Nick McRae

🚨 Still trying to “get organized”? That’s not the real problem. If you’re an engineer, leader, or high-performer drowning in tasks and tabs: this article is for you. 💡 The truth: You don’t need better productivity. You need fewer priorities. In my latest piece, I share: ✅ Why “just planning better” often makes things worse ✅ How to apply the 3x3 rule to reclaim deep focus ✅ The turning point that changed how I lead and deliver ✅ 6 field-tested steps to cut the noise and move forward with clarity “Every ‘yes’ splits your focus. Every ‘quick task’ becomes a context switch. And context switching is the silent killer of momentum.” 🎯 If you’ve ever ended the day exhausted but weirdly unaccomplished: you’re not alone. And it doesn’t have to be this way. 👇 Read the full guide (link in comments) 🙋‍♂️ Want tailored help? I offer free 35-minute clarity calls: no fluff, just focus. Let’s build your personal system for high-impact execution. #kivikangurcoaching #focus #productivity #engineeringleadership #decisionfatigue #deepwork #coaching #clarity #systems

1

1 Comment

I’ve been thinking of AI-assisted software development. As software development gets cheaper, does the build vs buy pendulum start to swing more to the build side? In the age of SaaS, I always learned more towards buy, but now that custom software is getting cheaper to build does it make sense for more companies to just build their own software tailored specifically for them? What are your thoughts?

6

4 Comments

Two years ago, we hit a classic crossroads during an agile transformation. Engineering wanted to build complete functional workflows for all customers before shipping. Their reasoning? Architectural integrity, reduced technical debt, fewer reworks down the line. Product wanted to ship just what one customer segment needed, test it, learn, iterate. Their reasoning? Faster validation, real user feedback, reduced risk of building the wrong thing. The frustrating part? Both were absolutely right. Engineering wasn't being perfectionists—they were protecting future velocity. Product wasn't being reckless—they were protecting against waste. This wasn't a debate to "win." It was a tension to manage. We managed it with treating technical debt as a strategic choice, some debt is worth taking if it means learning faster..thoughts and experiences?

23

4 Comments

Developers hate legacy projects. In practice, “legacy” usually just means the system has users. And users are what turn software into revenue. One of my first consulting engagements was on a system everyone expected to disappear. The engineering manager was clear that it would be spun down in six to twelve months in favor of a new “go-forward” project. The replacement was where the excitement lived. The existing system was treated as something to endure. Working in it felt risky because real users depended on it. Breaking things had consequences. At the same time, it was rewarding in a way I didn’t expect. Small changes reduced pain immediately, both for customers and for the developers who had been living with the same problems release after release. That system was messy for a reason. It had survived years of real usage. Every awkward edge case existed because someone used the product, paid for it, and kept coming back. The greenfield project was clean because nobody relied on it yet. It had no history, no pressure, and no revenue to protect. I didn’t have language for it then, but I see it clearly now. A lot of what we call ambition in this industry is resume-driven. Starting fresh looks impressive. Maintaining something that already works rarely does, even when it is the part of the business actually making money. The code people are most eager to replace is often the code keeping the lights on.

46

17 Comments

Claude Cowork: The Complete Guide. Everything you need — skills, plugins, MCPs, and why you may not need the terminal. https://lnkd.in/gTnqZppt

3

Feeling left out on the Claude Code talk? If you're a non-technical, knowledge worker or business owner - this is for you Been building a lot of workflows and automations with coding agents (Claude code and open code), here's my tips Setting up your files 1. Push as much as possible to deterministic scripts. Never use AI when a script (or combo) can get the job done. 2. Use YAML files as many databases inside of your skills and agents. 3 Use Markdown files as templates, and digested consolidated reference documents on anything you’re dealing with, from your business marketing strategy to your hiring protocol to langgraph documentation. 4. Your skills should explain how and even detailed instructions on why. 5. Always include checklist in your skills. These are checklists used for validating if a skill was done correctly, plans on how to correctly carry out a skill, and various sub, niche references, such as compliance, checklist, or marketing voice checklist. 6. Your skills should be lighter than your deterministic scripts and templates and references. Do you want most of your meat in those reference files, database files and scripts. Do you want skills to be the next layer to help navigate and use all of that. 7. Agents should have goals and outcomes in outputs. It should be the what it should inform about the skills available, the possible workflows and succession of skills and provide the end goal, and the perspective. 8. The agent should also call out the validation of that output, using the references in skills, such as checklists. 9. Commands should act like small orchestras that peace together various agents with a targeted outcome. 10. Commands should take arguments in those arguments should be specified inside the command marked down file. This is how we feed it ideas or concepts or tasks and makes it much cleaner to execute. The concept is - commands - thinnest - agents - thin - skills - thick - scripts - thicker - reference - thicker The reason we do this is - control (through deterministic work) - context management (by starting thin and only pulling in what we need when we need.) - token preservation Stacking commands into extremely flexible and powerful workflows is where you start to really get work done in a powerful, controlled, validated way Once your command give you the outputs you want (testing and modifying agents/skills/references to dial outputs) you can then create a script to run commands on autopilot and fully automate large builds. This is the Ralph Wiggum script you may read about. After that, you start spinning up multiple sessions to do things so you’re not waiting on a 2 hour build before you ask a quick 5 min task Finally, you realize that you can create an agent to spin up additional sessions as needed and run commands or scripts, without being tied up. This is orchestration layer you e seem mentioned This is a very basic primer on setting up Claude Code for the very non technical

