Is ChatGPT the next big startup platform?
Lately, I've been getting more DMs from founders trying to launch new products built on top of ChatGPT.
They ask me for my advice, and honestly, I don't know what to tell them.
I'm not convinced that ChatGPT will be the platform on which a wave of new indie startups will be built.
I'm not saying that tools like ChatGPT can't be helpful assistants.
But it seems increasingly unlikely that indie software products with "X powered by AI" as their main value prop will succeed long-term.
Why ChatGPT is not a traditional startup platform
Historically, bootstrapped startups have been built on top of open protocols (such as email, the web, and RSS), open platforms (like WordPress), and open-source tools (including Ruby on Rails, Laravel, Vue, and Tailwind CSS).
These indie founders used open protocols and open-source tools to level the playing field and give their ideas a chance to succeed. MailChimp and ConvertKit, for example, harnessed the email protocol without paying a single cent to Gmail or Microsoft for API access or postage fees.
Even more important, nearly every SaaS application leverages web browsers as their "platform" for launching their software product. There are no licensing fees for loading your web app in Chrome, Safari, Edge, or Firefox.
Early bootstrapping successes, such as 37signals, triumphed because of the open web. Built with Ruby on Rails (open source), accessed through the web browser, and promoted through their blog, Basecamp managed to carve out a thriving business despite competition from giants like Microsoft Project.
For decades, software startups built on the open web were more agile than their (larger) competition. They could move faster, leverage open source, and deploy new versions seamlessly to the web.
Web 2.0 democratized the software industry, giving smaller, bootstrapped companies a unique advantage. Indie founders could differentiate through simpler products, distinctive branding, and creative distribution channels. Usually, the only thing "cost of goods sold" was hosting. Most of the tech you built on top of was free, open-source, or open protocols.
However, the world of Web 2.0 startups wasn't without its challenges. Some founders built upon large, centralized platforms, often to their detriment. Twitter developers faced setbacks when their API access was blocked, while Shopify app creators saw their features integrated into the Shopify product. This became known as platform risk: where the platform changes its policies, algorithms, or features, which can negatively impact the startups built on it.
The risks of building your startup on ChatGPT (or other AI platforms)
OpenAI's ChatGPT is not open. Their source code is closed, and you must pay for API access. OpenAI also just announced their plugin store, a central marketplace they control.
By relying on ChatGPT, startups face the same platform risks they've encountered on Twitter, Facebook, and other centralized platforms. OpenAI has the power to:
Incorporate your product's features into their main ChatGPT offering
Increase the cost of API access
Control who is featured in the ChatGPT plugin store
Already, founders are noticing the costs of GPT4 vs GPT3:
OpenAI presents a particularly potent version of platform risk. As consumers increasingly adopt and pay for ChatGPT directly, the platform's influence resembles Google Search more than the Shopify Store.
Ben Thompson, in his article "The Accidental Consumer Tech Company," highlights this dynamic:
The most obvious business model for the research-focused company was the OpenAI API, access to which was sold on a usage basis.
Then came ChatGPT. Within a matter of weeks ChatGPT had over 100 million users, marking the fastest growth of a consumer app ever, and by all accounts it’s still growing rapidly; OpenAI, whether they intended to or not, suddenly found itself a consumer tech company.
In reality, many consumers are happy to access ChatGPT directly, bypassing the need for intermediary products.
Typically, markets have a "main product" that captures most of the value in a category (I describe this further here), and ChatGPT has emerged as the primary source of value in its domain.
The remaining opportunities for ChatGPT-based products will likely be integrated into existing products as additional features. With their market share, customer relationships, and use cases, established companies can easily incorporate AI as a value-add.
Unlike past technological breakthroughs, it's been surprising to see how fast mature companies like Microsoft have been able to incorporate AI/ChatGPT features within their existing apps.
This trend significantly reduces the demand for intermediary products, often developed by indie bootstrappers. For instance, customers may question the need for a standalone AI-driven transcription product when such functionality is already available in apps they use, like Descript, Microsoft Office, or Google Docs.
Ultimately, this landscape leaves fewer opportunities for small startups.
We're already seeing this dynamic play out:
I'm writing this mostly as a warning. I know a lot of indie founders are hopeful that AI is the next big platform that will produce the next wave of bootstrapped startups (similar to the iOS store, Shopify store, etc).
I believe there will be opportunities for established bootstrapped startups, who already have a strong customer base, to incorporate AI into their products.
And everyone can benefit from AI-assisted coding, problem-solving, and outline writing.
But it seems, with AI, I think most of the upside will go to big companies (especially to OpenAI), not new AI-driven startups.
For more on this, check out Arvid and Tyler's episode.
Cheers,
Justin Jackson
PS: I continue to be concerned about AI and that the pace of its development will outpace our ability to consider the ethics and ramifications on society.
If you're excited about AI, I urge you to check out some of the warnings being issued from long-time AI researchers:
On Twitter, I recommend following Professor Emily Bender
Dr. Stuart Russell (UC Berkeley): Are we living in an AGI world?
Gary Marcus: GPT-4’s successes, and GPT-4’s failures
Dr. Stuart Russell: How to get AI right
Gary Marcus: Deep Learning Is Hitting a Wall
Dan Shipper: Awe, Anxiety, and AI