ZD 25.11: The New Vibe at Work
The Vibe Work era starts now. Are you prepared?
In the Issue: Warning signs for the west. How to think about reliability and failures as accidental experiments. Guidance on which AI model to use in 2025. What is the vibe work era and how will it change your job?
The Distilled Spirit
Warning Signs
🤖 The West is Missing the Robot Wave (SemiAnalysis)
Dylan Patel shares a call to action to the US. China’s lead in manufacturing is translating to a growing lead in robotics and automation of labor. He follows with a detailed explanation of the technology involved and who is making what. If all he says is true the situation is dire indeed.
🔊 Bot Ridden Parallel Economy ()
The population is tired. Tariff and other general uncertainty induced market volatility are just adding to the pile. Legions of bots serve to inflame reactions, keeping people from focusing on the sorts of long-term, world building projects that powered pax Americana. Companies living in a parallel economy are setting up to take over for the collapsing media in this flooded information space.
🍏 Rotting Apples (Daring Fireball)
Apple missed on the delivery updated Siri with Apple Intelligence. John Gruber takes this as a sign that things might well be rotting inside of Apple. That they have squandered their credibility in some way.
Advisory Thought
🏛 Reliability Compounds (Scope of Work)
From TFA: ”You can’t touch reliability, or polish it. You can’t point to a single factor that’s made air travel safer, because reliability is fundamentally an emergent phenomenon. We don’t even truly know how it comes about, but it’s also absolutely inseparable from modern life – in aviation and elsewhere. We rely on water to come from our taps, we rely on supermarket shelves to be full, we rely on power to flow at the flip of a switch. We rely on the collected information of the internet to be in our pockets at all times. All this stuff just works.”
💔 Failures are Accidental Experiments (Jason Feifer)
Things failing are accidental experiments in many cases. Failures often include difficult to simulate conditions, unexpected inputs and new ideas. Always look to see what you can learn and take away from the experience.
💲 How to Find Exceptional Businesses ()
Buying stock in good companies and doing nothing is some of the best investing advice I ever got. Finding the good companies is harder. Oguz Erkan shares a solid eight point list of the kinds of companies that are exceptional and tend to be great do nothing investments.
Applied AI
✅ Which AI Model to Use in 2025 ()
Peter shares a really qualified guide to which AI platform to use today. The graphical checklist is the best feature, but the whole discussion is worth study. I don’t agree with the final conclusion — he prefers Claude over ChatGPT — but I think that is a matter of taste and what your primary uses are for your AI tool and how long your toolchain is in that platform.
🌟5 New AI Tools You Want to Try ()
Jeremy Caplan is back with another list of awesome AI tools worth trying. These are interesting examples of the new wave of much thicker, more capable tools than the relatively thin chatbots and API wrappers we have had before.
🏢 AI and the Corporate Pyramid ()
AI tools could completely reshape the corporate pyramid, spreading the powers outside of the traditional hierarchy as decision making gets delegated in a radically different way to thinking robots.
💼 Prompts are the New IP ()
As AI takes over the workplace, the prompts that guide it are fast becoming key business intellectual property. The more powerful agents we have today are able to consume much more instruction. Prompts are also much harder to secure than code or data, making it a challenge to keep this expertise in house.
Entertaining Interests
🥩 East Coast Pastrami Tour (John Tanner’s BBQ Blog)
There is lots of corned beef on sale post St. Patrick’s day to serve as an easy to start with base for your pastrami cook. Welcome to the season.
🏟 COSM is Like Being There (Yahoo Sports AM)
COSM is an attempt to combine the best of live sporting events with TV. It is a facilities-based approach that features an 87-foot, 12k resolution curved screen that creates a mini-Sphere like experience. They work with venues to capture immersive video onsite using svelte cameras that can get nearly into the action to make it feel like you are there.
💯100 Best Sports Moments of the Century (The Ringer)
The ringer has pulled together a list of the best sports experiences of the last 25 years into one top 100 list. Great format — including a section on explaining the moment to a non sports fan.
🏀 The Daily Cinderella ()
Rodger Sherman wanted an excuse to stream all of the basketball he could, so he is writing a daily post on college basketball for the foreseeable future. It is everything you need to get your head into March Madness in one handy substack.
