<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:psc="http://podlove.org/simple-chapters" xmlns:podcast="https://podcastindex.org/namespace/1.0" xmlns:itunes="http://www.itunes.com/dtds/podcast-1.0.dtd"><channel><title><![CDATA[complexity.fm Podcast]]></title><description><![CDATA[Conversations with people in tech]]></description><link>https://complexity.fm</link><generator>loudblog</generator><lastBuildDate>Sat, 03 Oct 2026 18:38:51 GMT</lastBuildDate><atom:link href="https://feed.complexity.fm/" rel="self" type="application/rss+xml"/><author><![CDATA[Bas Steins & Michal Martinka]]></author><pubDate>Sat, 03 Oct 2026 18:38:51 GMT</pubDate><copyright><![CDATA[© 2026ff. Bas Steins & Michal Martinka. All rights reserved.]]></copyright><language><![CDATA[en-US]]></language><ttl>3600</ttl><category><![CDATA[Technology]]></category><itunes:author>Bas Steins &amp; Michal Martinka</itunes:author><itunes:subtitle>Conversations with people in tech</itunes:subtitle><itunes:summary>complexity.fm is a podcast about technology, podcasting, and more.</itunes:summary><itunes:type>episodic</itunes:type><itunes:owner><itunes:name>Bas Steins &amp; Michal Martinka</itunes:name><itunes:email>feed@bas.fm</itunes:email></itunes:owner><itunes:explicit>false</itunes:explicit><itunes:category text="Technology"></itunes:category><itunes:image href="https://complexity.fm/cdn-cgi/image/onerror=redirect,width=3000,height=3000,format=webp/_astro/poster.BrTozYWL.png"/><item><title><![CDATA[The Heroku Killer That Got Them Into YC]]></title><description><![CDATA[📄 Show Notes
In this episode of complexity.fm, Bas Steins and Michal Martinka sit down with Brandon Bayer, founder of Flightcontrol and creator of Blitz.js, to talk about the messy reality of deploying to AWS without a DevOps team. Brandon walks through why he built Flightcontrol as a Heroku alternative that deploys directly into your own AWS account, why the product ended up being "too limiting" after years of growth, and why he and his co-founder decided to rebuild everything from scratch with a new product called Ravion. The conversation dives deep into platform engineering, Terraform vs OpenTofu, the "DevOps is a mindset, not a role" debate, and how AI agents are reshaping the relationship between developers, DevOps engineers, and infrastructure itself. Brandon also shares the wild story of applying to YC as a "Hail Mary" with zero expectations, getting accepted the same day, and raising a million dollars within a week.

⏱️ Topics Covered
00:00 Introduction to Brandon Bayer and Flightcontrol
01:56 What Flightcontrol does and why it deploys into your own AWS account
02:30 The Heroku free tier shutdown and its (non) impact on Flightcontrol
05:16 Flightcontrol's target customer: teams without a dedicated DevOps engineer
07:11 The typical company journey from PaaS to AWS to platform engineering
09:09 Tech stack behind Flightcontrol and the pain of Next.js app router
10:48 Forking Next.js and building a Rails-like framework, Blitz.js
11:00 Flightcontrol's YC journey: fundraising struggles before acceptance
12:12 The YC interview process and mock interviews with Stripe Atlas
15:34 Getting the YC acceptance call and raising 1 million dollars in a week
18:52 Being in YC's last fully remote COVID batch
19:16 Getting first customers through Twitter and organic growth
24:14 Realizing Flightcontrol was "too limiting" after years of growth
25:16 Introducing Ravion: rebuilding the product from scratch
28:52 How Ravion uses Terraform instead of custom orchestration
30:52 Terraform vs OpenTofu and the licensing drama
35:49 Bringing your own Terraform and custom modules in Ravion
41:14 Cost visibility, logs, and metrics in one unified UI
44:37 Language and cloud agnosticism, RailPack, and Docker generation
48:41 Is DevOps a mindset, not a role? Shift-left and self-service
50:02 AI agents, guardrails, and sandboxed preview environments
55:59 Comparing Ravion to Crossplane and other IDP tools
59:19 Why deterministic infrastructure still matters in the AI agent era
1:03:53 How AI is changing team dynamics between developers and DevOps
1:09:10 Why small teams may be disrupting big companies faster thanks to AI

