since it overshadows an amazing technical feat: a million lines of code ported to a new language in a week, then shipped to real customers, with a product that pulls in billions of dollars. Steve also made the mistake of complimenting Zig, which caused a 163 comment dog pile onto him. We've got supporter content, of course! This week that includes what it costs a community when the BDFL of a language says your code sucks, why the fear that LLMs can appear to think says more about us than about them, Steve on why you cannot ban a bunch of floats in a text file, Fil-C's promise of memory safe C and the trade-offs its author will not name, the pull request titled make Zig actually memory safe (unlike Rust), Odin not being notable enough, and Steve drops some exclusive new thoughts: agentic development is actually RTS macro versus micro. 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/xjDWZPBfvVI 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! Table of Contents: Prologue (00:00:00) Chapter 1: Bun Leaves Zig for Rust, in Public, by Accident (00:02:47) Chapter 2: Memory Problems, the Unsafe Burndown, and Andrew's Post (00:10:22) Chapter 4: A Million Lines in a Week (00:14:17) Chapter 8: Rue: The Compiler Steve Did Not Write (00:17:35) Chapter 9: 163 Replies for Complimenting Zig on Hacker News (00:22:50) Chapter 11: One Hundred Thousand Lines I Have Not Read (00:29:25) Chapter 12: Compilers Wrote More Assembly Than We Ever Did (00:33:49) Chapter 13: Vegan Plus Bacon, and the Cure for Binary Thinking (00:35:58) Chapter 14: Never Arguing About Tabs Versus Spaces Again (00:40:41) Chapter 15: Write the Documentation, Not the Code Review (00:44:37) Chapter 17: Effort Is What Makes It Not Slop (00:51:05) Chapter 18: Still the Chaotic Beginning (00:57:41) Chapter 19: Critical Thinking Is Curiosity, Not Skepticism (01:01:06) Epilogue (01:08:21) Hosts Matthew Sanabria - Host Kris Brandow - Host Steve Klabnik - Host Socials: Website Bluesky Threads X/Twitter LinkedIn Instagram TikTok Changelog Zulip Gophers Slack (00:00) - Start (00:00) - Prologue (02:47) - Chapter 1: Bun Leaves Zig for Rust, in Public, by Accident (10:22) - Chapter 2: Memory Problems, the Unsafe Burndown, and Andrew's Post (14:17) - Chapter 4: A Million Lines in a Week (17:35) - Chapter 8: Rue: The Compiler Steve Did Not Write (22:50) - Chapter 9: 163 Replies for Complimenting Zig on Hacker News (29:25) - Chapter 11: One Hundred Thousand Lines I Have Not Read (33:49) - Chapter 12: Compilers Wrote More Assembly Than We Ever Did (35:58) - Chapter 13: Vegan Plus Bacon, and the Cure for Binary Thinking (40:41) - Chapter 14: Never Arguing About Tabs Versus Spaces Again (44:37) - Chapter 15: Write the Documentation, Not the Code Review (51:05) - Chapter 17: Effort Is What Makes It Not Slop (57:41) - Chapter 18: Still the Chaotic Beginning (01:01:06) - Chapter 19: Critical Thinking Is Curiosity, Not Skepticism (01:08:21) - Epilogue

    • Chapters
  • #82
    August 1 · 1 hr 27 min

    Token Maxx: Fury Road

    On this week, we've got Kris, Matt, and Dylan! Yes, Dylan is back to talk about GopherCon and his experience as a self declared late adopter of agentic coding. Dylan chairs GopherCon, speaks at it next week, and describes his own AI hockey stick as happening in about two months. Dylan walks through what is on the GopherCon program, then explains his coding agent tipping point: a dependency injection refactor he had written off that Claude finished in an afternoon. But he's not all in yet: he still won't let one robot review and approve another robot's code. From there the conversation turns to the machine underneath: the model as a CPU and the harness as an operating system. The trio then arrives at money and the uncomfortable reality of how expensive tokens are, and whether that matters at all. They also answer an important question: should enterprises be subsidizing the cheap subscription plans individuals enjoy? We've got an Expanded edition for this episode which means we've got extra content! That includes what really happens to your conference proposal when 200 submissions compete for 25 slots, the mystery process that had Kris's Mac humming, the 200 gigabytes of Go module cache hiding on Dylan's drive, agents that shell out to Ruby and Perl when you asked for neither, caveman prompting, and the math on what a $250,000 engineer actually costs per month. 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. 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! Table of Contents: Prologue (00:00:00) Chapter 1: GopherCon Next Week: Community Day, Filippo, and slog (00:03:25) Chapter 3: Foot Dragger: The Refactor Claude Finished in an Afternoon (00:13:05) Chapter 4: Fashionably Late: CoreWeave, and Dylan's AI Hockey Stick (00:22:03) Chapter 5: Token Hunger and Overzealous Junior Engineers (00:30:25) Chapter 6: The Harness Is the Operating System (00:36:07) Chapter 7: Two Skills Worth Writing: Dependabot Chains and Chaos Forensics (00:42:12) Chapter 9: Context Is Everything: Bootloaders, Distros, and Second Opinions (00:50:04) Chapter 11: Thinking Tokens, and Why the Model Is the Product (00:58:22) Chapter 12: Who Pays for the Tokens: Personal Plans vs Enterprise Tiers (01:00:45) Chapter 14: If Everybody Does It: Token Maxx and the TOS Problem (01:08:54) Chapter 15: Talking Matt Into GopherCon, Live (01:17:39) Epilogue (01:23:09) Hosts Dylan Bourque - Host Matthew Sanabria - Host Kris Brandow - Host Socials: Website Bluesky Threads X/Twitter LinkedIn Instagram TikTok Changelog Zulip Gophers Slack (00:00) - Prologue (03:25) - Chapter 1: GopherCon Next Week: Community Day, Filippo, and slog (13:05) - Chapter 3: Foot Dragger: The Refactor Claude Finished in an Afternoon (22:03) - Chapter 4: Fashionably Late: CoreWeave, and Dylan's AI Hockey Stick (30:25) - Chapter 5: Token Hunger and Overzealous Junior Engineers (36:07) - Chapter 6: The Harness Is the Operating System (42:12) - Chapter 7: Two Skills Worth Writing: Dependabot Chains and Chaos Forensics (50:04) - Chapter 9: Context Is Everything: Bootloaders, Distros, and Second Opinions (58:22) - Chapter 11: Thinking Tokens, and Why the Model Is the Product (01:00:45) - Chapter 12: Who Pays for the Tokens: Personal Plans vs Enterprise Tiers (01:08:54) - Chapter 14: If Everybody Does It: Token Maxx and the TOS Problem (01:17:39) - Chapter 15: Talking Matt Into GopherCon, Live (01:23:09) - Epilogue

