Given a pool of very productive, sharp and useful code churning programmers, this one would NOT be on top of the list. A range of coding topics, mostly around .Net C#.
🚨 Don’t stop Vibe Coding… just understand the other options.
Vibe Coding isn’t the problem — it is simply misused.
In this video, I break down:
⚡ Vibe Coding vs Spec-Driven Development
⚡ The **AI Coding Spectrum**
⚡ When each approach actually works (and when it doesn’t)
If you’re using AI tools like Copilot or ChatGPT, this is something you NEED to understand 👇
Ever wanted to use **AI models in VS Code that aren’t available in Copilot?** 🤔
In this video, I show how to connect **OpenRouter** with the **Roo Code extension** so you can run alternative AI models directly inside **VS Code** — including **Qwen 3.6 Plus** and **DeepSeek R1**.
With just **two simple steps**, you can unlock dozens of AI models using a single API key and experiment with different coding assistants to see which works best for your workflow.
💡 **What you’ll learn:**
• How to create a **free OpenRouter API key**
• Installing and configuring the **Roo Code extension**
• Selecting different AI models in VS Code
• Testing **Qwen 3.6 Plus** and **DeepSeek R1**
• Monitoring **usage and cost** in the OpenRouter dashboard
If you're experimenting with **AI coding tools, alternative LLMs, or AI-assisted development**, this setup is one of the easiest ways to start.
Automating Development with Copilot Agent Skills in Visual Studio 2026**
Did you know **GitHub Copilot in Visual Studio 2026** now supports **Agent Skills**?
Instead of repeatedly writing prompts for AI, you can capture a complex development process once and reuse it anytime. In this video, I show how a simple **Skills.md file** can automate something as complex as scaffolding an entire **.NET Clean Architecture solution**.
In the demo you'll see how Copilot can:
• Create a solution
• Generate multiple projects
• Add references and packages
• Organize solution folders automatically
All from a single prompt.
If you're interested in **AI-driven development workflows** or improving **developer productivity with Copilot**, this is a feature you’ll want to understand.
I ran an experiment with **GitHub Copilot in Visual Studio 2026** to see if AI could create a realistic **Scrum/Kanban project plan** for rewriting an old codebase.
The idea was simple:
1️⃣ Give Copilot the **legacy code**
2️⃣ Ask it to **review the code**
3️⃣ Then generate **Sprint plans, effort estimates, and timelines**
But there were some constraints:
* Only **1 developer**
* **5 hours per day** available
* **2-week sprints** (with 7 actual dev days)
* **No historical sprint velocity data**
### What happened?
* **ChatGPT 5.1 Codex Mini** produced a plan that looked detailed… but was basically **waterfall disguised as agile**.
* Testing and documentation were pushed to the **end of the project** — a classic **Agile anti-pattern**.
Then I tried the **full ChatGPT 5.1 Codex** model.
The output looked much smarter:
✔️ Scrum terminology
✔️ Definition of Done / Ready
✔️ Structured sprint backlog
But when you actually analyze the plan…
⚠️ The estimates were unrealistic
⚠️ Most early sprints were **repetitive entity conversion work**
⚠️ The **real application logic** was largely missing
### The Reality
Sprint planning is difficult even for experienced teams.
Normally it depends on:
* **Sprint velocity**
* **Historical team data**
* **Consensus estimates (planning poker)**
* **Deep knowledge of the codebase**
AI simply **doesn’t have that context**.
### Final Verdict
AI can generate **convincing looking plans**, but realistic sprint planning still requires **human judgment and domain understanding**.
🚨 **New Video — OpenSpec Spec-Driven Development FAILED**
I ran a real experiment using **OpenSpec Spec-Driven Development** while building a **car classifieds web application** with AI coding tools.
The goal: redesign a **.NET Razor Pages front-end** into a **premium modern UI**.
But something unexpected happened.
Instead of improving development speed, the **Spec-Driven workflow consumed hours of developer time and a large number of AI tokens** — while producing results very close to the original design.
So I tried something much simpler:
➡️ **Instructions.md**
Just clear instructions for the AI agent — no framework, no proposals, no spec reviews.
The result?
⚡ Faster progress
💰 Far fewer tokens used
🧠 Much less overhead
This raises a serious question for developers experimenting with **AI coding frameworks**:
**Do we really need complex Spec-Driven workflows… or can simple instruction-based approaches work better in practice?**
I break down the full experiment and results in this video.
