▸ Latest Episode
#15 Agentic Engineering
Mar 13, 2026 · 01:01:57
In this episode, Toby and Krisztian share their hands-on experiment building a real group management app using Claude Code and agentic engineering. Toby spent roughly a month's worth of hours prompting Claude to build a cross-platform mobile and web app with Expo, a Node/Express API, Postgres on Scaleway, Hanko authentication and Terraform infrastructure — all without looking at the code. They discuss what worked surprisingly well, what fell apart, the token costs, how agentic engineering compares to managing juniors, and what they would do differently next time. https://techleaguepodcast.com/ LinkedIn: https://www.linkedin.com/company/techleague-podcast/ Instagram: https://www.instagram.com/techleaguepodcast/ Apple Podcasts: https://podcasts.apple.com/us/podcast/tech-league/id1852602975 Spotify: https://open.spotify.com/show/1zx8UIe2EjawuVU3I31fdP Chapters 0:00 Introduction: the experiment 1:15 Why Expo for cross-platform mobile and web 2:54 The architect document approach 5:50 How the initial prompt and brainstorming worked 7:45 Not looking at the code 9:05 Context7: giving Claude access to latest API docs 10:35 First pass results 12:46 API and database quality: better than expected 14:40 UI issues: the weak spot 17:20 The bug list testing session 24:23 How sub-agents and parallel work played out 26:07 The Claude usage limit dark pattern 28:15 What it cost: 247 euros for roughly 100 hours of work 31:30 Code quality and lessons from the spec 38:48 The testing problem: agents writing tests for their own code 45:20 The 80/20 rule: great at the fun stuff, weak on the boring 50:30 SaaS disruption: custom software at commodity prices 57:25 How to build LLM memory and learning loops 1:00:25 Summary and what we would do differently Technologies Mentioned Claude Code - https://claude.ai/code Expo - https://expo.dev Scaleway - https://www.scaleway.com Hanko - https://hanko.io Terraform - https://www.terraform.io Playwright - https://playwright.dev Context7 - https://context7.com Node.js - https://nodejs.org PostgreSQL - https://www.postgresql.org
View show notes →Clips
AI Code: Brilliant Logic, Hilarious Flaws 🤣
SaaS is Dead? AI Builds Your Perfect Tool Instantly.
The AI terms changed FAST. See the new 'Agentic Engineering'!
The sweet smell of SaaS death? 🤯 Custom code changes everything.
Can Europe ditch *all* US tech? Sovereignty debate heats up!
AWS's 'Sovereign Cloud' - Is it real or just a clever facade?
Recent Episodes
#14
EU Cloud Alternatives - Scaleway
In this episode, Toby and Krisztian kick off an ongoing series exploring the EU cloud and software stack. Following a previous episode on EU digital sovereignty, they have set themselves a challenge: build their side projects entirely on EU-based services. This episode covers hands-on experience with Forgejo for source code management, Scaleway as a cloud provider, and Hanko for authentication. They share honest feedback on what works, what doesn't, and where the gaps are compared to the big American hyperscalers. https://techleaguepodcast.com/ LinkedIn: https://www.linkedin.com/company/techleague-podcast/ Instagram: https://www.instagram.com/techleaguepodcast/ Apple Podcasts: https://podcasts.apple.com/us/podcast/tech-league/id1852602975 Spotify: https://open.spotify.com/show/1zx8UIe2EjawuVU3I31fdP Chapters 0:00 Introduction and the EU sovereignty challenge 1:48 Finding EU alternatives: european-alternatives.eu 3:47 Looking for a GitHub replacement 5:50 Forgejo: the open-source Gitea fork 8:27 What works in Forgejo and what doesn't 10:46 Hosting Forgejo on Scaleway 13:48 The gap between self-hosting and a managed service 15:18 Scaleway overview: regions, services and Terraform support 20:35 Scaleway serverless functions and containers 25:02 Service-to-service authentication 28:34 Deploying Forgejo, databases and runners on Scaleway 36:04 Logging, metrics and Cockpit observability 40:27 Scaleway regions: Amsterdam, Paris, Warsaw 42:25 IAM limitations and enterprise considerations 44:14 Hanko: EU-native user authentication 48:32 Comparing EU stack total cost vs AWS plus Datadog 50:05 What's next: OVH, Hetzner, Infomaniak Technologies Mentioned - EU alternatives: https://european-alternatives.eu - Codeberg: https://codeberg.org - Forgejo: https://forgejo.org - Scaleway: https://www.scaleway.com/en/ - OVHcloud: https://www.ovhcloud.com/en/ - UpCloud: https://upcloud.com - Hetzner: https://www.hetzner.com - Elastx: https://elastx.se/en - Hanko: https://www.hanko.io
Feb 27, 2026
00:53:45
#13
OpenClaw
In this episode, Toby is joined by Xavier (Zavi) for a relaxed conversation about OpenClaw, an open-source project that lets you build a personalised, memory-aware AI assistant running on your own hardware. They share hands-on experiences setting it up with Telegram, Claude and local models, and discuss what makes it feel different from a standard chat interface: persistent memory in markdown files, heartbeat schedules, proactive check-ins, and a soul file that shapes personality over time. The conversation also covers security, prompt injection risks, the skill ecosystem, local model options, and the cultural questions around long-running AI companions. https://techleaguepodcast.com/ LinkedIn: https://www.linkedin.com/company/techleague-podcast/ Instagram: https://www.instagram.com/techleaguepodcast/ Apple Podcasts: https://podcasts.apple.com/us/podcast/tech-league/id1852602975 Spotify: https://open.spotify.com/show/1zx8UIe2EjawuVU3I31fdP Chapters 0:00 Introduction 0:33 What is OpenClaw? 