    • Chapters
  • #81
    July 25 · 52 min

    I, Builder

    Kris and Matt have a pretty big question they're trying to answer this week: in the age of AI, what does it actually mean to build something? We start (for real this time!) with the "SDKs are dead" discourse, and whether you still need one when you can just wrap the HTTP calls yourself. Kris then jumps the conversation to Beej's article on what it means to "make" or "build" a thing, he argues that we already lived through this during the analog-to-digital shift in music, photography, and video. Matt adds his own perspective from when he remodeled his kitchen. The duo doesn't stop with just that. In this week's Expanded edition they discuss the "well, actually" brigade and when technical precision helps versus hurts, the "I versus we" of blame, the identities we get sold and comfort chambers we live in, how character limits and engagement incentives turned social media into rage bait, and Matt's take on why "a persistent machine I can run agents on" is just a computer. 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/LhtXfZHRyO0. 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! Notes: On Making Table of Contents: Prologue (00:00:00) Chapter 1: SDKs in the Age of AI, Do We Still Need Them? (00:01:43) Chapter 3: What Does It Mean to Build Something? (00:22:12) Chapter 4: From Analog to Digital, AI as Just Another Tool (00:28:37) Chapter 5: Who Actually Built It? Ownership and Attribution (00:36:06) Epilogue (00:51:14) Hosts Matthew Sanabria - Host Kris Brandow - Host Socials: Website Bluesky Threads X/Twitter LinkedIn Instagram TikTok Changelog Zulip Gophers Slack (00:00) - Prologue (01:43) - Chapter 1: SDKs in the Age of AI, Do We Still Need Them? (22:12) - Chapter 3: What Does It Mean to Build Something? (28:37) - Chapter 4: From Analog to Digital, AI as Just Another Tool (36:06) - Chapter 5: Who Actually Built It? Ownership and Attribution (51:14) - Epilogue

    • Chapters
  • #80
    July 18 · 1 hr 11 min

    Everything Agentic All at Once

    Another week, another deep episode! Kris and Matthew are back and this one goes right off the rails in the best way. What was supposed to be a rundown of the new model landscape turns into a wide-ranging conversation about what AI coding agents actually mean for the people who write software. Matt brings a real dilemma: he had an LLM generate a working program and could not decide whether he still needed to read the code. That kicks off Kris's case for separating validation (did you build the right thing?) from verification (did you build the thing right?), an assertion that we already trust our compilers without reading the machine code, and why prototyping something and throwing it away can be a feature rather than a failure. Matt eventually closes the loop with a story about rewriting an LLM's code from scratch, converging on the exact same design, and finally understanding it. This conversation just kept going and going, and you can listen to the whole thing in the Expanded edition! This week that includes why the US dollar is already a digital currency, a frank conversation about America at its best and worst, why every confident prediction about the future turns out wrong, the mental health toll of clicking approve at two in the morning, and the software identity crisis quietly plaguing many of us. 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. 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! Table of Contents: Prologue (00:00:00) Chapter 1: Catching Up and the AI News Cycle (00:01:57) Chapter 2: Do I Need to Read the Code? (00:08:00) Chapter 3: Trusting the Compiler and the Illusion of Reading Every Line (00:16:05) Chapter 4: Building It Twice: Prototyping and Throwing It Out (00:26:28) Chapter 5: Linus Says AI Is Fine; Cloud Code Is Melting My Brain (00:29:20) Chapter 6: Nobody Understands Computers; Architecture Is the Real Work (00:32:09) Chapter 7: Oxide: Architecting an Ephemeral Runner Pool, and Learning by Osmosis (00:39:35) Chapter 14: Space for Everyone, Matt's V2 Rewrite, and Comfort in Discomfort (00:49:54) Chapter 15: The Volatile Era: Adoption Phases, Subscriptions, and Token Christmas (00:59:06) Epilogue (01:09:06) Hosts Matthew Sanabria - Host Kris Brandow - Host Socials: Website Bluesky Threads X/Twitter LinkedIn Instagram TikTok Changelog Zulip Gophers Slack (00:00) - Prologue (01:57) - Chapter 1: Catching Up and the AI News Cycle (08:00) - Chapter 2: Do I Need to Read the Code? (16:05) - Chapter 3: Trusting the Compiler and the Illusion of Reading Every Line (26:28) - Chapter 4: Building It Twice: Prototyping and Throwing It Out (29:20) - Chapter 5: Linus Says AI Is Fine; Cloud Code Is Melting My Brain (32:09) - Chapter 6: Nobody Understands Computers; Architecture Is the Real Work (39:35) - Chapter 7: Oxide: Architecting an Ephemeral Runner Pool, and Learning by Osmosis (49:54) - Chapter 14: Space for Everyone, Matt's V2 Rewrite, and Comfort in Discomfort (59:06) - Chapter 15: The Volatile Era: Adoption Phases, Subscriptions, and Token Christmas (01:09:06) - Epilogue

    • Chapters
  • #79
    July 11 · 59 min

    Landlord of the Rings

    Kris and Matthew are back for a duo episode, and Kris has a hot take: ownership is terrible. What starts as a conversation about AI subscriptions versus API pricing turns into a wide-ranging discussion about what we actually gain by owning things, from Photoshop to your own compute to your house. Along the way, Kris and Matthew dig into why enterprises are drifting toward per-token pricing, why AMP's headless "own where it runs, rent the model" approach feels like the future, and why a number like "Meta spent 71 trillion tokens" tells you almost nothing about what got built. They also take apart the tired "now anyone can code" claim with a little help from Ratatouille, and land on a simple ask: think through your trade-offs before you repeat the take. We've got supporter content, of course! This week that includes a breakdown of flex pricing and overnight agents, a no-mercy ranking of the Claude, Codex, and Gemini ecosystems (plus whose face is more punchable), the hundred-million-token saga of fixing Time Machine backups, Jevons paradox coming for your codebase, and a full teardown of why the IRS can't just send you a bill. 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. 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! Table of Contents: Prologue (00:00:00) Chapter 1: Fable's Extended Window and a $300 Opus Surprise (00:02:36) Chapter 2: Subscriptions vs API Pricing: Where the Money Is Heading (00:03:51) Chapter 4: AMP's Headless Agent: Own Where It Runs, Rent the Model (00:09:30) Chapter 6: The Hot Take: Ownership Is Bad (00:14:34) Chapter 7: IP Hoarding, Palantir, and Why Token Pricing Is Correct (00:18:38) Chapter 8: Be Honest With Yourself: Own vs Rent Math (00:24:07) Chapter 9: Token Counts Aren't Output: The Meta 71 Trillion Myth (00:29:44) Chapter 11: Local Hardware Trade-offs: Speed, Availability, Budget (00:40:06) Chapter 13: Anyone Can Cook, But Not Everyone Can (00:43:52) Chapter 14: Why the IRS Can't Just Send You a Bill [Extended] (00:50:25) Chapter 15: Check Your Gut: Chat Search and Apple Intelligence (00:50:25) Chapter 16: Home Ownership, Raises, and the Chip Shortage (00:53:27) Epilogue (00:57:25) Hosts Matthew Sanabria - Host Kris Brandow - Host Socials: Website Bluesky Threads X/Twitter LinkedIn Instagram TikTok Changelog Zulip Gophers Slack (00:00) - Prologue (02:36) - Chapter 1: Fable's Extended Window and a $300 Opus Surprise (03:51) - Chapter 2: Subscriptions vs API Pricing: Where the Money Is Heading (09:30) - Chapter 4: AMP's Headless Agent: Own Where It Runs, Rent the Model (14:34) - Chapter 6: The Hot Take: Ownership Is Bad (18:38) - Chapter 7: IP Hoarding, Palantir, and Why Token Pricing Is Correct (24:07) - Chapter 8: Be Honest With Yourself: Own vs Rent Math (29:44) - Chapter 9: Token Counts Aren't Output: The Meta 71 Trillion Myth (40:06) - Chapter 11: Local Hardware Trade-offs: Speed, Availability, Budget (43:52) - Chapter 13: Anyone Can Cook, But Not Everyone Can (50:25) - Chapter 15: Check Your Gut: Chat Search and Apple Intelligence (53:27) - Chapter 16: Home Ownership, Raises, and the Chip Shortage (57:25) - Epilogue