🌐 Resources & Links
Flightcontrol: https://www.flightcontrol.dev
Flightcontrol Docs: https://www.flightcontrol.dev/docs
Ravion Docs: https://docs.ravion.com
Blitz.js: https://blitzjs.com
Brandon Bayer on X: https://x.com/flybayer
Brandon Bayer on LinkedIn: https://www.linkedin.com/in/brandonbayer1
Flightcontrol on GitHub: https://github.com/flightcontrolhq
Flightcontrol on Y Combinator: https://www.ycombinator.com/companies/flightcontrol
]]></description><link>https://complexity.fm/e/4</link><guid isPermaLink="false">https://complexity.fm/e/4</guid><dc:creator><![CDATA[Bas Steins & Michal Martinka]]></dc:creator><pubDate>Tue, 21 Jul 2026 16:30:00 GMT</pubDate><enclosure url="https://media.complexity.fm/episodes/0004/audio.mp3" length="0" type="audio/mpeg"/><itunes:author>Bas Steins &amp; Michal Martinka</itunes:author><itunes:summary>📄 Show Notes
In this episode of complexity.fm, Bas Steins and Michal Martinka sit down with Brandon Bayer, founder of Flightcontrol and creator of Blitz.js, to talk about the messy reality of deploying to AWS without a DevOps team. Brandon walks through why he built Flightcontrol as a Heroku alternative that deploys directly into your own AWS account, why the product ended up being &quot;too limiting&quot; after years of growth, and why he and his co-founder decided to rebuild everything from scratch with a new product called Ravion. The conversation dives deep into platform engineering, Terraform vs OpenTofu, the &quot;DevOps is a mindset, not a role&quot; debate, and how AI agents are reshaping the relationship between developers, DevOps engineers, and infrastructure itself. Brandon also shares the wild story of applying to YC as a &quot;Hail Mary&quot; with zero expectations, getting accepted the same day, and raising a million dollars within a week.

⏱️ Topics Covered
00:00 Introduction to Brandon Bayer and Flightcontrol
01:56 What Flightcontrol does and why it deploys into your own AWS account
02:30 The Heroku free tier shutdown and its (non) impact on Flightcontrol
05:16 Flightcontrol&apos;s target customer: teams without a dedicated DevOps engineer
07:11 The typical company journey from PaaS to AWS to platform engineering
09:09 Tech stack behind Flightcontrol and the pain of Next.js app router
10:48 Forking Next.js and building a Rails-like framework, Blitz.js
11:00 Flightcontrol&apos;s YC journey: fundraising struggles before acceptance
12:12 The YC interview process and mock interviews with Stripe Atlas
15:34 Getting the YC acceptance call and raising 1 million dollars in a week
18:52 Being in YC&apos;s last fully remote COVID batch
19:16 Getting first customers through Twitter and organic growth
24:14 Realizing Flightcontrol was &quot;too limiting&quot; after years of growth
25:16 Introducing Ravion: rebuilding the product from scratch
28:52 How Ravion uses Terraform instead of custom orchestration
30:52 Terraform vs OpenTofu and the licensing drama
35:49 Bringing your own Terraform and custom modules in Ravion
41:14 Cost visibility, logs, and metrics in one unified UI
44:37 Language and cloud agnosticism, RailPack, and Docker generation
48:41 Is DevOps a mindset, not a role? Shift-left and self-service
50:02 AI agents, guardrails, and sandboxed preview environments
55:59 Comparing Ravion to Crossplane and other IDP tools
59:19 Why deterministic infrastructure still matters in the AI agent era
1:03:53 How AI is changing team dynamics between developers and DevOps
1:09:10 Why small teams may be disrupting big companies faster thanks to AI

