Vibe Coding vs Paper Coding

Forget fancy AI or modern vibe coding tools. My journey started with something much more physical: “Painful Paper Coding.” My very first program was born on a stack of yellow punch cards.

Long before the cloud, we had the giant mainframe. To get these huge machines to do anything, I had to follow a strange old ritual…

  • Step one: Buy a stack of blank cards. They were cheap – about 25 cents for 50 tickets to total frustration.
  • Or, I could “borrow” a few cards from a friend or a rival lab when no one was looking ๐Ÿ˜Ž
  • Next, find a free keypunch machine. I had to type out my code line by slow, painful line.
  • The concluding step involved delivering my stack of cards to the data center’s small window, where the “high priests” (the operators) would process them through the massive computer (IBM S/360).

After waiting 15 minutes or so, Iโ€™d get a big printout, find one tiny typo, and have to start the whole nightmare all over again.

Still, I loved every minute of it.

I loved the noisy machines and the massive computer. I even enjoyed the careful planning and flowcharting I had to do before punching a single card.

I enjoyed the challenge of writing efficient code to save money and time, and I loved the feeling of being in total control of my code.

Even with today’s smart tools and easy coding, I still miss that feeling.

Reading Over AI Summaries

๐—ช๐—ต๐˜† ๐—œโ€™๐—บ ๐—–๐—ต๐—ผ๐—ผ๐˜€๐—ถ๐—ป๐—ด ๐——๐—ฒ๐—ฒ๐—ฝ ๐—ฅ๐—ฒ๐—ฎ๐—ฑ๐—ถ๐—ป๐—ด ๐—ข๐˜ƒ๐—ฒ๐—ฟ “๐—”๐—œ ๐—ฆ๐˜‚๐—บ๐—บ๐—ฎ๐—ฟ๐—ถ๐—ฒ๐˜€”

We are all drowning in tasks, and time for actual reading is shrinking by the second. With the rise of AI, itโ€™s tempting to just skim an AI-generated summary and call it a day.

But Iโ€™ve realized the real value of a book isn’t just the “data” – itโ€™s the ๐—บ๐—ฒ๐—ป๐˜๐—ฎ๐—น ๐˜€๐—ฝ๐—ฎ๐—ฐ๐—ฒ it creates. Itโ€™s about understanding the context, the nuance, and the “why” behind the “what.”

Over the Easter break, I finally finished “๐—–๐—ผ-๐—œ๐—ป๐˜๐—ฒ๐—น๐—น๐—ถ๐—ด๐—ฒ๐—ป๐—ฐ๐—ฒ: ๐—Ÿ๐—ถ๐˜ƒ๐—ถ๐—ป๐—ด ๐—ฎ๐—ป๐—ฑ ๐—ช๐—ผ๐—ฟ๐—ธ๐—ถ๐—ป๐—ด ๐˜„๐—ถ๐˜๐—ต ๐—”๐—œ” by ๐—˜๐˜๐—ต๐—ฎ๐—ป ๐— ๐—ผ๐—น๐—น๐—ถ๐—ฐ๐—ธ.

What makes this book stand out? Mollick isnโ€™t a pure “tech guy.” He approaches AI from a userโ€™s perspective, making his insights incredibly practical and grounded.

๐—ง๐—ต๐—ฒ ๐Ÿฐ ๐—š๐—ผ๐—น๐—ฑ๐—ฒ๐—ป ๐—ฅ๐˜‚๐—น๐—ฒ๐˜€ ๐—ผ๐—ณ ๐—”๐—œ

If you want to master AI, Mollick suggests these four principles:

โ€ข Always Invite AI to the Table: Use it for everything to see where it shines (and where it fails).
โ€ข Be the “Human in the Loop”: AI is a co-pilot; you are still the captain responsible for the final output.
โ€ข Treat AI Like a Person: (A very smart, slightly weird intern). Give it context, feedback, and clear instructions.
โ€ข Assume This Is the Worst AI You Will Ever Use: The technology is the “weakest” it will ever be right now. Imagine what’s coming next.