    • Chapters
  • #78
    July 4 · 1 hr

    Under Pressure

    On this week's episode, Kris and Matthew have a topic: tension. Not the bad kind you learn to avoid, but the productive kind that, like a code smell, is worth stopping to examine. They discuss the challenges of APIs and stability, whether SDKs are actually necessary, and as always a little bit of life advice for our wonderful viewers/listeners. We've got supporter content, of course! This week that includes Kris's deep dive on why it's actually called soccer (and the perfect apple-versus-fruit comeback for every European), his love letter to Gerrit and patch-based review, Matthew on why the pull request description means jack and what Jujutsu gets right, Kris's saga of building his own podcast platform out of Supercast, Transistor, and Riverside, and the great "machines that build machines, or is that just CICD?" debate. 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/3y6D1Gp4TSM. 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! Table of Contents: Prologue (00:00:00) Chapter 1: Heat, the World Cup, and America Through European Eyes (00:00:52) Chapter 3: Fable Returns and the One Person Agent Operation System (00:05:35) Chapter 4: Issue Trackers Aren't Dead: Linear, plan.md, and Keeping Context (00:10:47) Chapter 7: Solutions Versus Tools and the Meta Repo Problem (00:14:40) Chapter 9: Bending to the Tool: Forks, Filesystems, and the git flow Confession (00:18:34) Chapter 10: Code Smells and the Oxide API Versioning Tension (00:31:11) Chapter 11: Stopping the World: Backward Compatibility and the Summer of Hell (00:38:08) Chapter 12: Do We Still Need SDKs? (00:48:44) Epilogue (00:57:36) Hosts Matthew Sanabria - Host Kris Brandow - Host Socials: Website Bluesky Threads X/Twitter LinkedIn Instagram TikTok Changelog Zulip Gophers Slack (00:00) - Prologue (00:52) - Chapter 1: Heat, the World Cup, and America Through European Eyes (05:35) - Chapter 3: Fable Returns and the One Person Agent Operation System (10:47) - Chapter 4: Issue Trackers Aren't Dead: Linear, plan.md, and Keeping Context (14:40) - Chapter 7: Solutions Versus Tools and the Meta Repo Problem (18:34) - Chapter 9: Bending to the Tool: Forks, Filesystems, and the git flow Confession (31:11) - Chapter 10: Code Smells and the Oxide API Versioning Tension (38:08) - Chapter 11: Stopping the World: Backward Compatibility and the Summer of Hell (48:44) - Chapter 12: Do We Still Need SDKs? (57:36) - Epilogue

    • Chapters
  • #77
    June 27 · 27 min

    The Spy Who Banned Me

    It's Fallthrough time! In this episode, Kris, Matthew, and Steve get into the time Steve pointed Anthropic's Fable at a side project he hadn't touched since January and watched it shepherd 95 pull requests in three days, the US government export banning Fable and its sibling Mythos, the insanity of Mythos popping a national security agency's boxes in a matter of hours, and why a software engineer that runs 24/7 for anyone with an account is exactly the kind of thing governments get nervous about. We've got supporter content, of course! This week that includes whether your AI agents should file their bugs in your repo or on a private board where they can make as big a mess as they want, the model that forged itself a cookie to get around a testing problem and what it means for an AI to actually understand being a software developer, Steve's deliberately-out-of-context claim that millennials are the only people who know how computers work, Trump signing a quantum cryptography order while openly asking what the whole thing even is, Kris's year-long sales tax rabbit hole where nobody can agree where a person physically is, and the Diet Coke can math that says one exploding rocket is about one percent of an hour of making aluminum. 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/h6Ve7WKIQVc. 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! Table of Contents: Prologue (00:00:00) Chapter 1: Burning the Whole Allotment, 95 PRs in Three Days (00:01:56) Chapter 2: Beads vs Linear, Where Agents Keep Their Bugs [Extended] (00:07:24) Chapter 3: Is Fable Coming Back? Export Controls and the NSA Hack (00:07:24) Chapter 5: A Software Engineer in Every Pocket, Running 24/7 (00:12:57) Chapter 6: Anthropic Can't Explain Itself, Dario and the Death of Media Training (00:15:20) Chapter 13: People Are People, Contradictions and Closing the Loop (00:19:16) Epilogue (00:23:33) Hosts Matthew Sanabria - Host Kris Brandow - Host Steve Klabnik - Host Socials: Website Bluesky Threads X/Twitter LinkedIn Instagram TikTok Changelog Zulip Gophers Slack (00:00) - Prologue (01:56) - Chapter 1: Burning the Whole Allotment, 95 PRs in Three Days (07:24) - Chapter 3: Is Fable Coming Back? Export Controls and the NSA Hack (12:57) - Chapter 5: A Software Engineer in Every Pocket, Running 24/7 (15:20) - Chapter 6: Anthropic Can't Explain Itself, Dario and the Death of Media Training (19:16) - Chapter 13: People Are People, Contradictions and Closing the Loop (23:33) - Epilogue

