Feb 7, 2024 · 1h 6m · latent-space
The State of AI in production — with David Hsu of Retool
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
Retool founder and CEO David Hsu joins the Latent Space podcast to discuss the evolution of Retool, contrarian principles of startup building, and the practical implementation of AI across modern enterprise workflows. He contrasts raw industry hype with real-world production adoption, sharing insights on developer primitives, open-source infrastructure, and the philosophical underpinnings of intelligence.
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 18.5% of the talking time here. How this is scored →
The hosts as informed peer 4.0 Guest teaching 4.4 Guest disagreement 1.8 The hosts pushing back 1.2
speaking balance: gold is the hosts, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 14:56 Rejection of the low-code category
David directly rejects Swyx's description of Retool as the king of low-code, stating Retool hates the term and explicitly bans it from internal documentation.
Hardest push from the hosts ▶ 54:30 Challenging Retool's sales-led perceptionSwyx presses David on whether Retool is actually sales-led versus bottom-up, referencing observations from founder Joseph Nelson.
Biggest teaching moment ▶ 45:48 Macro reality of global software developmentDavid re-educates the hosts on software industry reality, pointing out that most engineers and software projects worldwide exist outside Silicon Valley building internal tools.
The host holds their own ▶ 53:07 Swyx's multimodal AI industry breakdownSwyx pitches a structured taxonomy of multimodal models (vision, audio, TTS) to David for Retool's upcoming industry survey.
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 |
|---|---|---|---|---|---|---|
| Skipping YC Demo Day and Focusing on Real Traction | 3 | 4 | 1 | 1 | Alessio asks David about skipping YC demo day and lessons on founder persistence. David explains his philosophy of under-promising and over-delivering instead of engaging in demo day hype. | |
| Lean Operations, Motivation, and Hiring Former Founders | 4 | 4 | 1 | 1 | Alessio queries how Retool reached $1M ARR with only four people, and Swyx shares his experience on founder-employee dynamics. David explains that early startups die from loss of motivation rather than lack of capital. | |
| Counter-Intuitive Fundraising: Raising Less at Lower Valuations | 3 | 5 | 2 | 1 | Swyx brings up Retool's contrarian 2021 funding strategy. David educates on the risks of excessive valuations, showing how high valuations dilute employees and trap companies in down-round spirals. | |
| What is Retool? Developer-First Internal Tooling Philosophy | 4 | 4 | 4 | 1 | Swyx asks if Retool is the king of low-code internal tooling, which David immediately pushes back on, stating Retool forbids the low-code label because it alienates developers. | |
| Retool AI: Building AI-Enabled Applications vs. Code Generation | 3 | 5 | 2 | 1 | David explains why code-generation assistants only provide a marginal 10-20% productivity boost due to the burden of debugging low-level code, contrasting it with the bigger opportunity in AI-enabled applications. | |
| Open Source Strategy for AI Infrastructure | 4 | 5 | 1 | 1 | Alessio asks about building vector database integrations, and David outlines his belief that open source infrastructure will dominate before illustrating novel enterprise AI use cases with a major apparel manufacturer. | |
| Retool Workflows and Automated Processes vs. Chat Interfaces | 4 | 5 | 2 | 1 | Swyx links Retool Workflows to AI chaining pipelines. David offers a contrarian perspective, asserting that chat interfaces are reactive toys and true economic productivity comes from automating full workflows step-by-step. | |
| Product Differentiation: Building Primitives vs. Consulting Solutions | 3 | 4 | 1 | 1 | Swyx asks about competition with Zapier and Notion. David explains Retool's philosophy of building reusable developer primitives rather than bespoke solutions or consulting services. | |
| State of AI Survey Insights: Hype, Stack Overflow, and Productivity | 4 | 4 | 2 | 1 | Alessio reviews data from Retool's State of AI survey regarding Stack Overflow decline and job impact. David explains why AI adoption must be led by engineers rather than operations teams. | |
| AI in Hiring, Internal vs External Production, and Model Wars | 4 | 5 | 2 | 1 | Alessio asks about internal vs external adoption. David explains that the majority of software in the global economy is internal-facing, which makes internal applications the natural primary domain for AI. | |
| Future AI Surveys, Multi-Modality, and Developer Ubiquity | 6 | 4 | 2 | 2 | Swyx demonstrates strong domain expertise by outlining a detailed framework for multimodal AI adoption questions, and presses David on balancing top-down enterprise sales with bottom-up developer ubiquity. | |
| Philosophical Perspectives on AGI, Evolution, and Functional Intelligence | 6 | 4 | 1 | 2 | The hosts and David explore philosophical frameworks around AGI, Hofstadter's GEB, evolutionary selection pressure, and Swyx's machine evolution analogies. |
Statements from this episode (28)
Retool skipped its initial YC Demo Day due to having zero traction
“So we were supposed to present at the March demo day, but then we basically felt like we had nothing really going on and no traction. We had no customers. And so we're like, okay, well, why don't we, you know, take six months to go find all that before present…”
Hsu: YC Demo Day is driven by startups over-promising and under-delivering
“And in demo day, I mean, maybe you two have seen a lot of videos, like it's a lot of, Honestly over-promising and usher delivery, because every startup, you know, says, oh yeah, you know, I'm going to be the next Google or something, and then you appear under …”
Hsu: Retool had 5-6 years of runway paying founders $30k-$40k
“I think that at that point we had raised something like maybe a million dollars, maybe a million and a half coming out of YC. So, I mean, to us to people, you know, that was a huge amount of money. I was like, wow, like how are you ever gonna spend a million a…”
Most early-stage startups die from losing motivation, not running out of money
“I think that's actually true in most startups. Actually, it's like most startups die in the early stages, not because you run out of money, but really because you run out of motivation.”