The Vibe Work Era
NEW FEATURE! 🎶 AI Generated Musical Coda: The Vibe Work Era. 🎶
Artificial Intelligence has demonstrated enormous potential over the last few years. We can do amazing things cutting and pasting things in and out of chatbots. Chat style interfaces are fun, easy to use, and very powerful, but also fundamentally limiting because their effectiveness diminishes as tasks become more complex, require integration with external data or software, or involve continuous updates without manual intervention. Having to copy and paste things in and out of them does not quite scale. There is a little too much cargo handling and context switching for most people.
That is starting to change dramatically. The first wave of this has been tools in your chatbots. Claude Artifacts and ChatGPT Canvas create a surface that functions as a real web server or document viewer you can truly work through, at least at a small scale. In your office suite, Microsoft Copilot attempts to provide advice. Google does a better job with Gemini in G-Suite. These sorts of tools show the way — where rather than moving work from tool into and out of chatbot you can communicate with the content and have chatbot do the work.
In software development this is moving very quickly. Code has always been a bit easier for LLMs than language — it is much less ambiguous and much better documented. GitHub Copilot was one of the first AI products integrated into a real-world application. It was quite useful, though I never got much more out of it than I did out of ReSharper in the last decade. It was a pretty fancy autocomplete, it could not build your application from whole cloth even if it helped you code better.
OpenAI co-founder Andrej Karpathy coined the term Vibe Coding. It describes building software not by sitting there writing code in your IDE, but by directing your AI agents to build software in your IDE in a constant feedback loop. You vibe through creation rather than struggle through syntax errors. Tools like Cursor, Bolt, Replit, and Claude Code are truly changing the way software development works. It is enabling a new wave of non-developers to build things. It makes experienced developers extremely powerful as they can mechanically apply their skills at an exponential rate.
I sat down with Cursor this week to try and build a few applications. It was an amazing and eye-opening experience. I was truly skeptical of this sort of tool, after all my years of experience in software development have told me not to trust do it all tools. The resulting codebases that were hard to modify or update, with lots of extremely verbose code and unnecessary baggage. As a guy who rarely got it right on the first shot, I saw no other way than typing one character at a time, albeit with lots of Intellisense to help me out.
I should have believed the hype. Advances in language models and tooling have come a long way. Claude or ChatGPT could make a small script. Running it in Artifacts or Canvas was useful but not deployable and fundamentally limited in scale. Cursor and its peers are on a different level. They create deployable applications, in some cases on your local disk where software development is sane. The truly impressive thing is how well it works with existing code. I took the kernel of an app I had been struggling with improving and blitz scaled it up to deployable in a few hours of typing. Building out tests and having the tool create the code to pass them was as sweet a TDD experience as I’ve ever had. This old developer is sold on the concept.
Computer programming is at the leading edge of this trend. The tooling was already meant for automation. The users tend to be a bit tech savvy and love to experiment. The major AI labs are working on making this possible for the rest of the world. Anthropic, makers of Claude, created Model Context Protocol to enable seamless communication between AI platforms and software applications. This week, OpenAI announced their AgentsSDK to enable similar functionality for other tasks. Combined with tools like Deep Research, you can achieve production-level output through coaching and sharing rather than direct research and content creation. Vibe coding is going to become vibe writing, accounting, and researching.
If you were an accountant in 1980, VisiCalc rocked your world. The magic of spreadsheet software like VisiCalc—being able to change one number and have all the other numbers update automatically, provided your formulas were programmed correctly—saved weeks of manual recalculation. There are stories of firms being hired for jobs that had once taken a month, doing all the work in 3 days and spending the next three weeks on the golf course. Over the next decade, the bookkeeping and accounting world changed dramatically. Easy data entry into computers replaced many human bookkeepers, while demand for the now more capable accounting services increased. The ability to data model turned accountants from human calculators to strategic advisors.
We are probably at a moment not too unlike the early 1980s. Are you ready for the vibe work era?
The Look
America and the west are falling far behind on robot installations. Learn more about this dire situation on SemiAnanlysis.
Did you enjoy reading this post? Hit the ♥ button above or below because it helps more people discover great Substacks like this one and it helps train your algorithm to get you more posts you like. Please share here or in your networks to help us grow!