    • Chapters
  • #76
    June 20 · 1 hr 7 min

    Of Fables and Mythos

    Another week, another Fallthrough! In this episode, Kris and Matthew get into the US government slapping export controls on Anthropic's Mythos and Fable models, the marketing machine that made them sound world-ending, and how much of the AI hype actually holds up. We've got supporter content, of course! This week that includes whether your phone is already an extension of your mind and the "Clean Talk" cleaning influencers who will happily tell you to mix two chemicals that quietly kill you, the Egyptologist on a date who proves a person is smart but people are dumb, why we throw you in jail for robbing a bank but only fine you for wrecking the economy, a no-holds-barred roast of Claude Code and the flickering bug that still will not die, and why AI in the creative fields is really just Photoshop with no shortcut to doing good work. 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/zzMFABZ0Rjs. 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! Table of Contents: Prologue (00:00:00) Chapter 1: Mythos and Fable, the Export Control, and Anthropic's Marketing Machine (00:02:18) Chapter 2: Freedom of Thought, Safety Gating, and Books in the Library (00:06:40) Chapter 6: Why You Can't Believe the AI Labs: Dogfooding, Lines of Code, and No Profit (00:11:11) Chapter 7: Nobody Knows Where the Ceiling Is, and the Coming Commodity Market (00:20:22) Chapter 8: Local Models, the Chocolate Shop AI Node, and Harness Over Model (00:22:27) Chapter 9: So Why Is Fable Actually Banned? Brain Drain, DOGE, and Jailbreaks (00:29:46) Chapter 10: It's Still All Hype, and the Government Did Anthropic's Marketing (00:37:37) Chapter 12: The Sentiment Shift: From Anthropic Darling to Team OpenAI (00:44:43) Chapter 13: The Future Is Harnesses, Not Models, and OpenAI's Better Bet (00:50:35) Chapter 14: Is Software Dead? Jobs, the Coming Shortage, and Don't Let Your Brain Atrophy (00:54:48) Chapter 15: AI in Creative Fields, Photoshop, and Why There's No Shortcut to Good Work [Extended] (01:05:39) Epilogue (01:05:39) Hosts Matthew Sanabria - Host Kris Brandow - Host Socials: Website Bluesky Threads X/Twitter LinkedIn Instagram TikTok Changelog Zulip Gophers Slack (00:00) - Prologue (02:18) - Chapter 1: Mythos and Fable, the Export Control, and Anthropic's Marketing Machine (06:40) - Chapter 2: Freedom of Thought, Safety Gating, and Books in the Library (11:11) - Chapter 6: Why You Can't Believe the AI Labs: Dogfooding, Lines of Code, and No Profit (20:22) - Chapter 7: Nobody Knows Where the Ceiling Is, and the Coming Commodity Market (22:27) - Chapter 8: Local Models, the Chocolate Shop AI Node, and Harness Over Model (29:46) - Chapter 9: So Why Is Fable Actually Banned? Brain Drain, DOGE, and Jailbreaks (37:37) - Chapter 10: It's Still All Hype, and the Government Did Anthropic's Marketing (44:43) - Chapter 12: The Sentiment Shift: From Anthropic Darling to Team OpenAI (50:35) - Chapter 13: The Future Is Harnesses, Not Models, and OpenAI's Better Bet (54:48) - Chapter 14: Is Software Dead? Jobs, the Coming Shortage, and Don't Let Your Brain Atrophy (01:05:39) - Epilogue

    • Chapters
  • #75
    June 13 · 48 min

    Shall We Play A Game?

    Matt and Kris welcome Fernando Duran, the creator of SAD Servers, the site that drops you onto a broken Linux box over SSH with a timer running and dares you to fix it. They go through what SAD Servers is, the spite-fueled interview story that started it all, how companies use it to hire, and the surprisingly lean infrastructure (plus the tiny Go "checker") that makes the whole thing work. We've got supporter content, of course! This week that includes why technical interviews are broken and the spite that turned one into a business, taking the number two spot on Hacker News without toppling over and the layers that keep the bad actors out, why three availability zones might just be expensive cargo-culting, the metrics that quietly lie to you (from Google-style error budgets to mean time to recovery versus mean time between failures), and the company that accidentally spent half a billion dollars on tokens. 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. 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! Table of Contents: Prologue (00:00:00) Chapter 1: What SAD Servers Is, Click a Button, Get a Broken Server (00:01:57) Chapter 2: Why He Built It, Troubleshooting, Learning, and Bad Interviews (00:03:33) Chapter 3: From Idea to AMI, How Scenarios Get Built (00:08:53) Chapter 4: Why Tech Interviews Are Broken [Extended] (00:13:58) Chapter 5: How Companies Use SAD Servers to Hire (00:23:27) Chapter 6: Kubernetes Playgrounds, SAD Pager, and Scenario Design (00:26:37) Chapter 7: The Architecture, Django, Ephemeral VMs, and the Consul Proxy (00:33:52) Chapter 8: Running Lean, Spot Instances, SQLite, and Surviving in Ohio (00:43:06) Chapter 9: Keeping the Bad Actors Out, Rate Limits and Layered Defense [Extended] (00:46:28) Chapter 10: The Checker, a Tiny Go Agent and the Problem of "Did You Win?" (00:55:49) Chapter 11: AI Evaluations, SRE Battle and the Agent Gold Rush (01:00:41) Chapter 12: Spicy Takes I, Availability Zones Are Overrated [Extended] (01:04:37) Chapter 13: Spicy Takes II, Error Budgets and the Google Cargo Cult [Extended] (01:15:34) Chapter 14: Spicy Takes III, MTBF vs MTTR, Bad Metrics, and Load Testing Theater [Extended] (01:26:07) Chapter 15: Spicy Takes IV, the Token Economy and Why Deleting Code Still Matters [Extended] (01:32:49) Epilogue (01:38:26) Hosts Matthew Sanabria - Host Kris Brandow - Host Fernando Duran - Guest Socials: Website Bluesky Threads X/Twitter LinkedIn Instagram TikTok Changelog Zulip Gophers Slack (00:00) - Prologue (01:57) - Chapter 1: What SAD Servers Is, Click a Button, Get a Broken Server (03:33) - Chapter 2: Why He Built It, Troubleshooting, Learning, and Bad Interviews (08:53) - Chapter 3: From Idea to AMI, How Scenarios Get Built (13:58) - Chapter 4: Why Tech Interviews Are Broken [Extended] (23:27) - Chapter 5: How Companies Use SAD Servers to Hire (26:37) - Chapter 6: Kubernetes Playgrounds, SAD Pager, and Scenario Design (33:52) - Chapter 7: The Architecture, Django, Ephemeral VMs, and the Consul Proxy (43:06) - Chapter 8: Running Lean, Spot Instances, SQLite, and Surviving in Ohio (46:28) - Chapter 9: Keeping the Bad Actors Out, Rate Limits and Layered Defense [Extended] (55:49) - Chapter 10: The Checker, a Tiny Go Agent and the Problem of "Did You Win?" (01:00:41) - Chapter 11: AI Evaluations, SRE Battle and the Agent Gold Rush (01:04:37) - Chapter 12: Spicy Takes I, Availability Zones Are Overrated [Extended] (01:15:34) - Chapter 13: Spicy Takes II, Error Budgets and the Google Cargo Cult [Extended] (01:26:07) - Chapter 14: Spicy Takes III, MTBF vs MTTR, Bad Metrics, and Load Testing Theater [Extended] (01:32:49) - Chapter 15: Spicy Takes IV, the Token Economy and Why Deleting Code Still Matters [Extended] (01:38:26) - Epilogue

    • Chapters
  • #74
    June 6 · 1 hr 28 min

    Pull the (AI) Lever, Kronk!

