Artificial Intelligence on Coderrob

Agentic Development and the Review-Centric Workflow

Thu, 14 May 2026 00:00:00 -0500

AI is going to change software teams whether we pretend to be ready or not.

The lazy version of that change is simple: give everyone an agent, tell them to go faster, and hope review catches whatever falls out the other side.

That is not a strategy.

The better version is this:

Move humans closer to design and review, then let agents help execute well-defined units of work.

That does not mean developers stop being technical. It means the team’s technical energy shifts toward architecture, decomposition, design patterns, refactoring, review, verification, and shared understanding.

From Agent to a Soul

Sun, 12 Apr 2026 10:00:00 -0500

I’ve been thinking about how we talk about AI systems, and Anthropic’s use of the word “soul” keeps sticking with me.

At first, that sounds too poetic for software.

Then you start layering personality, persistent memory, defined capabilities, curated tools, operating constraints, and a consistent communication style onto an agent.

At that point, “agent” starts to feel a little too small.

More Than a Task Runner

An agent executes.

A more carefully designed AI collaborator does something slightly different.

Fast Is Slow When You're Neck Deep in AI Slop

Fri, 10 Apr 2026 12:00:00 -0500

The Pattern Nobody Wants to Talk About

I’ve seen a distinct pattern where AI slows down software development.

I know. Heresy. But hear me out.

Agents are fast, but in the old saying kind of way:

Fast is slow, and slow is fast.

People push back on this. “Humans are faster,” they say.

Honestly? I doubt that’s true either.

What I believe is that most people are single-threaded. One brain, one task, one context.

Trust but Verify: The Math of Rigorous Software Development

Fri, 10 Apr 2026 10:00:00 -0500

Between Satire and Ridicule

It started, as these things often do, with a comparison nobody asked for.

I made a post saying that, with AI-assisted workflows, I can contribute anywhere from 20 to 300,000 lines of change in a day.

That number was not meant to say, “Look at me, I make ALL the tokens go brrrrrrrt…”

It was me counting additions plus deletions across real work: feature creation, refactoring, documentation, automation, test updates, generated assets, project scaffolding, cleanup, and the supporting files needed to make the work sustainable.

Announcing ESLint Zero-Tolerance AI Anti-Slop Rules

Tue, 10 Feb 2026 10:00:00 -0600

AI-assisted development has a trust problem.

Generated code can look polished, type-safe, and reviewable while still smuggling in shortcuts that quietly erode a codebase over time. The faster teams move with agents, the more those shortcuts compound.

ESLint Zero-Tolerance is my attempt to put hard boundaries around those patterns in TypeScript projects: not as abstract guidance, but as enforceable rules and presets that make low-trust code harder to ship.

The Announcement

I’ve published the ESLint Zero-Tolerance AI anti-slop rules here:

Reverse Engineering Agentic Workflows from Copilot Debug Logs

Fri, 24 Oct 2025 09:00:00 -0500

Here’s a secret weapon for building your own agentic workflows: GitHub Copilot Chat’s debug logs.

You know how everyone’s out there wrestling with hallucinating AI agents? Trying to figure out how to structure those prompts, which tools to call when, how to handle errors without pulling your hair out, what context to pass between steps…

The answer?

It’s sitting right there in your Copilot Chat debug view. Already solved. Already tested. Already proven to work for your specific use cases.

The UI of AI is CLI

Wed, 22 Oct 2025 05:00:00 -0500

We spent decades trying to make computers easier to use. We went from punch cards to command lines to graphical user interfaces to touch screens. We added buttons, menus, icons, gestures, or really anything to avoid making people type commands into a terminal.

And now? Now we’re instructing AI to use… command lines.

Plot twist of the century, right there.

The Great UI Circle of Life

Here’s the evolution in a nutshell:

Planning the Planning: The Agentic Software Development Paradox

Tue, 21 Oct 2025 10:00:00 -0500

You know what’s wild about working with AI agents to build software? The planning. Oh boy, the planning.

Not just a plan - that would be too simple. No, no. We’re talking about:

Creating a planning document
Planning the planning (because that first plan will encounter the enemy)
Planning the planning of the planning (we need to go deeper)
Having an agent review the plan to identify any planning not planned in the plan
Having another agent revise the plan after being told to plan the planning based on the planning and the current implementation

It’s like Inception, but instead of dreams within dreams, it’s plans within plans within plans. Plan-ception, if you will.

Building a Fully Local, Privacy-First AI Chat with Ollama and Open WebUI

Wed, 14 May 2025 05:14:13 -0500

Sometimes I want full control, not most, but total, end-to-end control over the tools I use to think, build, and create.

