Posts by Viktor Farcic
The bottleneck moved. Code is now cheap and abundant. AI generates it faster than any human. The scarce resource? Architecture, context, verification, judgment.
youtu.be/sDy55DIkIJ0
The gap between a generic coding assistant and an AI agent that truly understands YOUR company is enormous.
Here's the complete blueprint for building custom AI agents: system context, tools, knowledge retrieval, security, and cost optimization.
youtu.be/98yYcWwD95I
Fixed replicas are always wrong. You're either wasting resources or under-provisioned.
Here's how to build Kubernetes autoscaling that goes from 0→5 replicas based on actual traffic, using tools you probably already have.
youtu.be/K0aaB2L3sXI
Built AI inference platform using Kubernetes + Crossplane.
One custom resource = any team can deploy a model.
No GPU expertise required. Your data never leaves your network.
Full walkthrough:
youtu.be/odZ4WclCX5s
"Just tell the AI how we do things" doesn't scale. Tribal knowledge fails. Copy-pasting docs fails. In this video, I show how a vector database + embeddings + MCP server gives your coding agent real company context—automatically. No more fixing AI output by hand.
youtu.be/P-Sv6tKPeLQ
Today's AMA session is likely to be violent. There might be yelling, a broken nose or two, blood being spilled, ... We'll have two guests, bitter rivals, each one thinking that the other one is terribly wrong. Wars have been started for less.
youtube.com/live/dd9pAta...
Traditional UIs assume predictable data. AI outputs are never predictable.
So who designs the UI when you don't know what the data looks like until it appears?
I explored two approaches — and why the future of UIs might not be designed by humans at all.
🎥👇
youtu.be/czffWAXTy2Y
"Just use a Mac Studio!" Sure—it'll run a trillion-parameter model about 100x slower than H100s. A Mac Mini can't even fit it in memory. Stop romanticizing self-hosting and look at the actual numbers. New video is live. 🍎💀
https://youtu.be/pWtDTkfNaUU
Commands, MCP tools, Skills, Subagents, Hooks, Plugins, Memories... 🤯
Every week there's a new way to extend coding agents.
Let's cut through the noise and explain what each actually does, how they work under the hood, and which ones you should actually use.
https://youtu.be/xAIN7YHXfCY
Me: "Find all databases in the cluster"
kubectl: *struggles through 356 API definitions*
Vector DB + AI: "Found PostgreSQL StatefulSet, CloudNativePG cluster, Qdrant, and an AWS RDS instance"
This is how K8s querying should work. New video shows how 👇
https://youtu.be/uUdbQkq5c4k
I got addicted to AI. Now I use it to dub my YouTube videos into other languages automatically. Built a CLI that handles everything - upload, dubbing, translation, scheduling. Here's how I integrated ElevenLabs API into my workflow 🎬
https://youtu.be/_tvKc9kKrhA
AI agents are unpredictable by design. The LLM decides the path, chooses tools, loops as needed.
You can't debug what you can't see.
Good news: OpenTelemetry tracing works for AI agents too. New video explains how.
https://youtu.be/Dh7jAYHcAlo
I wrote a simple prompt. Port's MCP provided the context. Claude Code did the rest. Result? A fully configured developer portal with K8s integration, GitHub sync, and self-service actions. AI + the right context = 🔥
https://youtu.be/6r6ITHCYZ3c
Most Dockerfiles in production are security nightmares: running as root, using :latest tags, copying secrets. Writing a good Dockerfile isn't hard, you just need to know the rules. Here's every best practice you need for production-ready containers 🐳
https://youtu.be/ueTe-VQaD7c
Your API is a product. Who's using it? Are they abusing it? Should they even have access? New video on API Management: developer portals, analytics, versioning, monetization with Zuplo review 😤
https://youtu.be/mOy2mhKE2_8