    Matt and Kris welcome back Bill Kennedy who's been working on Kronk, AI tooling in Go. They go through what Kronk (and it's dependent library Yzma) is, the power of running models locally, and how they see coding agents and LLMs shaping the future of the industry. We've got supporter content, of course! This week that includes the message-caching trick that turns seconds into thirty milliseconds, why hybrid models quietly wreck the KV cache, Bill's "thirty billion knobs" explainer for dense versus mixture-of-experts models, why demo day is destroying engineering teams, and the stealth-startup math nobody says out loud. 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. 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! Table of Contents: Prologue (00:00:00) Chapter 1: Why Go for AI, and the Origin of Kronk (00:02:14) Chapter 2: What Kronk Is, an SDK to Kill the Model Server (00:08:56) Chapter 3: Running It Yourself: Hugging Face, GGUF, and Hardware (00:14:58) Chapter 8: Context, Attention, and the Honesty of Claude 4.8 (00:22:28) Chapter 9: A Coding Agent Is a Skill, Not a Magic Wand (00:27:22) Chapter 10: Semantics vs Mechanics (00:44:18) Chapter 11: Rebuilding From Scratch: Specs and the Machine That Builds the Machine (00:47:29) Chapter 14: Why AMP Wins: Pedigree and Economics (01:00:20) Chapter 15: How to Actually Drive a Model: Plinko, Rocky Balboa, and Attention (01:14:59) Epilogue (01:25:08) Hosts Matthew Sanabria - Host Kris Brandow - Host Bill Kennedy - Guest Socials: Website Bluesky Threads X/Twitter LinkedIn Instagram TikTok Changelog Zulip Gophers Slack (00:00) - Prologue (02:14) - Chapter 1: Why Go for AI, and the Origin of Kronk (08:56) - Chapter 2: What Kronk Is, an SDK to Kill the Model Server (14:58) - Chapter 3: Running It Yourself: Hugging Face, GGUF, and Hardware (22:28) - Chapter 8: Context, Attention, and the Honesty of Claude 4.8 (27:22) - Chapter 9: A Coding Agent Is a Skill, Not a Magic Wand (44:18) - Chapter 10: Semantics vs Mechanics (47:29) - Chapter 11: Rebuilding From Scratch: Specs and the Machine That Builds the Machine (01:00:20) - Chapter 14: Why AMP Wins: Pedigree and Economics (01:14:59) - Chapter 15: How to Actually Drive a Model: Plinko, Rocky Balboa, and Attention (01:25:08) - Epilogue

    • Chapters
  • #73
    May 30 · 1 hr 1 min

    Dependencies All The Way Down

    Another week, another Fallthrough! In this episode, Kris and Matthew discuss a personal realization about what AI actually unlocks, their workflow philosophies, local models, a recent hot take from Mitchell Hashimoto, and a thought experiment that would break the NPM ecosystem overnight. Like the previous episode (a pseudo part 1), this episode returns to one question: are you moving in a direction, or just moving fast? Kris argues AI has lifted his coding domain ceiling entirely, while Matthew brings some nuance to the "just submit to the LLM, you're an architect now" take. The duo then turns outward: frontier lab CEOs claiming all white-collar work is automated while posting 300 open roles, the strong case for running local models (free, fast, and available right now), and a GitHub supply chain wake-up call triggered by a malicious VSCode plugin installed by an employee. We've got supporter content, of course! This week that includes Matthew's Shopify in-store pickup problem and whether building your own cash register on Stripe is actually feasible, Kris going deep on typography after Claude critiqued its own publication designs as screaming AI, the pre-fill versus decode speed gap and what it means for how you write prompts, how the US federal budget actually works day to day and why simple answers are the wrong frame, and Stripe's two-business-day payout model versus merchant of record, plus Rhode Island's ongoing war against Matthew's chocolate shop. 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. 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! Table of Contents: Prologue (00:00:00) Chapter 1: AI as a Capability Unlock (00:02:21) Chapter 2: The Chocolate Shop Problem, or: Building vs. Understanding [Extended] (00:05:18) Chapter 3: The Adam Jacob Take and Multitudes of AI Workflows (00:05:44) Chapter 4: Personal Software is the Thing Now (00:18:55) Chapter 5: Fear of Being Left Behind, Speed vs. Velocity (00:23:43) Chapter 6: The Fog of Knowledge, Typography, and AI-Generated Mediocrity [Extended] (00:30:11) Chapter 7: Tuning Your AI Tools, AGENTS.md, and the Out-of-Box Problem (00:30:29) Chapter 8: The LLM as Kernel: Understanding Agents from the Inside Out [Extended] (00:33:47) Chapter 10: Tech CEO Hype, Anthropic's Kool-Aid, and Self-Exemption (00:34:22) Chapter 11: Owning Your Commerce Stack: E-Commerce, Stripe, and Sales Tax Hell [Extended] (00:40:46) Chapter 12: Personal AI is the Future, Don't Sleep on Local Models (00:41:06) Chapter 13: Cerebras IPO, Oxide Rack Speculation (00:44:35) Chapter 14: GitHub Compromised via VSCode Plugin, Supply Chain Reckoning (00:47:05) Chapter 15: Open Source Sustainability, Joy vs. Product, and Pay-Per-Pull (00:54:15) Epilogue (01:00:06) Hosts Matthew Sanabria - Host Kris Brandow - Host Socials: Website Bluesky Threads X/Twitter LinkedIn Instagram TikTok Changelog Zulip Gophers Slack (00:00) - Prologue (02:21) - Chapter 1: AI as a Capability Unlock (05:18) - Chapter 2: The Chocolate Shop Problem, or: Building vs. Understanding [Extended] (05:44) - Chapter 3: The Adam Jacob Take and Multitudes of AI Workflows (18:55) - Chapter 4: Personal Software is the Thing Now (23:43) - Chapter 5: Fear of Being Left Behind, Speed vs. Velocity (30:11) - Chapter 6: The Fog of Knowledge, Typography, and AI-Generated Mediocrity [Extended] (30:29) - Chapter 7: Tuning Your AI Tools, AGENTS.md, and the Out-of-Box Problem (33:47) - Chapter 8: The LLM as Kernel: Understanding Agents from the Inside Out [Extended] (34:22) - Chapter 10: Tech CEO Hype, Anthropic's Kool-Aid, and Self-Exemption (40:46) - Chapter 11: Owning Your Commerce Stack: E-Commerce, Stripe, and Sales Tax Hell [Extended] (41:06) - Chapter 12: Personal AI is the Future, Don't Sleep on Local Models (44:35) - Chapter 13: Cerebras IPO, Oxide Rack Speculation (47:05) - Chapter 14: GitHub Compromised via VSCode Plugin, Supply Chain Reckoning (54:15) - Chapter 15: Open Source Sustainability, Joy vs. Product, and Pay-Per-Pull (01:00:06) - Epilogue