1:40 Why does it feel different from a standard AI chat? 3:51 Setting it up: first impressions 4:45 Practical use cases: standups, workshop manuals, tractor parts 7:04 How the heartbeat and memory systems work 9:15 Cron jobs, proactive tasks and the soul file 12:06 The internals: TypeScript, service daemon, CLI and web UI 14:23 Security model: token auth, Tailscale, least-privilege access 17:42 Prompt injection risks 21:30 The skill ecosystem and supply chain risks 28:25 Local model support and failover between providers 32:55 Running local models: gaming laptops, Apple Silicon, VRAM 38:35 Different bot instances developing different personalities 41:45 Long-running AI companions and what they mean for society 44:55 Manipulation risk and the corporate AI companion future 48:15 Practical advice: what to give it access to, and what not to Technologies Mentioned OpenClaw - https://openclaw.dev Claude (Anthropic) - https://www.anthropic.com Telegram - https://telegram.org Ollama - https://ollama.ai Tailscale - https://tailscale.com
Feb 20, 2026
00:52:50
#12
DevOps and SecOps
In this episode, Toby and Krisztian take a deep look at DevOps and SecOps: where the ideas came from, what they were supposed to mean, how they got warped by the industry, and what good looks like in practice. They cover the waterfall origins of ops as a separate team, the shift-left movement, the build-it-you-run-it principle, why DevOps as a job title makes no sense, platform engineering, and how security is going through the same transformation. They also cover common anti-patterns, DORA metrics, how to get buy-in for a transformation, and what it looks like when it works at scale. https://techleaguepodcast.com/ LinkedIn: https://www.linkedin.com/company/techleague-podcast/ Instagram: https://www.instagram.com/techleaguepodcast/ Apple Podcasts: https://podcasts.apple.com/us/podcast/tech-league/id1852602975 Spotify: https://open.spotify.com/show/1zx8UIe2EjawuVU3I31fdP Chapters 0:00 Introduction 0:28 What DevOps was actually supposed to mean 1:57 The waterfall origins: why ops and dev were separate 5:45 Full stack and the rise of the developer-operator 8:40 Why the old model produced poor software quality 11:04 The move to agile and SaaS changed everything 14:15 DevOps as a term: what went wrong 16:08 Platform engineering: the natural next step 21:00 Breaking down the dev vs ops cultural divide 25:47 Real-world example: 10x performance improvement through shared ownership 30:29 Security is going through the same transformation 32:49 Shifting security left: from IDE to CI/CD pipeline 37:02 Reachability scanning and avoiding false positives 40:25 The strangler pattern for security posture improvement 43:34 SecOps as enablers, not gatekeepers 45:34 Common DevOps anti-patterns 53:48 Four-eyes principle done right vs done as Jira ping-pong 1:00:00 DORA metrics: how to measure if your DevOps is working 1:05:39 Management buy-in: why it matters and why it's hard 1:11:43 Real transformation stories 1:20:00 Internal platforms and giving teams real autonomy Technologies Mentioned Kubernetes - https://kubernetes.io AWS - https://aws.amazon.com Grafana Cloud - https://grafana.com/products/cloud Checkov - https://www.checkov.io GitHub Actions - https://github.com/features/actions
Feb 13, 2026
01:25:49
#11
How to build great engineering organizations
In this episode, Toby and Krisztian dive into how to structure and scale engineering organisations effectively. Drawing on years of consulting experience, they cover autonomous teams, domain-driven ownership, reducing cross-team handovers, internal platform teams, Conway's Law, the dangers of gatekeeping in ops and security, why self-service tooling is non-negotiable, and what it looks like when organisations are run like a portfolio of internal startups. A practical guide for engineering leaders and anyone building out an eng org. https://techleaguepodcast.com/ LinkedIn: https://www.linkedin.com/company/techleague-podcast/ Instagram: https://www.instagram.com/techleaguepodcast/ Apple Podcasts: https://podcasts.apple.com/us/podcast/tech-league/id1852602975 Spotify: https://open.spotify.com/show/1zx8UIe2EjawuVU3I31fdP Chapters 0:00 Introduction 1:43 When does an organisation become the bottleneck? 3:39 Starting with the problem space: divide and conquer 6:22 Autonomous teams and moving away from top-down command 8:03 How to detect misalignment: count the handovers 9:15 Conway's Law: use it intentionally 12:27 Single ownership and full accountability per domain 14:32 Internal service teams: when to spin one up 17:09 Each department as its own startup 19:57 Hero syndrome and knowing what not to build in-house 25:13 Self-service tooling: make it so good they choose it 28:33 KPIs, review cadences and cost visibility 36:24 Common anti-patterns: top-down command, founders who don't let go 41:42 Internal tooling teams as natural monopolies 45:26 The operations and security gatekeeper trap 48:20 Shifting from gatekeeper to enabler 53:02 Why developers must own production 57:34 How to set cross-divisional standards 1:07:09 Good internal platforms embed standards in golden paths 1:14:29 Entrepreneurial mindset within organisations 1:18:45 Summary and closing thoughts Technologies Mentioned AWS EKS - https://aws.amazon.com/eks Apache Airflow - https://airflow.apache.org Terraform - https://www.terraform.io Kubernetes - https://kubernetes.io
Feb 6, 2026
01:21:31