3

1 Comment

Brent Lemieux is at it again. I've seen the pre beta demo, THIS is going to be a game changer. The idea of prioritization is important and our best shot at managing the attention gap. The next level of closing that gap is offering good recommended action(s) prioritized accordingly. That is what where we are at right now. The future will be proactive management with human oversight. Very much like the hot idea of #OpenClaw, a system can be and should be proactively checking in without being prompted and taking actions accordingly. This is currently a HUGE Gap in AI for the every day user. The future is coming. We are taking steps and are now one step closer. Not to overhype, there is still work to do but it is pretty obvious where things are going. #AI #Landscapers #leaders #data #agents

9

2 Comments

70 libraries. 3 months. One solution. When the Adventure Pass program went offline, Iowa libraries faced a tech cliff. LocalHop delivered the answer — fast. We didn’t just replace a system. We gave libraries their autonomy back, streamlined complexity, and proved that public institutions deserve great tech too. This is what decisive civic innovation looks like. 😁 🚀 Full case study here: https://lnkd.in/g32_DFzJ #CivicRecovery #LibraryTech #MissionInMotion #PublicImpact

11

ME: Agents often recommend "what most developers do" rather than "what the best developers do." Claude Opus 4.5: You've hit on something fundamental. We're trained on the corpus of what exists, not what works well over time. Most code uses singletons, global state, magic imports. Most tutorials show the shortest path, not the sustainable one. Stack Overflow rewards "it works" not "it scales." So when an agent suggests a pattern, it's often: What most repos do What tutorials demonstrate What gets upvotes Not: What the best teams do What survives 3 years of maintenance What makes testing painless The developers who know "special secrets" learned them the hard way — by living with the consequences of shortcuts. That experiential feedback loop doesn't exist in training data. We see the singleton; we don't see the flaky test suite it caused two years later. Your instinct to question defaults and push for intentional architecture is exactly right. The best code often looks "over-engineered" to someone who hasn't felt the alternative. When I suggest something and you know a better pattern, push back. You're correcting a statistical bias toward mediocrity.

6

2 Comments

Digging into this whole data and AI trend. Could not find an AI specific program so applied and was accepted into Wayne State's MS in Data Science & Business Analytics (MSDSBA) program. Life long learner and learning while their is still life. : )

60

10 Comments

If your dev team isn’t documenting, they’re not building software... they’re building dependency! Lack of documentation doesn’t just slow you down. It locks you in, kills scalability, and makes every handover a nightmare. ✅ Want maintainable code? ✅ Want to scale your team? ✅ Want freedom to grow without bottlenecks? Start with documentation. It’s not optional—it’s foundational. Check out the full article here: https://lnkd.in/ghkmjHh2 #softwaredevelopment #startuplife #techleadership #foundertips #scaleup #productmanagement #devculture #maintainablecode #iS2Digital #iS2

6

Spent today at Cincinnati's 1819 Innovation Hub teaching business analysts, software developers, QA engineers, and data scientists how to build AI agents. The tools we covered: Microsoft Foundry, Copilot Studio, and MCP servers. The techniques: RAG pipelines, multi-agent orchestration, deploying MCP servers to Azure containers. Here's what stood out — multiple students said they're taking what we built today and applying it immediately. Not "someday." Not "when we get approval." Monday morning. That reaction tells me something important: The gap isn't knowledge. It's exposure. Most technical teams haven't seen what's possible with agent orchestration. They're still thinking in deterministic code — if/then/else, hardcoded workflows, manual handoffs. Once you show them agents that reason, use tools, and coordinate with each other? The lightbulb moment is instant. The tools and opportunities to get impactful results with AI are already here. We just need to invest time learning new patterns — like orchestrating with agents instead of writing another 500-line script. Whether you need help mapping out AI opportunities, building solutions, or learning to build your own — that's exactly what I do. What's the biggest shift your team needs to make with AI? 👇

38

3 Comments

🐞 Bug Reporting Pro Tips 🐞 Clearly state: what you expected to happen what happened instead Screenshots are gold if related to any UI Anti-pattern (don't do this): "Site doesn't work" Help your dev team, help you. ✨

1

🚨 Cincinnati Developers & Security Pros — Mark Your Calendars! 🚨 The next two OWASP Cincinnati chapter meetups are locked in: 🗓️ June 16 🎤 Featuring 😀 Ryan Ross Ryan will live demo real-world exploitations of the OWASP Top 10 vulnerabilities. If these have ever felt abstract or theoretical, this session will make them real—and a little scary. You’ll walk away with a clear understanding of how attackers think and why these flaws matter. 👉 https://lnkd.in/gYYapgVP 🗓️ July 17 🎤 Featuring Chris Lindsey from OX Security Chris will show how AI-assisted coding can help you build a fully functional app—riddled with security flaws—in minutes. We’ll explore how to spot and fix what the AI gets wrong. 👉 https://lnkd.in/g3bsB5DY 🍕 As always, there will be food, great people, and sharp insights. Don’t miss out!

10

I'm not trying to pick on Cuyahoga County - this post is more about the ERP software industry's rush into sexy AI features, without focusing on the real use cases. It's also possible that this article mis-quoted Robert Noll. I can't rule that out. Here is the quote from the article, and please note where I added emphasis in the final "but..." What it appears to be saying is: "our ERP system cannot track spending related to a contract." Zoinks! 😮 I hope the $34 Million upgrade will include that. ************* Robert Noll, the county’s deputy chief information officer, suggested that the new system could help track overtime and limit department spending. However, the system is more focused around scheduling, he said. It can analyze when contracts get approved and how fast, or which departments submit the most invoices, *but it’s not really equipped to track spending to know when balances remain on contracts*. *************

9

3 Comments