That’s where Ollama and Open WebUI comes in. This is my go-to local AI chat setup: a fully local, OpenAI-style interface that runs quietly on your machine without whispering a word to the cloud.

What I Use This For

This isn’t just an academic exercise in privacy. I use this:

When Pragmatism Meets Silence

Thu, 20 Mar 2025 03:35:16 -0500

Had one of those surreal conversations at work recently.

I needed to onboard with an internal AI service—because, frankly, there’s only one option available: and it’s run by a team building their own wrapper around Azure OpenAI.

From the start, it was a mess.

Their onboarding SharePoint was broken, documentation was a year out of date, and the API FAQ pointed to a “Coming Soon w/ RAG API FAQ” page hiding the currently available API docs. I posted in their support space about the issues and was redirected to another team to get help with their onboarding documentation.

Sora Text to Video: Playing with AI Like It's 2049

Sat, 15 Feb 2025 14:42:11 -0600

I finally spent some time experimenting with Sora, and the shift from text-to-image to text-to-video feels bigger than I expected.

I have used Stable Diffusion for a long time. Image generation already changed how I think about prompts, style, references, and iteration.

Video adds another layer.

You are not just describing what something looks like anymore.

You are describing what happens.

Prompting a Moment

With image generation, a prompt describes a frame.

Local LLMs as a Code Assistant in Visual Studio Code

Tue, 03 Dec 2024 10:00:17 -0600

I have been using a Visual Studio Code extension called Continue, and it has become one of my favorite ways to work with local LLMs.

The short version:

Continue lets you bring AI-assisted coding into VS Code while keeping the model, prompts, and workflow under your control.

That matters more than it sounds like it does.

Why Local Matters

Local models are not always the strongest models.

They are not always the fastest models.

Microsoft TinyTroupe for UI/UX Persona Focus Groups

Wed, 13 Nov 2024 08:38:07 -0600

TinyTroupe is incredible.

Imagine running your own focus group composed of multi-personality AI agents. From a UI/UX designer’s perspective, the ability to explicitly define user personas and have them evaluate your designs with constructive feedback.

Add vision agents into the mix… Just… wow.

Check it out:

https://github.com/microsoft/TinyTroupe

You Know What's Going to Be Funny When AI Goes Offline

Wed, 21 Aug 2024 21:23:44 -0500

You know what’s going to be funny? When AI services go offline and teachers finally get to see what a student’s real writing would be like. Imagine the shock—“Wait, why does this essay look like a Neanderthal cave drawing scribbled on paper?” It’ll be like watching someone try to cook without YouTube tutorials—pure chaos.

Today, I Introduced the Interns to Dogfooding

Wed, 21 Aug 2024 20:47:54 -0500

Today, I got to introduce the AI interns to new topics like dogfooding*, cyclomatic complexity, BEM, JSON:API, and why interfaces are so useful. It was great to nerd out with them! Their excitement was contagious, and I was impressed to learn their code review for the chatbot was their first time working with React and web development.

Seeing that “aha!” moment when they first create something—there’s nothing like it.

For you Gen Z and Alphas: “dogfooding” means using your own product, just like your users do, so you feel the same issues and can fix them. It’s like making sure the dog food you’re serving is good—because you’re eating it too**!

** Don’t eat dog food - unless it’s a biscuit, I guess, but you do you.

Embracing Multi-Model Approaches for Enhanced Workflow Efficiency

Fri, 16 Aug 2024 22:09:29 -0500

Leveraging multiple models can significantly streamline the software development process. Here’s a strategy to consider:

Requirement Analysis: Use one model to gather user requirements.
Edge Case Identification: Deploy another model to spot edge cases while drafting the requirements document.
Code Instruction: Utilize a code generation model to create detailed how-to guides.
Language-Specific Models: Pass the how-to guide to a model specialized in the relevant programming language.

By creating a self-referencing workflow of requests and responses, you can enhance accuracy and efficiency. Ensure your code includes both positive and negative unit tests, runs them, and identifies issues with self- corrections.

Stable Diffusion Explained Like I Am 5

Thu, 10 Nov 2022 15:09:17 -0600

Imagine “a world, earth, seen from space, 8k, unreal engine, detailed, photorealistic”…

Remember the common phrase, “a picture is worth a thousand words”? Well, it’s time to rethink that.

With advancements in image generation technology, we may need to start saying, “a few words are worth thousands of pictures.”

This technology is simply astonishing.

Any image. Any concept. Any quality. Any style. Near-instant results.