RSSAmplifier

Blog

Tiberriver256

tiberriver256.github.ioRSS feed ↗20 posts

Latest posts

Stop Reviewing Lint (AI Agents & Theory of Constraints)

Part of a series on getting more done with AI coding agents, using the 5 Focusing Steps from Theory of Constraints. Here’s the situation. You have AI coding agents. They write code for you. You want to ship more of it, faster. The obvious move: make the agents faster. Run more of them in parallel. That move is wrong. A 40-year-old idea from manufacturing explains why. A recipe from a factory floor…

Why I Prefer Stable Hours Over 'Get Your Work Done' in Software Dev

I like stable, predictable work hours. I learned that when you try to hook your self-worth to “commitments” or “getting your work done” in an environment like software development where work is so difficult to estimate your schedule can quickly become chaos. How I got stuck… I started out treating software development like the work I had laid out for me when I worked at Staples (the office supply…

Approvals Are a Security Wet Blanket for AI Agents

I was running an AI agent in a remote VM, exposed only via HTTP through a Node web server so I could send it commands. The agent asked to kill all Node processes. I approved it and immediately lost access to the VM. In hindsight, obvious. In the moment, I was deep in a task and the approval prompt looked like every other one I’d clicked that hour. Approval fatigue is the real problem Approval…

Do Newer Models Hold Up as Context Fills?

My experience has always been that AI models get pretty dumb after about 50% of their context window is filled. @vikhyatk on x.com challenged my assumption and said it’s been getting better. I decided to put on a little test. The below experiment was fully conducted and documented by gpt-5.2-codex and the codex CLI. We used the copilot cli for the sonnet 4.5 part of the experiment. Experiment…

Sharing Skills: Claude Marketplace to VS Code Workspaces

Part 3 of 3 in the Skills Catalog series ( Part 1 Part 2 ) You’ve built a skills catalog. Nested domains. Scripts that execute on demand. Now the final question: How do you share this across your team? Anthropic distributes skills through their marketplace and Skills API—portable, reusable knowledge packages. GitHub Copilot doesn’t have a marketplace. But VS Code workspaces achieve the same…

Learning from Anthropic: Nested Skills and Scripts

Part 2 of 3 in the Skills Catalog series ( Part 1 Part 3 ) Part 1 covered the catalog pattern. Now: how do you organize knowledge that scales to hundreds of skills? Anthropic’s skills repository shows patterns for complex domains: Nested structures for complex domains Scripts as executable knowledge Supporting resources (templates, configs) Anthropic’s Nested Skills The mcp-builder skill :…

Bringing Anthropic Skills to GitHub Copilot

Part 1 of 3 in the Skills Catalog series ( Part 2 Part 3 ) Anthropic just announced Skills —modular, discoverable knowledge for Claude. Each skill is a SKILL.md file with metadata. Claude discovers them through the description field. No central catalog, no manual wiring. Can we bring this to GitHub Copilot? Yes—with one key adaptation: a catalog table that replaces Claude’s auto-discovery. What…

Why Use AI CLIs When GUIs Exist?

People keep asking me why they’d ever use an AI command-line tool when the GUI works fine. The answer: CLIs let you mix deterministic scripting with non-deterministic AI judgment calls. Pure AI is hard to steer. Pure scripting breaks on edge cases. CLI tools blend both. The Difference in Action Here’s the example I show: copilot -p "Write me three reports on Bob, Dave, and Larry" versus "Bob" ,…

VC-Funded AI Coming to a Close

The era of artificially cheap AI tools is ending, and Cursor’s recent pricing changes are just the beginning. For over two years, venture capital has subsidized AI tools, keeping prices below true costs while companies built market share. The Cursor team’s recent pricing model overhaul signals this honeymoon period is ending. What changed at Cursor: Previous: Unlimited usage with throttling after…