⚠️ **I Tried Spec-Driven Development… and the reality surprised me.**
Everyone says Spec Driven Development is the *“correct way”* to build applications with AI.
But after actually using frameworks like **OpenSpec** and comparing them with **Vibe Coding**, I noticed some things that people don’t really talk about.
Things like:
• The **huge number of .md files** generated
• The **time required to review specs**
• The **AI model being the real bottleneck**
• And the **hidden cost in tokens and workflow overhead**
In this video I share my **honest first impressions** and explain where Spec Driven Development works well — and where it might not.
• A simple “review this code” prompt = ❌ misleading results
• AI tends to be *too nice* and misses serious issues
• Without structure, scores get inflated
But when I introduced a **custom AI review scorecard**…
💥 The same project scored **15/100 — Technical Bankruptcy**
That changed everything.
⚠️ Key insight:
AI code reviews are only as good as the **framework you give them**.
In this video, I break down:
✔️ Reviewing a real legacy codebase
✔️ How to structure AI reviews properly
✔️ Comparing legacy vs modern rewritten code
✔️ Why AI results are NOT deterministic
If you're using Copilot for serious development, this is something you need to understand.
👇 Question for you:
Have you tried using AI for code reviews yet? What was your experience?
Can AI generate a full **.NET Data Access Layer** from an existing database schema using **Spec Driven Development**?
In **Episode 3 of the Car Classifieds project**, we use **OpenSpec + GPT Codex in VS Code** to generate domain entities and Entity Framework configurations using a **database-first approach**.
This episode also includes an **honest look at when Spec Driven Development may be overkill** and where AI automation might actually be faster.
Topics covered:
• Proposal → Design → Tasks workflow
• Generating domain entities from schema
• Entity Framework configuration
• Clean Architecture DTO patterns
• Working with GPT Codex as an AI coding agent
If you're interested in **AI-assisted .NET development, OpenSpec, and Spec Driven Development**, this episode shows how the workflow actually looks in practice. ⚙️
🚀 New Video: **Spec-Driven Development Explained | AI Coding Workflow**
As AI tools become a bigger part of software development, the question is no longer *whether* to use AI — but **how to use it effectively on real projects**.
In this video I explain **Spec-Driven Development**, a structured approach where AI generates code from specifications instead of simple prompts.
Topics covered:
• Why **vibe coding** works for small tasks but struggles with larger systems
• How **structured specifications** give AI better context
• The typical **spec-driven workflow**
• Popular frameworks like **OpenSpec, GitHub Spec Kit, BMAD, Kiro, and Tessl**
If you're experimenting with **AI-assisted software development**, this approach might become increasingly important as projects get more complex.
Incomplete Developer
AI can now make music in minutes… no skills needed.
So I tested Suno AI as a developer.
Honestly?
It feels like vibe coding for music.
⚡ Fast
⚡ Impressive
⚠️ But not quite “real”
Are musicians in trouble… or is this just hype?
Watch here 👇
https://youtu.be/HZRmY4siCzA
👇 Be honest:
Is AI music the future… or just noise?
#aimusi #sunoai
1 month ago | [YT] | 0
View 0 replies
Incomplete Developer
🚨 Don’t stop Vibe Coding… just understand the other options.
Vibe Coding isn’t the problem — it is simply misused.
In this video, I break down:
⚡ Vibe Coding vs Spec-Driven Development
⚡ The **AI Coding Spectrum**
⚡ When each approach actually works (and when it doesn’t)
If you’re using AI tools like Copilot or ChatGPT, this is something you NEED to understand 👇
🎥 Watch here: [https://youtu.be/d_QXKM5snzo](https://youtu.be/d_QXKM5snzo)
💡 The truth?
It’s not about choosing one method…
It’s about knowing **when to use each one**.
2 months ago | [YT] | 0
View 0 replies
Incomplete Developer
Ever wanted to use **AI models in VS Code that aren’t available in Copilot?** 🤔
In this video, I show how to connect **OpenRouter** with the **Roo Code extension** so you can run alternative AI models directly inside **VS Code** — including **Qwen 3.6 Plus** and **DeepSeek R1**.
With just **two simple steps**, you can unlock dozens of AI models using a single API key and experiment with different coding assistants to see which works best for your workflow.