๐—ข๐—ป๐—ฒ ๐—ง๐—ผ๐—ผ๐—น, ๐— ๐—ฎ๐—ป๐˜† ๐—ฃ๐—ฒ๐—ฟ๐˜€๐—ผ๐—ป๐—ฎ๐˜€

The book dives deep into how AI shifts roles depending on your needs. It can be your:

โ€ข Creative Partner for brainstorming.
โ€ข Coworker for heavy lifting.
โ€ข Tutor for learning new skills.
โ€ข Coach for personal growth.

๐—ง๐—ต๐—ฒ ๐—•๐—ผ๐˜๐˜๐—ผ๐—บ ๐—Ÿ๐—ถ๐—ป๐—ฒ: AI shouldn’t just replace our thinking; it should augment it. If you have the chance, put down the summary and pick up the book. Your brain will thank you for the “empty space.”

Link to the book: https://amzn.to/4tvk1j7

#AI #CoIntelligence #EthanMollick #ContinuousLearning #FutureOfWork #DeepWork

Human In the Loop, Or Not?

“Human in the loop” (HITL) is the most overused phrase in AI today. Itโ€™s also the most misunderstood.

In many workflows, the human has become a mechanical bottleneck – a “rubber stamp” meant to click ‘Approve’ or ‘Next’ without truly engaging. This isn’t just a waste of talent; itโ€™s a recipe for mediocrity.

In 2026, we donโ€™t just need a human in the loop. We need an Expert Human in the Loop.

The difference?
โ€ข HITL (Mechanical):ย Checking for typos or formatting. Approving output because it “looks right.”
โ€ข EHITL (Expert):ย Challenging the AIโ€™s logic. Applying domain-specific nuance. Spotting the subtle hallucinations that only a pro with 10+ years of experience can see.

AI can give us the 80% in seconds. But that final 20% – the part that actually moves the needle – requires us to apply our expertise toย the AI, not just afterย it.

Donโ€™t just check the AIโ€™s homework. Teach it how to think.

#AI #FutureOfWork #Expertise #HumanCentricAI #WorkflowInnovation #DigitalTransformation

Everything Is Different

“Isn’t it funny how day by day nothing changes, but when you look back, everything is different?” – C.S. Lewis.

Looking back at the trajectory of AI from 2024 to 2026, the transformation is staggering. We’ve moved from simple chatbots to sophisticated Multi-Agent systems. We are no longer just “using AI”; we are orchestrating it.

Welcome to the ‘Agent Manager’ era, where the most valuable skill is no longer just ‘prompting’ – it’s ‘intent engineering.’

To thrive, we must adapt:

โ€ข From Chatbots to Agents: GPT-5.4 and Claude 4.6 now execute complex workflows autonomously.
โ€ข The Multi-Agent Norm: Orchestration (via OpenClaw) mirrors high-functioning human teams.
โ€ข Human Edge: Our value lies in judgment, curation, and defining clear goals.

How are you evolving for the Agent Manager era?

#AI #FutureOfWork #AgenticWorkflows #TechTrends2026 #Innovation #DigitalTransformation

Avoiding AI Failure

Charlie Munger famously said: “All I want to know is where I’m going to die, so I’ll never go there.”

When it comes to implementing AI, most leaders are desperately trying to figure out how to be brilliant. They want to disrupt the market, revolutionize their workflows, and implement the most bleeding-edge models available.

But Mungerโ€™s philosophy of inversion suggests a much more practical starting point. Instead of asking how to win, ask: How do we guarantee this AI project will completely fail?

If your goal is to kill an AI initiative, here is the exact playbook:

  • Start with the tech, not the problem: Buy an expensive enterprise LLM license first, then wander around the company looking for a vague use case to justify the cost.
  • Ignore your data foundation: Assume the AI will magically untangle years of undocumented, siloed, and messy legacy data. Garbage in, garbage out – at scale.
  • Remove the human immediately: Automate a high-stakes, customer-facing workflow end-to-end on day one without a “human-in-the-loop” to catch the inevitable edge cases or hallucinations.
  • Skip change management: Drop a powerful new AI tool on your employees’ desks without training them on how to use it, or adjusting their KPIs to reflect their new workflows.

In the rush to adopt AI, the tech world is obsessed with seeking brilliance. But in complex systems, avoiding stupidity is often the faster path to ROI.

Don’t ask how AI is going to make your company a genius. Figure out what will cause the implementation to die – and then just don’t go there.

#ArtificialIntelligence #TechLeadership #AIStrategy