    • Chapters
  • #72
    May 23 · 32 min

    Terminal Velocity

    Has AI eaten all of tech? Is 600 billion tokens too many use in a single month? On this week's episode Matt joins Kris to talk about some of the current news in the tech industry. From OpenClaw's maintainer burning through $1.3 million worth of tokens to why we need more epistemic markers, to the rather absurd $725 billion that big tech plans to spend on AI CapEx this year. And of course, there are layoffs coming as well. Support content? We've got a whole bunch of that! This week that includes a ZFS war story where a thousand-fold write amplification bug stalls every file write, the case for why LLMs hand you mediocre output by default, today's OAuth and GitHub security incidents traced back to a standards fight twenty years ago, a spicy take on who actually holds power inside big tech, and Matthew hand-writing an HTTP server in Rust just to fight off skill atrophy. 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. 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! Table of Contents: Prologue (00:00:00) Chapter 1: AI Has Eaten All of Tech (00:01:11) Chapter 2: Building the Machine That Builds Your Machine (00:06:30) Chapter 3: You Still Have to Understand the Layer Below [Extended] (00:14:56) Chapter 4: Say the Quiet Part: Put a Number on Your Predictions (00:15:07) Chapter 5: LLMs Give You Mediocre by Default [Extended] (00:17:53) Chapter 6: Speed Is Not Velocity (00:18:21) Chapter 7: OAuth, GitHub, and the Bill for Skipped Homework [Extended] (00:25:40) Chapter 9: $725 Billion in CapEx and a Wave of Layoffs (00:26:26) Chapter 10: Who Actually Has the Power [Extended] (00:30:40) Epilogue (00:31:45) Hosts Matthew Sanabria - Host Kris Brandow - Host Socials: Website Bluesky Threads X/Twitter LinkedIn Instagram Changelog Zulip Gophers Slack (00:00) - Prologue (01:11) - Chapter 1: AI Has Eaten All of Tech (06:30) - Chapter 2: Building the Machine That Builds Your Machine (14:56) - Chapter 3: You Still Have to Understand the Layer Below [Extended] (15:07) - Chapter 4: Say the Quiet Part: Put a Number on Your Predictions (17:53) - Chapter 5: LLMs Give You Mediocre by Default [Extended] (18:21) - Chapter 6: Speed Is Not Velocity (25:40) - Chapter 7: OAuth, GitHub, and the Bill for Skipped Homework [Extended] (26:26) - Chapter 9: $725 Billion in CapEx and a Wave of Layoffs (30:40) - Chapter 10: Who Actually Has the Power [Extended] (31:45) - Epilogue

    • Chapters
  • #71
    May 16 · 1 hr 25 min

    He Who Controls the Source

    This week we've got Kris and Jamie! They open with the Shai-Hulud worm chewing through the npm supply chain and close on the messy economics of who actually pays for open source labor. And there's plenty of great stuff in between: GitHub's everything-platform creep, the case for LLMs as a way out of dependency hell, and the forge alternatives finally maturing into real options. We've got supporter content, of course! This week that includes an expansion on Kris's "pull requests as original sin" theory, the everything-platform rant that compares GitHub to Meta Business Suite, a monologue on money, knowledge gaps, and LLMs as a way out of open source debt. 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. 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! Table of Contents: Prologue (00:00:00) Chapter 2: The Shai-Hulud Worm and the GitHub Actions Attack Surface (00:00:59) Chapter 3: Pull Requests as Original Sin [Extended] (00:19:41) Chapter 4: GitHub Enterprise Cloud and the Forge Alternatives (00:19:59) Chapter 5: The Everything Platform Problem [Extended] (00:28:17) Chapter 6: GitLab Counterpoint, Kubernetes, Feature Flags, and Friction (00:37:58) Chapter 7: Walled Gardens and the Business Model of Open Source [Extended] (00:43:06) Chapter 8: AI Safety, Napalm Grandmas, and Agentic PRs (00:43:24) Chapter 9: Kris's Theory: Money, Knowledge Gaps, and LLMs as a Way Out of Open Source Debt [Extended] (00:46:37) Chapter 10: Jamie Pushes Back: Forks, Vulnerabilities, and OAPI CodeGen Stalled on OpenAPI 3.1 (00:46:58) Chapter 11: Making Open Source a Surfaceable Cost (00:54:29) Chapter 12: The Open Source Resistance (01:07:35) Chapter 13: Capital One, Director Sign-off, and the Lawyer's Perspective [Extended] (01:16:52) Chapter 16: FSNotify Cleanup and the Composer Token Leak (01:17:05) Epilogue (01:23:46) Hosts Kris Brandow - Host Jamie Tanna - Host Socials: Website Bluesky Threads X/Twitter LinkedIn Instagram Changelog Zulip Gophers Slack (00:00) - Prologue (00:59) - Chapter 2: The Shai-Hulud Worm and the GitHub Actions Attack Surface (19:41) - Chapter 3: Pull Requests as Original Sin [Extended] (19:59) - Chapter 4: GitHub Enterprise Cloud and the Forge Alternatives (28:17) - Chapter 5: The Everything Platform Problem [Extended] (37:58) - Chapter 6: GitLab Counterpoint, Kubernetes, Feature Flags, and Friction (43:06) - Chapter 7: Walled Gardens and the Business Model of Open Source [Extended] (43:24) - Chapter 8: AI Safety, Napalm Grandmas, and Agentic PRs (46:37) - Chapter 9: Kris's Theory: Money, Knowledge Gaps, and LLMs as a Way Out of Open Source Debt [Extended] (46:58) - Chapter 10: Jamie Pushes Back: Forks, Vulnerabilities, and OAPI CodeGen Stalled on OpenAPI 3.1 (54:29) - Chapter 11: Making Open Source a Surfaceable Cost (01:07:35) - Chapter 12: The Open Source Resistance (01:16:52) - Chapter 13: Capital One, Director Sign-off, and the Lawyer's Perspective [Extended] (01:17:05) - Chapter 16: FSNotify Cleanup and the Composer Token Leak (01:23:46) - Epilogue