🌐 Resources &amp; Links
Flightcontrol: https://www.flightcontrol.dev
Flightcontrol Docs: https://www.flightcontrol.dev/docs
Ravion Docs: https://docs.ravion.com
Blitz.js: https://blitzjs.com
Brandon Bayer on X: https://x.com/flybayer
Brandon Bayer on LinkedIn: https://www.linkedin.com/in/brandonbayer1
Flightcontrol on GitHub: https://github.com/flightcontrolhq
Flightcontrol on Y Combinator: https://www.ycombinator.com/companies/flightcontrol
</itunes:summary><itunes:explicit>false</itunes:explicit><itunes:image href="https://media.complexity.fm/episodes/0004/episode-artwork.png"/></item><item><title><![CDATA[Building in Public Is Dead (And AI Killed It)]]></title><description><![CDATA[📄 Show Notes

In this episode of complexity.fm, Bas Steins and Michal Martinka sit down with Dominik Sumer, the solo developer behind Snapify and Vimetric, to talk about a decade of building products in public and what happens when that playbook stops working. Dominik walks through how Snapify started as a Stack Overflow clone that flopped, then pivoted into a viral code-snippet-to-video editor built almost entirely through building-in-public momentum on old Twitter. He explains why that same strategy feels broken today, how AI-generated "one shot this" content has flooded social feeds and killed genuine audience engagement, and why he still misses the tight-knit developer communities of 2022 Twitter. The conversation moves into his newer project Vimetric, a privacy-first, open-source web analytics tool, where Dominik unpacks why he open-sourced it, how self-hosting could become the product's next big growth lever, and why he doesn't believe AI coding tools are actually killing SaaS businesses the way social media hot takes claim. The episode closes with a broader debate on whether AI chatbots will eventually replace dashboards entirely, and what that means for indie makers building tools today.

⏱️ Topics Covered

00:00 - Introduction to Dominik Sumer, 12 years as a professional developer
01:20 - The origin story of Snapify, from failed Stack Overflow clone to viral code-snippet tool
04:06 - Why Dominik stepped back from Snapify to start something new
04:29 - Introducing Vimetric, a privacy-first web and product analytics tool
08:00 - Building products from personal need as a solo developer
08:57 - Distribution and marketing through building in public
11:11 - Psychological vulnerability of sharing progress and failures publicly
15:08 - How AI has changed audience reactions and the loss of the "wow" factor
16:33 - The shift in social media from genuine engagement to attention-grabbing content
17:58 - Nostalgia for RSS readers and slower, deeper content consumption
19:13 - Why AI-generated "catchy" posts still perform despite feeling inauthentic
20:40 - The search for a modern equivalent of 2022 Twitter
23:07 - Hacker News as one of the last places for brutally honest feedback
24:22 - Riffing on building a template-free, AI-free social network
26:11 - Finally digging into AI directly
26:57 - Does AI actually cause customer churn for indie tools
30:58 - Why Vimetric is open source and what that decision was based on
33:10 - Does open sourcing actually help discovery and marketing
35:11 - Self-hosting adoption numbers and the missing one-click Docker setup
38:43 - Infrastructure details, Coolify on Hetzner, microservices, multi-tenant setup
40:25 - The chickens-and-eggs analogy for why people still pay for convenience
42:36 - Whether AI meaningfully changes the self-hosting versus SaaS calculus
44:15 - The real threat to SaaS, chatbots replacing dashboards entirely
44:40 - Closing thoughts on calorie trackers, workout apps, and the future interface