Taming Your Copilot: A Practical Guide to Precision AI in VS Code

Guest post authored by Jules . Based on an audio transcript of Micah yammering to himself about this topic. GitHub Copilot. It’s in your VS Code, whispering sweet code suggestions. Sometimes it’s a mind-reading marvel. Other times? It’s like trying to explain astrophysics to a Golden Retriever – well-intentioned, but ultimately… fuzzy. The truth is, Copilot’s raw power, like any sophisticated…

ChatGPT needs some humility

It’s hot today in Michigan. High temps. High humidity. We turned down the thermostat a little bit, the AC kicked on, stayed on and didn’t really seem to be able to move the internal temperature. Several hours later I was curious… So… I opened up my Beestat app and gave ChatGPT a screenshot of the system graph and asked it to calculate my degrees per hour. It spun for about 4 minutes, running…

I got my GitHub Copilot Certification

How I studied Used a community-owned practice exam Followed the Microsoft Learn modules Who should probably take this exam? People administering a GitHub Copilot rollout at their organization. My thoughts As a user of GitHub copilot, I thought the courses on prompt engineering and actual usage were pretty good but minimal. I’ve learned a lot more about GitHub copilot reading through the GitHub…

An interesting learning with SignGemma and the deaf community

Google announced a new tool called SignGemma which they describe as “our most capable model for translating sign language into spoken text” I absent mindedly shared this in the GR Slack group I hang out in and a friend encouraged me to dig a little deeper into learning about how this new tool is received within the deaf community. The interesting piece for me to hunt for was why this tool wouldn’t…

Getting the related work items for an Azure Pipeline run in a monorepo

The problem.. I absolutely love this little ‘related work items’ feature in Azure DevOps pipelines: You can click on it and quickly get a view of the work items associated with that pipeline: This information is beneficial in two ways: It is useful information for the person running a deployment It can be used via ADO APIs to block deployments if related stories are in a certain status So… what’s…

The quest for an ergonomic keyboard - Part 1

If you didn’t read my previous blog post I’m hunting for a good method for keeping my tech experiments going from my living room lazy boy. One of the key components to that is having a bluetooth ergonomic keyboard Why Bluetooth? If I’m going to be popping up a little workspace here and there all over the house, I want setup time and teardown time to be seconds. I also don’t want anything dangling…

Blogging from Android

I love to write blog posts and tinker with technology. I haven’t been able to do this for awhile for several reasons: I have two autistic boys that require constant attention when they are awake I have severe pain in my hands that prevents me from typing on laptop keyboards My home office is in my youngest child’s bedroom which keeps me from blogging when my kids are sleeping Needless to say… I’ve…

ThingWorx local development environment with Docker - Part 1

ThingWorx is a platform for developing “Industrial IoT solutions”. They’ve got a lot of interesting documentation on various setup guides but I wanted to document how to get one up and running locally in a few minutes for the express purposes of learning and experimentation. A local setup doesn’t usually need all the heavy duty resilience and redundancy requirements that come with setting up a…

Estimate, target, plan and commit

I’ve started into “ Software Estimation: Demystifying the Black Art ” by Steve McConnell, and of course… as is my style. I’m making some conjectures about what I think software development should look like based off of reading the first chapter. I am going to finish the book, but jotting some quick thoughts down as I go helps me to digest the book as I go, so don’t be surprised if I actually…

How to run multiple versions of React side-by-side using Single Spa

This seems like it should be easy right? Single-spa is a micro-frontend framework designed to handle running multiple SPA frameworks side-by-side. It’s pretty straightforward in all the other frameworks… but not so in React. Let’s dig into why that is. What’s the problem? In order to use React you always have to install two packages: react - The main runtime for React Some other package that…

Stages of ramp up on software development projects

This is a list I’m going to be maintaining on different ideas and thougths around ramping up on new software projects. Grasp of existing problems One component of ramping up on a new project is understanding existing problems. You’ll never find a project without problems. These problems come in all shapes and sizes. They might be code quality issues, they might be slow running tests, they might be…