Hsu: Roughly half of Retool's early 20-40 employees were former founders
“One thing we did actually early on that I think has paid a lot of dividends, especially, you know, which was a lot larger now is we hired a lot of former founders. So I want to say like when we were, I don't know, 2030, 40 people, we were probably like half fo…”
Many 2013–2017 startup founders held only single-digit equity by IPO
“If you look at actually a lot of founders in sort of, you know, the operations, statistics space, or, you know, those that fundraise, like, you know, 2013, 20 17, a lot of the founders by IPO only have a few percentage points, actually, for company.”
Hsu: Higher valuation for the same quality company is always worse
“Higher valuation, given the same company quality, is always worse.”
Retool has internal docs barring employees from using the term 'low code'
“We actually hate the low code moniker. We actually never, in fact, we have docs saying we will never use it internally. Or even to customers.”
Hsu: 95% of Retool customers are developers
“Like, 95% of our customers actually are developers, actually.”
Hsu: AI coding assistants only speed up development by 10% to 20%
“If you talk to most engineers today, like, a lot of engineers do use Copilot but if you ask them, like, how much time does Copilot save you, it's not like coding is tenx faster than before. You know, coding is maybe, like, 10% faster, maybe 20% faster, or some…”
Every internal enterprise app will be AI-infused within three years
“Practically every app, every internal app especially, is going to be AI infused over the next, like, three years.”
Hsu: Open-source AI infrastructure and tooling is going to win
“We think the open source movement in AI, especially when it comes to all the supporting infrastructure, is going to win.”
Silicon Valley is still largely searching for real AI use cases
“Especially in Silicon Valley, where a lot of the innovation is happening, I think there's actually not that many AI use cases, to be honest, and, ah, AI, to me, even as of what, like January, 19, yeah, 19 to 24, still feels like, ah, in search of truly good us…”
Hsu: The best AI use cases exist in non-tech legacy enterprises
“If you go outside of Silicon Valley, there are actually a lot of use cases for AI, but, ah, a lot of this is not obvious. Like, you have to get into the businesses themselves”
Hsu: 60% to 70% of AI business utility comes from ChatGPT
“I think a lot of the utility in AI today like, I would probably argue 60, 70% of the utility, like, you know, businesses I found in AI is mostly via ChatGPT and across the world, too.”
Hsu: AI chat only boosts productivity 10-20% and won't replace employees
“The chat, again, will help you, like, 10 or 20%, but it's unlikely that you're going to replace an employee with chat. Like, you know, you're not going to be like, oh, I have a relationship manager at JPMorgan Chase, and I've replaced him with an, you know, AI…”
Hsu: AI workflows, not chat or agents, are the next value driver
“So we think sort of the AI workflow space is a really exciting space, and I think that is the next step in sort of delivering a lot of business value by AI. I personally don't think it's, you know, via chat or, you know, via agents quite yet, so.”
Hsu: Developers prefer modular building blocks over finished end solutions
“I think one huge advantage of some of the developers is that they actually don't, like, developers don't want to be given an end solution. They want to be given the building blocks to themselves to build the end solution.”
Hsu: Retool has never built custom features for a single customer
“We have, I think, basically never built anything specific for one customer”
Enterprise AI skepticism will surge without clear returns within a year
“I do think that based on what we are hearing from customers, if we're not seeing returns in like a year or something, there will be more skepticism. So I think there is like a it is time bound, if you will, so.”
Hsu: Best AI adoption is engineering-driven, not operations-driven
“I think the companies that adopt AI the best, it is going to be engineering driven, I think, rather than, like, operations driven or anything else.”
Hsu: Most global software engineers write internal-facing code
“Most of the software engineers in the world and most of the code-written in the world actually goes towards these internal facing applications.”
OpenAI would decisively defeat any AI startup competitors
“I think if I was competing in startups, OpenAI would win for sure. Like, at this point, OpenAI is so far ahead from both a model and a pricing perspective that, like, there is no reason for it to go really, I think, in my opinion, at least, a startup model.”
Future Llama models will match frontier GPT models if progress slows
“As AI progress slows down, so if we get like Llama-IV, Llama-V for example, maybe it's a comparable at that point, like GPT-V or GPT-VI, like, It made it to the point where it was like, look, I just want to use Lama. Like, it's, you know, safe for me to, you k…”
Retool survey found GPT-4V NPS was around 14 vs 45 for GPT-4
“GPT-IV.V. MPS, I want to say it was, like, 14 or something, like, it was, like, not high, actually. But the GPT-IV MPS thing was, like, 45 or something like that”
Retool changed its north star metric from revenue to active developers
“Our growth star metric used to always be, sort of revenue-generated, or in a new AR-generated. We actually changed it to be number of developers building on the platform, actually, last year.”
Hsu: Retool has 11 separate business units using its platform at Amazon
“Today at Amazon, actually, we have, I think, 11 separate business units that use Retool at this point”
Hsu: AGI should be judged by functional output, not human-like cognition
“The trick is, if the output is the same as a human, then I'm happy, basically. I don't really care about what's going on inside. And so it feels like caring about the inside is like a pretty high bar. Like, why do you care? It's kind of like the plane thing, l…”