🎥 **Watch the full video here:**
[https://youtu.be/dYqiE0WJsYY](https://youtu.be/dYqiE0WJsYY)
💡 **What you’ll learn:**
• How to create a **free OpenRouter API key**
• Installing and configuring the **Roo Code extension**
• Selecting different AI models in VS Code
• Testing **Qwen 3.6 Plus** and **DeepSeek R1**
• Monitoring **usage and cost** in the OpenRouter dashboard
If you're experimenting with **AI coding tools, alternative LLMs, or AI-assisted development**, this setup is one of the easiest ways to start.
2 months ago | [YT] | 0
View 0 replies
Incomplete Developer
Automating Development with Copilot Agent Skills in Visual Studio 2026**
Did you know **GitHub Copilot in Visual Studio 2026** now supports **Agent Skills**?
Instead of repeatedly writing prompts for AI, you can capture a complex development process once and reuse it anytime. In this video, I show how a simple **Skills.md file** can automate something as complex as scaffolding an entire **.NET Clean Architecture solution**.
In the demo you'll see how Copilot can:
• Create a solution
• Generate multiple projects
• Add references and packages
• Organize solution folders automatically
All from a single prompt.
If you're interested in **AI-driven development workflows** or improving **developer productivity with Copilot**, this is a feature you’ll want to understand.
▶️ Watch the full video here:
[https://youtu.be/wm_LdAVNYpA](https://youtu.be/wm_LdAVNYpA)
Let me know in the comments:
**What development tasks would you turn into an Agent Skill?**
2 months ago | [YT] | 0
View 0 replies
Incomplete Developer
🚧 **Can AI Really Do Sprint Planning?** 🤖📅
I ran an experiment with **GitHub Copilot in Visual Studio 2026** to see if AI could create a realistic **Scrum/Kanban project plan** for rewriting an old codebase.
The idea was simple:
1️⃣ Give Copilot the **legacy code**
2️⃣ Ask it to **review the code**
3️⃣ Then generate **Sprint plans, effort estimates, and timelines**
But there were some constraints:
* Only **1 developer**
* **5 hours per day** available
* **2-week sprints** (with 7 actual dev days)
* **No historical sprint velocity data**
### What happened?
* **ChatGPT 5.1 Codex Mini** produced a plan that looked detailed… but was basically **waterfall disguised as agile**.
* Testing and documentation were pushed to the **end of the project** — a classic **Agile anti-pattern**.
Then I tried the **full ChatGPT 5.1 Codex** model.
The output looked much smarter:
✔️ Scrum terminology
✔️ Definition of Done / Ready
✔️ Structured sprint backlog
But when you actually analyze the plan…
⚠️ The estimates were unrealistic
⚠️ Most early sprints were **repetitive entity conversion work**
⚠️ The **real application logic** was largely missing
### The Reality
Sprint planning is difficult even for experienced teams.
Normally it depends on:
* **Sprint velocity**
* **Historical team data**
* **Consensus estimates (planning poker)**
* **Deep knowledge of the codebase**
AI simply **doesn’t have that context**.
### Final Verdict
AI can generate **convincing looking plans**, but realistic sprint planning still requires **human judgment and domain understanding**.
👉 Watch the full experiment:
https://youtu.be/ErwuATHHXw4](https://youtu.be/ErwuATHHXw4
3 months ago | [YT] | 0
View 0 replies
Incomplete Developer
🚨 **New Video — OpenSpec Spec-Driven Development FAILED**
I ran a real experiment using **OpenSpec Spec-Driven Development** while building a **car classifieds web application** with AI coding tools.
The goal: redesign a **.NET Razor Pages front-end** into a **premium modern UI**.
But something unexpected happened.
Instead of improving development speed, the **Spec-Driven workflow consumed hours of developer time and a large number of AI tokens** — while producing results very close to the original design.
So I tried something much simpler:
➡️ **Instructions.md**
Just clear instructions for the AI agent — no framework, no proposals, no spec reviews.
The result?
⚡ Faster progress
💰 Far fewer tokens used
🧠 Much less overhead
This raises a serious question for developers experimenting with **AI coding frameworks**:
**Do we really need complex Spec-Driven workflows… or can simple instruction-based approaches work better in practice?**
I break down the full experiment and results in this video.