🌐 Resources & Links

Snapify: https://snappify.com
Vimetric: https://vimetric.com
Dominik Sumer on Twitter/X: https://twitter.com/dominiksumer
complexity.fm podcast: https://complexity.fm
Hosts: Bas Steins and Michal Martinka
Referenced: Plausible Analytics (mentioned as inspiration for open-source analytics tools)
Referenced: Carbon (mentioned as an early code-snippet sharing tool competitor to Snapify)
]]></description><link>https://complexity.fm/e/3</link><guid isPermaLink="false">https://complexity.fm/e/3</guid><dc:creator><![CDATA[Bas Steins & Michal Martinka]]></dc:creator><pubDate>Tue, 07 Jul 2026 16:30:00 GMT</pubDate><enclosure url="https://media.complexity.fm/episodes/0003/audio.mp3" length="0" type="audio/mpeg"/><itunes:author>Bas Steins &amp; Michal Martinka</itunes:author><itunes:summary>📄 Show Notes

In this episode of complexity.fm, Bas Steins and Michal Martinka sit down with Dominik Sumer, the solo developer behind Snapify and Vimetric, to talk about a decade of building products in public and what happens when that playbook stops working. Dominik walks through how Snapify started as a Stack Overflow clone that flopped, then pivoted into a viral code-snippet-to-video editor built almost entirely through building-in-public momentum on old Twitter. He explains why that same strategy feels broken today, how AI-generated &quot;one shot this&quot; content has flooded social feeds and killed genuine audience engagement, and why he still misses the tight-knit developer communities of 2022 Twitter. The conversation moves into his newer project Vimetric, a privacy-first, open-source web analytics tool, where Dominik unpacks why he open-sourced it, how self-hosting could become the product&apos;s next big growth lever, and why he doesn&apos;t believe AI coding tools are actually killing SaaS businesses the way social media hot takes claim. The episode closes with a broader debate on whether AI chatbots will eventually replace dashboards entirely, and what that means for indie makers building tools today.

⏱️ Topics Covered

00:00 - Introduction to Dominik Sumer, 12 years as a professional developer
01:20 - The origin story of Snapify, from failed Stack Overflow clone to viral code-snippet tool
04:06 - Why Dominik stepped back from Snapify to start something new
04:29 - Introducing Vimetric, a privacy-first web and product analytics tool
08:00 - Building products from personal need as a solo developer
08:57 - Distribution and marketing through building in public
11:11 - Psychological vulnerability of sharing progress and failures publicly
15:08 - How AI has changed audience reactions and the loss of the &quot;wow&quot; factor
16:33 - The shift in social media from genuine engagement to attention-grabbing content
17:58 - Nostalgia for RSS readers and slower, deeper content consumption
19:13 - Why AI-generated &quot;catchy&quot; posts still perform despite feeling inauthentic
20:40 - The search for a modern equivalent of 2022 Twitter
23:07 - Hacker News as one of the last places for brutally honest feedback
24:22 - Riffing on building a template-free, AI-free social network
26:11 - Finally digging into AI directly
26:57 - Does AI actually cause customer churn for indie tools
30:58 - Why Vimetric is open source and what that decision was based on
33:10 - Does open sourcing actually help discovery and marketing
35:11 - Self-hosting adoption numbers and the missing one-click Docker setup
38:43 - Infrastructure details, Coolify on Hetzner, microservices, multi-tenant setup
40:25 - The chickens-and-eggs analogy for why people still pay for convenience
42:36 - Whether AI meaningfully changes the self-hosting versus SaaS calculus
44:15 - The real threat to SaaS, chatbots replacing dashboards entirely
44:40 - Closing thoughts on calorie trackers, workout apps, and the future interface

🌐 Resources &amp; Links

Snapify: https://snappify.com
Vimetric: https://vimetric.com
Dominik Sumer on Twitter/X: https://twitter.com/dominiksumer
complexity.fm podcast: https://complexity.fm
Hosts: Bas Steins and Michal Martinka
Referenced: Plausible Analytics (mentioned as inspiration for open-source analytics tools)
Referenced: Carbon (mentioned as an early code-snippet sharing tool competitor to Snapify)
</itunes:summary><itunes:explicit>false</itunes:explicit><itunes:image href="https://media.complexity.fm/episodes/0003/episode-artwork.png"/></item><item><title><![CDATA[Python Is Not Enough: Why Pythonistas Are Falling in Love with Rust]]></title><description><![CDATA[📄 Show Notes

Python Is Not Enough: Why Pythonistas Are Falling in Love with Rust

In this episode of complexity.fm, hosts Bas Steins and Michal Martinka sit down with Bob Belderbos — Python veteran, educator, and coach with over 20 years of software development experience. Bob shares his journey from Sun Microsystems sysadmin to co-founding PyBites, to now coaching developers one-on-one and in cohorts on Python, agentic AI, and Rust. We dive deep into why the Python community is increasingly drawn to Rust, what AI is really doing to how developers learn, and whether vibe coding is a superpower or a trap.

