Dec 6, 2024 · 38m · no-priors
No Priors Ep. 92 | With StackBlitz CEO and Co-Founder Eric Simons
0:00 / 0:00
▶ Watch on YouTube →
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
In this episode of No Priors, StackBlitz co-founder and CEO Eric Simons joins Sarah Guo to discuss the explosive rise of Bolt.new, an in-browser AI tool enabling users to generate and deploy full-stack web applications via natural language prompts. Simons details the underlying WebContainers architecture, explains how AI code generation is reshaping software economics, and shares personal insights on founder resilience and the future of agentic coding.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. The hosts hold 23.1% of the talking time here. How this is scored →
The hosts as informed peer 5.0 Guest teaching 3.4 Guest disagreement 1.3 The hosts pushing back 2.4
speaking balance: gold is the hosts, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 17:52 Eric rejects traditional no-code paradigms
Eric firmly rejects the historical framing of no-code GUI builders, asserting they were merely stopgap compromises that are rendered obsolete by modern code-generation models.
Hardest push from the hosts ▶ 16:57 Sarah questions the long-term viability of no-code toolsSarah directly questions the premise of no-code platforms, arguing that walled-garden GUI tools inevitably trap developers and create architectural dead-ends.
Biggest teaching moment ▶ 10:23 Eric highlights academic benchmark failure in real app evaluationEric educates the host on why academic benchmarks like SWE-bench fail to capture full-stack development, explaining how Bolt Local became an industry-standard stress test.
The host holds their own ▶ 28:24 Sarah brings investor perspective on browser-native tooling and IDE economicsSarah connects Eric's thesis directly to her early investment in Figma while citing historical venture market realities about the difficulty of monetizing IDE software.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | The hosts as informed peer | Guest teaching | Guest disagreement | The hosts pushing back | Why |
|---|---|---|---|---|---|---|
| WebContainers Architecture and In-Browser Full-Stack Execution | 6 | 4 | 2 | 4 | Sarah frames the common skepticism around AI wrappers and asks Eric to defend open-sourcing Bolt's system prompts. Eric articulates StackBlitz's underlying WebContainer architecture in WebAssembly and why prompt open-sourcing builds a sustainable open-source ecosystem. | |
| Community Education, Power Users, and Bolt as an AI Benchmark | 6 | 5 | 1 | 2 | Sarah draws an insightful parallel to Midjourney's power-user community and cites SWE-bench limitations. Eric elaborates on how Bolt Local evolved into a real-world testing benchmark for frontier LLMs akin to gaming benchmarks. | |
| Real-World Software Case Studies and Development Cost Arbitrage | 4 | 3 | 1 | 1 | Sarah asks for concrete product examples, and Eric walks through real-world cost and time arbitrage cases like ViralHooks and Chilled CRM. Sarah chimes in with market perspective on development pricing. | |
| The Shift from Traditional No-Code Platforms to AI Code Generation | 6 | 4 | 2 | 5 | Sarah challenges the viability of no-code platforms based on historical lock-in and ecosystem friction. Eric validates her skepticism but explains why Sonnet 3.5 represents an architectural inflection point toward standard code generation. | |
| StackBlitz Origins and the Technical Vision of WebContainers | 7 | 2 | 1 | 2 | Eric shares StackBlitz's founding story and the technical ambition behind WebContainers. Sarah demonstrates deep context as an early Figma investor, noting the historical difficulty of monetizing IDEs and recalling Eric's scrappy AOL days. | |
| Navigating Startup Uncertainty and the Ironman Athletic Mindset | 3 | 2 | 1 | 2 | Sarah gently ribs Eric about taking on an Ironman during a newborn phase and startup pivot. Eric reflects on channelizing extreme physical discipline to handle startup stress. | |
| Future Predictions: Agentic Workflows and Deterministic AI Training | 3 | 4 | 1 | 1 | Sarah asks for short-term predictions in code generation. Eric breaks down why software engineering models will continue to outperform other domains due to deterministic execution feedback loops. |
Statements from this episode (19)
Simons: Bolt Reached Nearly 200,000 Weekly Users Building Web Applications
“Last week, we had, you know, almost 200,000 software composers, you know, we like to call them, that use Bolt to build web applications.”
