complexity.fm Podcast

Bas Steins & Michal Martinka

0.0 (0)
Technology
Updated Biweekly

Conversations with people in tech

5d ago

The Heroku Killer That Got Them Into YC

📄 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
Jul 7

Building in Public Is Dead (And AI Killed It)

📄 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)
Jun 24

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

📄 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
Jun 9

Read-Only Mode: Why a Senior Dev Quit Writing AI Code

📄 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)

4 Episodes

Conversations with people in tech

Creator

Bas Steins & Michal Martinka
Years Active

2026
Episodes

4
Rating

Clean
Copyright

© 2026ff. Bas Steins & Michal Martinka. All rights reserved.
Show Website

complexity.fm Podcast