⏱️ Topics Covered

0:00 - Introduction and Bob's background: from economics to Sun Microsystems to Python
3:30 - What makes Python a great language and where it falls short
8:00 - The Python vs. Ruby vs. Go vs. TypeScript debate
14:00 - Why Python tooling (Ruff, UV, Polars) is being rewritten in Rust
18:00 - Bob's Rust origin story: a PyCon talk that changed everything
24:00 - How Rust makes you a better Pythonista
28:00 - Type safety in Python: MyPy, Ty, and Pyrefly compared
33:00 - Why Rust is a natural fit for agentic AI systems
38:00 - Bob's coaching philosophy: project-based, high-stakes, real-world
44:00 - The payroll SaaS built in 12 weeks from scratch in Django
50:00 - How AI is changing the way developers learn (for better and worse)
56:00 - Vibe coding, attention span, and the danger of rubber-stamping PRs
1:02:00 - "Deslopification": helping developers clean up AI-generated mess
1:08:00 - Junior to senior, vibe coder to engineer: Bob's three coaching verticals

🌐 Resources & Links

complexity.fm podcast
https://complexity.fm

Bob Belderbos - coaching and courses
https://bobbelderbos.com

PyBites - Python training platform
https://pybit.es

Ruff - fast Python linter written in Rust
https://github.com/astral-sh/ruff

UV - Python package manager written in Rust
https://github.com/astral-sh/uv

PyO3 - Rust bindings for Python
https://pyo3.rs

Polars - DataFrame library with a Rust core
https://pola.rs

MyPy - static type checker for Python
https://mypy-lang.org

Ty - fast type checker by Astral
https://github.com/astral-sh/ty

Pyrefly - type checker by Meta
https://pyrefly.org

The Rust Programming Language (book)
https://doc.rust-lang.org/book

HTMX - hypermedia for the modern web
https://htmx.org

Datastar - reactive hypermedia framework
https://data-star.dev
]]></description><link>https://complexity.fm/e/2</link><guid isPermaLink="false">https://complexity.fm/e/2</guid><dc:creator><![CDATA[Bas Steins & Michal Martinka]]></dc:creator><pubDate>Wed, 24 Jun 2026 16:30:00 GMT</pubDate><enclosure url="https://media.complexity.fm/episodes/0002/audio.mp3" length="0" type="audio/mpeg"/><itunes:author>Bas Steins &amp; Michal Martinka</itunes:author><itunes:summary>📄 Show Notes

Python Is Not Enough: Why Pythonistas Are Falling in Love with Rust

In this episode of complexity.fm, hosts Bas Steins and Michal Martinka sit down with Bob Belderbos — Python veteran, educator, and coach with over 20 years of software development experience. Bob shares his journey from Sun Microsystems sysadmin to co-founding PyBites, to now coaching developers one-on-one and in cohorts on Python, agentic AI, and Rust. We dive deep into why the Python community is increasingly drawn to Rust, what AI is really doing to how developers learn, and whether vibe coding is a superpower or a trap.