▶ Watch here:
[https://youtu.be/dnqLfT3waEI](https://youtu.be/dnqLfT3waEI)
If you're experimenting with **AI agents, GitHub Copilot, Codex, or AI-assisted development**, this might save you a lot of time (and tokens).
#openspec #specdrivendevelopment #aicoding
3 months ago | [YT] | 0
View 0 replies
Incomplete Developer
⚠️ **I Tried Spec-Driven Development… and the reality surprised me.**
Everyone says Spec Driven Development is the *“correct way”* to build applications with AI.
But after actually using frameworks like **OpenSpec** and comparing them with **Vibe Coding**, I noticed some things that people don’t really talk about.
Things like:
• The **huge number of .md files** generated
• The **time required to review specs**
• The **AI model being the real bottleneck**
• And the **hidden cost in tokens and workflow overhead**
In this video I share my **honest first impressions** and explain where Spec Driven Development works well — and where it might not.
📺 **Watch the full video here:**
[https://youtu.be/wgV-tEYLEts](https://youtu.be/wgV-tEYLEts)
💬 **Curious what developers think:**
**Would you use Spec Driven Development for real projects, or stick with Vibe Coding?**
Vote below and share your experience 👇
👍 Spec Driven Development
🔥 Vibe Coding
🤔 A hybrid approach
3 months ago | [YT] | 0
View 0 replies
Incomplete Developer
🚨 Can AI actually do a *real* code review… or is it just being polite?
I put **GitHub Copilot in Visual Studio 2026** to the test on a **legacy .NET app from 2013** — and the results were surprising.
👉 Watch here: [https://youtu.be/omDvFGu8Vtc](https://youtu.be/omDvFGu8Vtc)
💡 Here’s what I found:
• A simple “review this code” prompt = ❌ misleading results
• AI tends to be *too nice* and misses serious issues
• Without structure, scores get inflated
But when I introduced a **custom AI review scorecard**…
💥 The same project scored **15/100 — Technical Bankruptcy**
That changed everything.
⚠️ Key insight:
AI code reviews are only as good as the **framework you give them**.
In this video, I break down:
✔️ Reviewing a real legacy codebase
✔️ How to structure AI reviews properly
✔️ Comparing legacy vs modern rewritten code
✔️ Why AI results are NOT deterministic
If you're using Copilot for serious development, this is something you need to understand.
👇 Question for you:
Have you tried using AI for code reviews yet? What was your experience?
3 months ago | [YT] | 0
View 0 replies
Incomplete Developer
🚗 **New Video in the OpenSpec Series!**
Can AI generate a full **.NET Data Access Layer** from an existing database schema using **Spec Driven Development**?
In **Episode 3 of the Car Classifieds project**, we use **OpenSpec + GPT Codex in VS Code** to generate domain entities and Entity Framework configurations using a **database-first approach**.
This episode also includes an **honest look at when Spec Driven Development may be overkill** and where AI automation might actually be faster.
▶ Watch the video:
https://youtu.be/2OckOvSGI2Q
📺 Full series playlist:
https://www.youtube.com/watch?v=2OckO...
Topics covered:
• Proposal → Design → Tasks workflow
• Generating domain entities from schema
• Entity Framework configuration
• Clean Architecture DTO patterns
• Working with GPT Codex as an AI coding agent
If you're interested in **AI-assisted .NET development, OpenSpec, and Spec Driven Development**, this episode shows how the workflow actually looks in practice. ⚙️
3 months ago | [YT] | 0
View 0 replies
Incomplete Developer
🚀 New Video: **Spec-Driven Development Explained | AI Coding Workflow**
As AI tools become a bigger part of software development, the question is no longer *whether* to use AI — but **how to use it effectively on real projects**.
In this video I explain **Spec-Driven Development**, a structured approach where AI generates code from specifications instead of simple prompts.
Topics covered:
• Why **vibe coding** works for small tasks but struggles with larger systems
• How **structured specifications** give AI better context
• The typical **spec-driven workflow**
• Popular frameworks like **OpenSpec, GitHub Spec Kit, BMAD, Kiro, and Tessl**
If you're experimenting with **AI-assisted software development**, this approach might become increasingly important as projects get more complex.
Watch here 👇
https://youtu.be/0atkW_janVg
Curious to hear how others are using AI in development.
Are you mostly **vibe coding**, or trying more structured workflows?
3 months ago | [YT] | 0
View 0 replies
Load more