    • Chapters
  • #70
    May 9 · 1 hr 26 min

    Forging Ahead

    Steve is back to talk JJ (Jujutsu version control) and the related product, ChangeSet, that he works on at East River Source Control. Kris and Steve trace why the GitHub monoculture is finally cracking, what JJ does that Git can't, and Steve's hypothesis that AI agents are pushing companies toward monorepos. Then the pair discuss the Opus 4.7 regression debate, the shift from "always use the frontier model" to using LLMs as one tool among many, and a quick discussion of the Tim Cook to John Ternus handoff at Apple. We've got supporter content, of course! This week that includes Steve's broader thesis that it's easier to scale a big tool down than scale a small one up, why GitHub's pull request model warped how people use Git, auto-rebase and conflicts as first-class citizens in JJ, AI subsidies, and a Bun being ported to Rust. 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. Thanks for tuning in and happy listening! Show Notes: JJ Table of Contents: Prologue (00:00:00) Chapter 1: Episode 70, New Branding, and Catching Up (00:01:14) Chapter 2: What is JJ (Jujutsu) and Why It Exists (00:03:43) Chapter 3: Change IDs, Tangled, and the Federated Forge Wave (00:08:59) Chapter 4: Healthy Diversity After the GitHub Monoculture (00:14:33) Chapter 5: Scaling Big Tools Down: Bazel, Buck, Kubernetes, Symfony [Extended] (00:18:03) Chapter 6: JJ Is Simpler AND More Powerful: No Index, No Stash (00:18:10) Chapter 7: The JJ Workflow: Snapshots, Watchman, and JJ Undo (00:26:06) Chapter 8: GitHub Warped Git: Why Patches and Gerrit Are Better [Extended] (00:32:14) Chapter 9: Auto-Rebase and Conflicts as First-Class Citizens [Extended] (00:32:14) Chapter 10: Getting Started with JJ: Tutorials and Workflows (00:32:29) Chapter 11: East River Source Control and ChangeSet: A Forge for Mono-Repo Scale (00:42:28) Chapter 12: Why AI Forces Companies into Monorepo Scaling Sooner (00:50:01) Chapter 13: AI Subsidies and the Scaling Wall [Extended] (00:54:55) Chapter 14: The Opus 4.7 Regression Debate, Goodhart's Law, and Custom Styles (00:55:24) Chapter 15: LLMs as Tools, Not Solutions: Local Models, Gemini, and Custom Pipelines (01:04:49) Chapter 16: Bun Rewriting in Rust: A Branch That Wasn't Supposed to Be News [Extended] (01:15:48) Chapter 17: The Apple CEO Transition: Tim Cook to John Ternus (01:16:08) Epilogue (01:24:55) Hosts Kris Brandow - Host Steve Klabnik - Host Socials: Website Bluesky Threads X/Twitter LinkedIn Instagram Changelog Zulip Gophers Slack (00:00) - Prologue (01:14) - Chapter 1: Episode 70, New Branding, and Catching Up (03:43) - Chapter 2: What is JJ (Jujutsu) and Why It Exists (08:59) - Chapter 3: Change IDs, Tangled, and the Federated Forge Wave (14:33) - Chapter 4: Healthy Diversity After the GitHub Monoculture (18:03) - Chapter 5: Scaling Big Tools Down: Bazel, Buck, Kubernetes, Symfony [Extended] (18:10) - Chapter 6: JJ Is Simpler AND More Powerful: No Index, No Stash (26:06) - Chapter 7: The JJ Workflow: Snapshots, Watchman, and JJ Undo (32:14) - Chapter 8: GitHub Warped Git: Why Patches and Gerrit Are Better [Extended] (32:29) - Chapter 10: Getting Started with JJ: Tutorials and Workflows (42:28) - Chapter 11: East River Source Control and ChangeSet: A Forge for Mono-Repo Scale (50:01) - Chapter 12: Why AI Forces Companies into Monorepo Scaling Sooner (54:55) - Chapter 13: AI Subsidies and the Scaling Wall [Extended] (55:24) - Chapter 14: The Opus 4.7 Regression Debate, Goodhart's Law, and Custom Styles (01:04:49) - Chapter 15: LLMs as Tools, Not Solutions: Local Models, Gemini, and Custom Pipelines (01:15:48) - Chapter 16: Bun Rewriting in Rust: A Branch That Wasn't Supposed to Be News [Extended] (01:16:08) - Chapter 17: The Apple CEO Transition: Tim Cook to John Ternus (01:24:55) - Epilogue

    • Chapters
  • #69
    May 2 · 1 hr 11 min

    Regression to the Mean

    Kris and Ian dig into the slow collapse of GitHub, starting with Ghostty off the platform after years of reliability problems. From there they trace Gary Bernhardt's old observation that we took a decentralized source control system and immediately put it behind a single point of failure, then widen the lens into AI as the engine of enshittification. The episode lands on a more optimistic note: maybe AI is also the tool that lets individuals rebuild the apps they used to have to buy. We've got supporter content, of course! This week that includes Kris's "AI was trained on the median of human code" argument, the institutional-memory thesis for why AI can't just replace people, a tour of the local-AI hardware landscape with Zig and llama.cpp, Ian's relatable "what is Cursor? what is Warp?" rant, a policy pitch on share-flipper voting rules, and a deep dive on typography as an anti-AI signal. 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. No episode of Break 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! Table of Contents: Prologue (00:00:00) Chapter 1: GitHub Is Falling Over: Ghostty Leaves and the Merge Bug (00:01:46) Chapter 2: Centralizing the Decentralized: Git, Forges, and the Linux Kernel Workflow (00:08:04) Chapter 3: Self-Hosting vs. The Cloud Tax (00:16:23) Chapter 6: AI as the Engine of Enshittification (00:32:02) Chapter 9: Epistemic Enshittification: Hedges, Hype, and Sci-Fi Brain (00:37:10) Chapter 10: Self-Checkout, Junior Devs, and Broken Hiring (00:46:52) Chapter 12: Fallthrough Rebrand and the Claude Design Experiment (00:54:16) Chapter 13: Typography as Anti-AI Signal: Buying Real Typefaces [Extended] (01:04:33) Chapter 14: Bleeps, Shorts, and Producing With AI (01:04:36) Chapter 15: AI as the De-Enshittification Tool (01:08:08) Epilogue (01:10:22) Hosts Kris Brandow - Host Ian Wester-Lopshire - Host Socials: Website Bluesky Threads X/Twitter LinkedIn Instagram (00:00) - Prologue (01:46) - Chapter 1: GitHub Is Falling Over: Ghostty Leaves and the Merge Bug (08:04) - Chapter 2: Centralizing the Decentralized: Git, Forges, and the Linux Kernel Workflow (16:23) - Chapter 3: Self-Hosting vs. The Cloud Tax (32:02) - Chapter 4: AI Code Quality and the Mediocre Mean [Extended] (37:10) - Chapter 9: Epistemic Enshittification: Hedges, Hype, and Sci-Fi Brain (46:52) - Chapter 10: Self-Checkout, Junior Devs, and Broken Hiring (54:16) - Chapter 12: Fallthrough Rebrand and the Claude Design Experiment (01:04:33) - Chapter 13: Typography as Anti-AI Signal: Buying Real Typefaces [Extended] (01:04:36) - Chapter 14: Bleeps, Shorts, and Producing With AI (01:08:08) - Chapter 15: AI as the De-Enshittification Tool (01:10:22) - Epilogue