Simons: AI Software Composers Could Eventually Reach 100 Million
“We think that, that number should be a hundred million. And yeah, and we, we're on this growth clip it seems like you know, maybe, maybe we'll get there sooner or later.”
Simons: In-browser dev environments eliminate server latency and cloud compute costs
“It's very complicated and expensive to typically do this if you're going to use servers. So it's very valuable to like do it in the browser because it's extremely fast. There's no latency. You're not paying by the minute, you know, for some cloud.”
Simons: Simple AI wrappers quickly die when foundation labs update products
“Whenever the next model comes or whenever, you know, one of the AI labs eventually integrates that into their chat products or whatever those companies, you know, kind of tend to go away pretty quickly.”
Simons: StackBlitz spent five years building its WebContainer technology
“One of the big advantages we have is we've been building this web container technology for five years, and it's like pretty, pretty difficult stuff to do.”
Simons: System prompts are not a moat, like viewing source on Google
“It's not going to be, you know, the system prompt kind of a, you know, for those who have done web development, it's kind of like view sourcing on a webpage. Like you can go to google.com, you can view source. No one so far has kind of done that and built anot…”
Simons: Most Bolt users are non-technical, led by former dev managers
“That's like the majority of people using Bolt at this point are, you know, non-technical. And it's interesting, the most successful people that are using Bolt are actually people that have had to interact with or manage development teams.”
Simons: Managing AI models is extremely similar to managing human developers
“Because it turns out kind of managing an AI is extremely similar to managing actual software developers.”
Simons: Startups are launching and making money on Stripe using Bolt.new
“We've been online for just under two months now, and we've already had the first startups launch out of this thing. You know, they've used Bolt to build their startup and are making money, like, you know, charging on Stripe or whatever have you.”
Simons: Bolt.new reduced a $5k, 3-month app project to $50 and 2 weeks
“A week before we launched Bolt she went onto Upwork and listed this project and said, Hey, I want to build this product. Da da da. Asked for quotes on the thing. She got a quote for 5000 dollars from, I think it was Dev in like Ukraine or something estimated t…”
Simons: Web dev shops use AI to build faster while charging standard rates
“A lot of folks you know, especially in the web dev shops, et cetera, They're just, they're able to punch out incredible web applications faster than ever before for clients and are able to charge the same price.”
Simons: Claude 3.5 Sonnet crossed tipping point to write production applications
“And just from a technical perspective, AI code gen models there's a tipping point specifically with Sonnet three, five, there's a chasm that's been crossed here as far as AI models are,
Have gone over the tipping point of being good enough to really write real…”
Simons: StackBlitz shelved Bolt in February until Claude 3.5 preview arrived
“Earlier this year, I think like in February, we had the idea for Bolt.
We tried to build it with some of the, you know, frontier models available at that time.
Wasn't possible to do.
the models were just did not spit back quality, accurate code that was const…”
Simons: Large number of Cursor downloads come from non-developers
“There is, and there was, and is a large number of people
They've been downloading cursor that are not developers because it let them meaningfully dip their toes into kind of like clicking accept change from the AI.”
Simons: Successful computing platforms must natively support building their own software
“If you look at every other platform that's ever existed, that's been an important capability of every platform that's ever
Succeeded in a meaningful way.
You know, Windows can build Windows apps.
Macs can build Mac apps.
The web does not have a built-in wa…”
Simons: Figma's original pitch was a WebGL 3D demo, not a design tool
“When they, their first pitch for Figma,
Was not like a design tool.
They didn't have a design tool.
They had a demo, like a WebGL demo of a three D ball dropping.
And the pitch was what browsers have gotten powerful enough to do meaningful three D graphics…”
Simons: StackBlitz has 3 million monthly active developers
“I think we're doing like three million developers a month using StackBlitz today.”
Simons: Engineers will increasingly operate as high-level software composers with AI agents
“I think engineers are going to more and more just be instructing these things at a higher level than just, Hey, tab, complete this thing. It's like, Hey, go and do X, Y, Z and send it off.”
Simons: Code Generation Is Not Hitting AI Scaling Limits
“I think folks have been kind of concerned about are we hitting kind of limits of this stuff, et cetera, and no, I don't think so. Like, in the specific realm of CodeGen, I think we're going to see, I think we're going to see a lot of improvements pretty rapidl…”