It’s interesting times for technology and the “non-tech person”- someone who never wanted to be involved with AI or any kind of digital undertaking in particular, who now finds themself surrounded by all sorts of tools requiring little to no prior understanding for their use.
People talk about the “democratization of tech,” with everything that comes along with it. There’s the opportunity for people with no coding background to create a codebase or app, although articles like this one point to potential security snafus.
Covering some of the pitfalls of DIY vibe coding for Tech Times, Richard Wells cites a Stanford study, writing:
“Developers using AI coding tools not only wrote less secure code than those who did not — they simultaneously reported higher confidence in its security. The trust paradox has practical consequences for non-developers, who have no engineering baseline from which to calibrate that misplaced confidence.”
In Which The Original Vibe Coder Provides a Caveat
Even Andrej Karpathy, commonly considered the godfather of vibe coding, has shifted stance a little bit.
Here’s his original post from Feb. 2 of last year:
“There’s a new kind of coding I call ‘vibe coding’, where you fully give in to the vibes, embrace exponentials, and forget that the code even exists. It’s possible because the LLMs (e.g. Cursor Composer w Sonnet) are getting too good.”
Having had a year to think about it, Karpathy is now referring to “agentic engineering” as a revised outlook.
“The goal is to claim the leverage from the use of agents but without any compromise on the quality of the software,” he writes. “Many people have tried to come up with a better name for this to differentiate it from vibe coding, personally my current favorite ‘agentic engineering’: – ‘agentic’ because the new default is that you are not writing the code directly 99% of the time, you are orchestrating agents who do and acting as oversight. – ‘engineering’ to emphasize that there is an art and science and expertise to it. It’s something you can learn and become better at, with its own depth of a different kind.”
More on the Future of Agentic Engineering
I saw a panel at this year’s Imagination in Action event at MIT this April where a panel discussed this phenomenon.
Maria Gorskikh, who is connected to MIT partly through her work on the NANDA protocol for agentic internet, interviewed Ryan Meadows of Lovable, Christopher Harris of Blitzy, and Serge Vasylechko of the Harvard Medical School about how this type of AI has evolved, and what to do with it now.
Gorskikh and Meadows talked about how Lovable appeals to a non-technical user base.
“Non-technical folks can actually build something,” Meadows said.
Blitzy, on the other hand, is designed for enterprise. Harris explained how the project’s founders developed an ideology around model ensembles, and how one of them pioneered the technology toward a coherent intended use:
“He realized that you could create the ability for agents to dynamically recruit other agents, and stitch context across agents that could create a knowledge graph of the entire codebase, and that solves the fundamental problem that we see with a lot of these enterprise usages, which is the limited context window.”
What emerged, he said, is “a whole new way of engineering.”
“It’s a bigger sort of transformation,” he said.
The Debugging Process
Gorskikh asked Harris about what happens if an AI creates a codebase quickly, but it turns out to need weeks’ worth of debugging, because it was created in haste, by a model.
Harris responded with the idea that the overarching architecture can come first, to be refined on an incremental basis, with that idea of a “knowledge graph” that reaches across the entire scope of the codebase.
“The first thing Blitzy does is ingest and understand (things) down to the file and module level, and that becomes really powerful,” he said, “because that’s a knowledge graph based on everything that’s in there.”
Feeding Founders
Later, Meadows talked about how Lovable wants to help founders.
“We are their co-founder,” he said, “and so right now, we’re very good at helping them build the product, but we also want to help them grow the product.”
In looking at the process of agentic coding, Gorskikh asked the panel how they feel about the term itself: vibe coder.
“Some people use this, ‘vibe coder,’ as a negative connotation of a person who doesn’t know how to code and doesn’t care about the code, but personally, again, I think it’s a great term,” she said.
“The general feel in the room here is that we we’ve solved coding somehow,” Vasylechko said, noting some of the “unsolved challenges” remaining in the vibe coding era. “We can build amazing stuff, but there are so many different things that we cannot build with these vibe coding tools, and I think I hit on that issue every single day.”
He continued:
“I train AI models, and if I try to write code for very complicated training loops, you know, if I actually want to read out all of the logs from my models as they train, it’s impossible to hold it in context.”
“I love the term vibe coding,” Meadows said. “It took me a little while, but we’ve really embraced it. I think it’s great. What can you not build?”
He did, however, mention some challenges.
“You should be able to deploy everywhere,” he added. “That’s tricky today, and then one of the things that we’re really working on is when you have a proliferation of apps, so let’s not talk about the app itself, but when you have a proliferation of apps, How do you then manage that?”
Those Who Build
Invoking the idea of “great builders,” Gorskikh asked the panel what they see as the values of great people working on AI.
Serge’s feedback was simple: a good pro, he suggested, can explain what they have vibe coded, because they don’t just rely on AI to do it all, or go into waters that they themselves have never waded through.
Another thing that’s important, Meadows added, is knowing the customer.
He also mentioned the value of creativity.
“For the most part, you’re staring at a blank canvas,” he said. “I would (say) creativity, and creativity might be born from pain in the past, but creativity is the thing that I see hold everybody back.”
“You have to be forward leaning, you must be forward leaning. You must be willing to remove biases and be open to embracing and educating yourself on the new normal, and then know that the new normal is going to change, probably next quarter, or in many cases, your prompts are going to have to change across all the foundationals as their capabilities change.”
He had this to say on personal initiative
“A lot of folks are very comfortable and familiar with how they’ve developed software, how they’ve created and maintained software, developed code over their careers, and a lot of folks are really comfortable with how they’ve been using Copilot tools, but if you’re not forward leaning or willing to challenge yourself and think about things differently, you’re going to limit your upside.”
Who Wins?
In closing, Gorskikh asked another open question: who is better poised to take advantage of the new reality: large enterprise users, or small builders?
“Nobody really knows,” Vasylechko said. “Everyone’s still learning.”
“In lines of code, it’s probably in enterprise,” Meadows said, “but I’m most impressed by the number of new entrepreneurs that have come out of this, and I’m most optimistic that whatever sort of consolidation we see in the enterprise, because of agentic workflows, will be easily made up for in the explosion in entrepreneurship that we see, and that’s the fun part of my job. We just see millions and millions and millions of new entrepreneurs, folks that wanted to get in the game, had this idea that had been sitting on the shelf for a long time, could never build – now they can build.”
“I think the new entrepreneurs and the founders who are using it, they’re not the big oil tanker that’s going to take forever to turn, they have their agility … they had no other option but to leverage today’s tools, where at the big enterprise, a big part of what you need to do is forget the way you used to work, and that is a challenge in itself.”
However, Harris and Meadows both agreed that enterprise is in the act of catching up.
“If you work at a major enterprise, your entire career is going to be based on if you’re forward leaning and you get good at this stuff,” he continued. “I think, like, as that becomes very apparent, there’s no other option. They have to also be creating value.”
All in all, some very interesting thoughts on what AI has brought us, and how we will use it. What do you think about vibe coding? Is it a bad word? Drop me a comment, and let me know.