    • Chapters
  • #68
    April 25 · 49 min

    No Country for Old Maintainers

    Jamie returns, co-hosting with Kris for quite an eventful episode. They start with the Vercel breach, the Axios attack, nvim-treesitter, and Gorilla Mux. Kris draws parallels between the current AI hype cycle to everything that came before: Photoshop was going to destroy photographers, DAWs were going to destroy musicians, and now Claude Code is going to destroy software engineers. Just like the last episode, this one is filled with supporter only content. It's actually an extra episode and a half! This includes a deep dive on OAuth scope design and why consent screens need the Let's Encrypt treatment, Anthropic locking out third-party harnesses, sandboxing LLMs with sandbox-exec and agent-safehouse, Jamie on "no is a complete sentence" and Renovate's unusually pro-maintainer code of conduct, Kris's pitch for $100M of LLM spend going to help maintainers triage backlogs instead of Mythos-style vulnerability hunting, the real complexity behind finance departments and why $250M wire transfers need entire treasury teams, the math on a $800 used 3090 running Qwen 3.5 at Sonnet-level capability, and Anthropic's pivot toward non-engineers with Claude Design and Claude Cowork. 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. No episode of Break 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! Table of Contents: Prologue (00:00:00) Chapter 1: Eventful Spring, Banter and Setup (00:01:20) Chapter 2: The Vercel Breach: Roblox to Production Via One Employee (00:04:25) Chapter 6: Deepfakes, the Axios Attack, and the $5 Wrench (00:06:49) Chapter 7: The XZ Utils Attack and nvim-treesitter Gets Archived (00:11:22) Chapter 8: GorillaMux, the Go Standard Library, and Drive-By Forkers (00:14:13) Chapter 11: The Hype Cycle: Opus Got Dumber, and Every Tool Was Supposed to Replace Us (00:20:44) Chapter 13: Claude Code Going Max-Only, Copilot Pro Losing Opus (00:30:56) Chapter 16: Kris the Former Hater, and the Slop Zone [Extended] (00:34:12) Chapter 17: Deep Blue, Collaboration, and Why LLMs Aren't Replacing Us (00:38:33) Epilogue (00:44:03) Hosts Kris Brandow - Host Jamie Tanna - Host Socials: Website Bluesky Threads X/Twitter LinkedIn Instagram (00:00) - Prologue (01:20) - Chapter 1: Eventful Spring, Banter and Setup (04:25) - Chapter 2: The Vercel Breach: Roblox to Production Via One Employee (06:49) - Chapter 6: Deepfakes, the Axios Attack, and the $5 Wrench (11:22) - Chapter 7: The XZ Utils Attack and nvim-treesitter Gets Archived (14:13) - Chapter 8: GorillaMux, the Go Standard Library, and Drive-By Forkers (20:44) - Chapter 11: The Hype Cycle: Opus Got Dumber, and Every Tool Was Supposed to Replace Us (30:56) - Chapter 13: Claude Code Going Max-Only, Copilot Pro Losing Opus (34:12) - Chapter 16: Kris the Former Hater, and the Slop Zone [Extended] (38:33) - Chapter 17: Deep Blue, Collaboration, and Why LLMs Aren't Replacing Us (44:03) - Epilogue

    • Chapters
  • #67
    April 18 · 1 hr 9 min

    Supply Chain Reaction

    After last week's "Another Spectre In The Shell" episode, we felt we needed a follow up! This week Kris is joined by returning co-host Jamie Tanna, with Ian arriving fashionably late. They pick apart the post-Mythos announcement hype cycle, cover counter-narratives from the security community, and examine why our supply chain is already so broken that more powerful LLM barely changes the threat model. The conversation builds toward Jamie's unpop: "a little copying is better than a little dependency" is wrong, we should have more dependencies and then into practical supply chain defense. The episode closes on an infinite-mindset note: supply chain security will never be done and we'll always have things to improve. Supporter content? This episode has a ton of it! If you haven't become a supporter yet, this is the episode to do it! In this week's extras we've got a critique of the "37,000 lines of code a day" claim, Kris's argument that we've squandered most productivity gains in history so why would AI be different, Claude Code fatigue and learning versus delegating, a deep yak-shave into why Kris hates Markdown and is building his publishing platform in raw HTML, a love letter to the web as the best technology we have, a merchant-of-record tangent on VAT and why Stripe isn't enough, Jamie's proposal for a Fallthrough bingo card, and a tangent on Anthropic having a therapist talk to Claude for twenty hours. If you aren't a supporter already, head over to https://fallthrough.fm/subscribe where you'll get not only extra content but also higher quality audio. If you prefer to watch this episode, you can view it on YouTube. No episode of Break 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! Table of Contents: Prologue (00:00:00) Chapter 1: Jamie Returns: Weather and Mythos Recap (00:01:11) Chapter 2: Post-Mythos Developments and Counter-Narratives (00:03:11) Chapter 3: $20K Tokens, Wrench Attacks, and Log4J Toasters (00:08:41) Chapter 8: Offshoring to the Digital Land (00:10:54) Chapter 16: Jamie's Unpop: The Go Proverb Is Wrong, Use More Dependencies (00:18:42) Chapter 17: More Dependencies, Bad Infrastructure, and Author Signing (00:24:02) Chapter 18: The Go Module Proxy Doesn't Understand Its Place (00:28:18) Chapter 19: Cooldowns and Minimum Release Age in Renovate (00:42:16) Chapter 20: Hype Cycles, Security Professionals, and End-User Responsibility (00:52:18) Chapter 25: Life Finds a Way: T-Rex Leather and Infinite Mindset (01:03:28) Epilogue (01:07:29) Hosts Kris Brandow - Host Ian Wester-Lopshire - Host Jamie Tanna - Host Socials: Website Bluesky Threads X/Twitter LinkedIn Instagram (00:00) - Prologue (01:11) - Chapter 1: Jamie Returns: Weather and Mythos Recap (03:11) - Chapter 2: Post-Mythos Developments and Counter-Narratives (08:41) - Chapter 3: $20K Tokens, Wrench Attacks, and Log4J Toasters (10:54) - Chapter 8: Offshoring to the Digital Land (18:42) - Chapter 16: Jamie's Unpop: The Go Proverb Is Wrong, Use More Dependencies (24:02) - Chapter 17: More Dependencies, Bad Infrastructure, and Author Signing (28:18) - Chapter 18: The Go Module Proxy Doesn't Understand Its Place (42:16) - Chapter 19: Cooldowns and Minimum Release Age in Renovate (52:18) - Chapter 20: Hype Cycles, Security Professionals, and End-User Responsibility (01:03:28) - Chapter 25: Life Finds a Way: T-Rex Leather and Infinite Mindset (01:07:29) - Epilogue

    • Chapters
Showing 1–20 of 23 episodes