⏱️ Topics Covered

0:00 - Introduction and Bob&apos;s background: from economics to Sun Microsystems to Python
3:30 - What makes Python a great language and where it falls short
8:00 - The Python vs. Ruby vs. Go vs. TypeScript debate
14:00 - Why Python tooling (Ruff, UV, Polars) is being rewritten in Rust
18:00 - Bob&apos;s Rust origin story: a PyCon talk that changed everything
24:00 - How Rust makes you a better Pythonista
28:00 - Type safety in Python: MyPy, Ty, and Pyrefly compared
33:00 - Why Rust is a natural fit for agentic AI systems
38:00 - Bob&apos;s coaching philosophy: project-based, high-stakes, real-world
44:00 - The payroll SaaS built in 12 weeks from scratch in Django
50:00 - How AI is changing the way developers learn (for better and worse)
56:00 - Vibe coding, attention span, and the danger of rubber-stamping PRs
1:02:00 - &quot;Deslopification&quot;: helping developers clean up AI-generated mess
1:08:00 - Junior to senior, vibe coder to engineer: Bob&apos;s three coaching verticals

🌐 Resources &amp; Links

complexity.fm podcast
https://complexity.fm

Bob Belderbos - coaching and courses
https://bobbelderbos.com

PyBites - Python training platform
https://pybit.es

Ruff - fast Python linter written in Rust
https://github.com/astral-sh/ruff

UV - Python package manager written in Rust
https://github.com/astral-sh/uv

PyO3 - Rust bindings for Python
https://pyo3.rs

Polars - DataFrame library with a Rust core
https://pola.rs

MyPy - static type checker for Python
https://mypy-lang.org

Ty - fast type checker by Astral
https://github.com/astral-sh/ty

Pyrefly - type checker by Meta
https://pyrefly.org

The Rust Programming Language (book)
https://doc.rust-lang.org/book

HTMX - hypermedia for the modern web
https://htmx.org

Datastar - reactive hypermedia framework
https://data-star.dev
</itunes:summary><itunes:explicit>false</itunes:explicit><itunes:image href="https://media.complexity.fm/episodes/0002/episode-artwork.png"/></item><item><title><![CDATA[Read-Only Mode: Why a Senior Dev Quit Writing AI Code]]></title><description><![CDATA[📄 Show Notes

In this first episode of complexity.fm, Bas and Michal sit down with AI consultant and educator Isaac Flath to get real about AI-assisted coding. Is it actually making senior developers faster? Is vibe coding dead on arrival for production systems? And why is memory still the unsolved problem at the heart of every AI product? No hype, no dogma — just three experienced practitioners comparing notes.

Isaac's AI Engineering Community on Skool
https://www.skool.com/ai-eng/about

Isaac Flath's website and blog
http://isaacflath.com

Next-PLAID retrieval benchmark (LightOn AI)
https://lightonai.github.io/next-plaid

Colgrep — contextual code search tool
https://github.com/colgrep (search "Colgrep" on GitHub)
]]></description><link>https://complexity.fm/e/1</link><guid isPermaLink="false">https://complexity.fm/e/1</guid><dc:creator><![CDATA[Bas Steins & Michal Martinka]]></dc:creator><pubDate>Tue, 09 Jun 2026 16:30:00 GMT</pubDate><enclosure url="https://media.complexity.fm/episodes/0001/audio.mp3" length="0" type="audio/mpeg"/><itunes:author>Bas Steins &amp; Michal Martinka</itunes:author><itunes:summary>📄 Show Notes

In this first episode of complexity.fm, Bas and Michal sit down with AI consultant and educator Isaac Flath to get real about AI-assisted coding. Is it actually making senior developers faster? Is vibe coding dead on arrival for production systems? And why is memory still the unsolved problem at the heart of every AI product? No hype, no dogma — just three experienced practitioners comparing notes.

Isaac&apos;s AI Engineering Community on Skool
https://www.skool.com/ai-eng/about

Isaac Flath&apos;s website and blog
http://isaacflath.com

Next-PLAID retrieval benchmark (LightOn AI)
https://lightonai.github.io/next-plaid

Colgrep — contextual code search tool
https://github.com/colgrep (search &quot;Colgrep&quot; on GitHub)
</itunes:summary><itunes:explicit>false</itunes:explicit><itunes:image href="https://media.complexity.fm/episodes/0001/episode-artwork.png"/></item></channel></rss>