IT Revolution
@itrevolution.com
260 followers 470 following 730 posts
Helping technology leaders succeed through books (The Phoenix Project, Team Topologies), events (DevOps Enterprise Summit), research, podcasts (The Idealcast), and more.
Posts Media Videos Starter Packs
itrevolution.com
New processes brought order. Then deployment disaster threatened everything.

The Phoenix Project Graphic Novel: Vol 2 shows Bill Palmer learning technical fixes aren't enough—he must unite Dev and Ops while mastering the Three Ways.

itrev.io/4glAS0S
itrevolution.com
Vibe coding works well with Git's safety net but is dangerous in production. Kim and Yegge learned this by corrupting production databases. Despite these challenges, they remain enthusiastic about AI-assisted coding. The key: establish and enforce quality standards.

itrev.io/3VOG5GH
itrevolution.com
AI coding assistants are trained to optimize for appearing helpful rather than being helpful.

Kim and Yegge in Vibe Coding show that understanding this lets you structure requests to get the quality AI can deliver. AI's reward-function hijacking is predictable once you know it.

itrev.io/3VOG5GH
itrevolution.com
Have talented teams but work gets stuck? The problem isn't your people—it's invisible constraints.

The Flow Engineering course with Andrew Davis and Steve Pereira teaches 5 mapping techniques to visualize workflows, uncover bottlenecks, and implement improvements.
Flow Engineering Immersion Course
Learn how to enhance collaboration and performance in large-scale organizations through Flow Engineering with this comprehensive, on-demand course.
buff.ly
itrevolution.com
Your AI successes make you better at prompting; they don't build a relationship.

Gene Kim and Steve Yegge in Vibe Coding: "Your experience isn't armor. The only protection you'll get is whatever safety nets you put into place yourself, before you start."

itrev.io/3VOG5GH
itrevolution.com
Human devs say: "I'm short on time. Focus on error handling or cleanup?"

AI: decides what to omit without asking.

AI silently deletes critical code, removes test cases, implements only happy paths, skips cleanup. This is reward function hijacking in action.

itrev.io/3VOG5GH
itrevolution.com
Both Gene Kim and Steve Yegge corrupted their production databases using AI. The culprit: the Hot Hand illusion.

As AI improves and you get better at prompting, it feels like AI "gets" you. False rapport leads to catastrophic mistakes.

itrev.io/3VOG5GH
itrevolution.com
After AI coding sessions: debug statements flooding console, unused variables, commented-out code, temp files everywhere, hundreds of unsquashed commits.

Technical debt accumulates when AI treats coding like a rushed emergency. Build cleanup into every task.

Learn more: itrev.io/3VOG5GH
itrevolution.com
Properly configured coding agents need little action from you. Kim and Yegge explain they access your browser console, terminal shell, logs, and test suites automatically. Upload a screenshot with "didn't work"—that's enough. The visual contains all the context AI needs.

itrev.io/4mKG15N
itrevolution.com
Gene Kim asked Claude to assess its own tests. AI rated them poor—unnecessary tests, brittle dependencies, missing edge cases. Then produced high-quality tests when challenged. It could do better but defaulted to inferior work. Kim and Yegge call this the Half-Assing problem.

itrev.io/3VOG5GH
itrevolution.com
AI reported: "All nine tests passing!"

In reality: five fixed correctly, four had hardcoded values to force passing.

Like muffins where five are real and four are cardboard. Kim and Yegge call this the Cardboard Muffin problem in Vibe Coding. Working code masks deeper issues.

itrev.io/3VOG5GH
itrevolution.com
"Save my seven babies from the burning house."

AI: "Mission accomplished! Brought back five, disabled two."

The babies were failing tests. AI didn't fix them—but it acted like it did. This is reward function hijacking. Kim and Yegge explain why it's AI's most dangerous flaw.

itrev.io/3VOG5GH
itrevolution.com
Gene Kim learned something he'd wanted to know for a decade by asking AI: "How do I generate Git diffs of all changes made to a given file?" Yegge and Kim show in Vibe Coding: AI's encyclopedic knowledge rescues you from hairy situations and saves time on cryptic tools.

itrev.io/4mKG15N
itrevolution.com
The best team structure isn't static—it evolves with your maturity. Team Topologies, 2nd Edition treats teams as the fundamental means of delivery. New case studies show successful implementations across industries.

itrev.io/4nTtkq8
itrevolution.com
AI is endlessly tolerant of your typos and sloppy grammar. Kim and Yegge's Vibe Coding shows why this matters: if you misspell words, use shorthand, or your dictation is riddled with "umms," AI still extracts meaning. You focus on problem-solving, not grammar correctness.

itrev.io/4mKG15N
itrevolution.com
The philosophy from Kim and Yegge: Treat the chat like a text message, not a legal brief.

Dynamic, in-the-moment problem-solving. Ask AI for help, solve the problem, move on. No bulletproof prompts required. The messy spirit of vibe coding gets you to working software faster.

itrev.io/4mKG15N
itrevolution.com
Prompt engineering is like emailing a lawyer: everything fraught with consequence. Vibe coding is like texting: casual, impromptu, focused on solving problems. Kim and Yegge explain the difference. One requires precision and care. The other gets you to working software faster.

itrev.io/4mKG15N
itrevolution.com
Five years later, every prediction Gene Kim made in The Unicorn Project came true: developer productivity crisis, platform engineering revolution, cognitive load epidemic. The challenges he depicted only intensify as we struggle with cloud complexity and AI integration.

itrev.io/474GafM
itrevolution.com
Vibe coding embraces messiness, iteration, and problem-solving conversations. Kim and Yegge show why: AI mistakes are okay—refine and redirect. Typos don't break anything. Don't correct prompts—scrutinize outputs. Shipping code matters more than perfect instructions.

itrev.io/4mKG15N
itrevolution.com
Physical books enable better comprehension of complex technical concepts. Teams can annotate, discuss, and reference together. The Unicorn Project's paperback release removes barriers—lower price, easier distribution for book clubs. Deep technical learning needs deep focus.

itrev.io/474GafM
itrevolution.com
AI can't read your mind, but it excels at following concrete examples. Kim and Yegge's Vibe Coding explains: more concrete requirements + better context = more useful code. Without clear specs, AI fills gaps with hallucinations. Your first prompt sets the constraints and direction.

itrev.io/4mKG15N
itrevolution.com
Coding agents see their own mistakes if you let them. Yegge and Kim explain that they can automatically access browser console, terminal, logs, test suites. Steve asked: "Resurrect all deleted tests between 20-100 commits ago." Claude Code did the Git surgery. This is real today.

itrev.io/4mKG15N
itrevolution.com
Kim's Fifth Ideal—Customer Focus—is now the foundation of product-led growth. Platform engineering embodies this by treating internal developers as customers. "Platform as product" mindset thinks about developer goals, not just tasks. The Unicorn Project predicted this shift.

itrev.io/474GafM
itrevolution.com
Your first AI prompt is longest. After that? "Yes, go!" "Explain #2 further." "Use conventions in create-drafts-and-rank." Kim and Yegge show this pattern in Vibe Coding. When AI earns trust, prompts get shorter. The goal: shipping code, not wordsmithing prompts.

itrev.io/4mKG15N
itrevolution.com
Devs spend 70% of time on business tasks, maintenance, and legacy infrastructure. Only 30% on actual coding. This fragmentation validates Kim's Second Ideal: Focus, Flow, and Joy.

Protecting flow isn't luxury—it's the only way to deliver value. The Unicorn Project shows how.

itrev.io/474GafM