Thursday 29 January 2026 5:45 am | Updated: Wednesday 28 January 2026 12:01 pm
I Tried to Create ChatGPT in 1999 – This Is What Failure Taught Me
By: Fred Plais
Image generated by ChatGPT
Fred Plais developed a natural language search engine during the early days of the internet. Here, he shares the invaluable lessons that today's AI founders can glean from his missteps and experiences.
How Did You Feel When ChatGPT Launched in 2022?
Were you astonished by the technological leaps? Perhaps you were fascinated by the new possibilities it unveiled? Or maybe you harboured concerns about what artificial intelligence might displace?
For me, the release of ChatGPT and other large language models left me almost speechless and sleepless. These advanced systems accomplished precisely what my company, founded in 1999, had strived to achieve before being derailed by the dot-com crash.
However, from that failure emerged a wealth of insights – not only for myself but also for contemporary AI startups navigating today's rapidly evolving landscape.
ChatGPT Before AI: The Infoclic Story
In 1999, I co-founded Infoclic, a search engine conceived during an era when Google had not yet become ubiquitous. Many leading search tools at the time were relatively basic, often delivering poor relevancy in their results.
It was a period of explosive growth as the online world expanded to include non-technical users. A natural language search engine, where individuals could pose questions rather than rely on keywords, seemed not only logical but revolutionary.
We constructed the initial version of Infoclic within six months and continued refining it through partnerships with content providers. Traffic steadily increased, and our advertising-based business model appeared robust, with projections pointing toward profitability.
Then, the bubble burst. Advertising rates collapsed dramatically, rendering Infoclic unsustainable. Google endured as the dominant search engine due to its innovative approach: ranking websites by popularity based on incoming links and utilising keyword searches.
Consequently, ChatGPT's debut in 2022 was particularly poignant. It embodied the vision I had aimed to realise in 1999, yet the necessary technology was simply not available then. Over two decades of technological progress, coupled with access to vast datasets and immense processing power, now enable large language models to process natural language and deliver precisely what users seek.
Lessons on Timing and Resilience
It would be effortless to resent another enterprise fulfilling your aspirations, even if it occurred more than twenty years later. However, in retrospect, Google was the superior product and likely would have eclipsed Infoclic regardless of the dot-com crash. Our strategy was misaligned with the era – the requisite technology had not yet matured.
Undoubtedly, startups thrive on a compelling idea and a talented team. Yet, the final and arguably most critical piece of the puzzle is timing, which demands considerable luck and a touch of serendipity. The business realm is replete with calamities stemming from delayed action, such as Blockbuster and Kodak, alongside innovations that were prematurely introduced, like the Nintendo Virtual Boy and, indeed, Infoclic.
Establishing a business is akin to undertaking an accelerated MBA programme. You rapidly acquire skills in team recruitment, investor persuasion, product development, engineering management, and customer service. Essentially, you must become proficient in every facet essential to a company's operation.
Navigating the AI Boom and Potential Corrections
Similar to the dot-com bubble, the current AI sector undoubtedly faces analogous dynamics. Startups are attracting exorbitant valuations and substantial investments, and we can anticipate at least one market correction in the future. Based on my experience, I would counsel AI founders to exercise caution regarding these valuations. While they may appear enticing now, they could imperil your business's long-term viability.
The dot-com crash did not obliterate the internet, and any forthcoming AI correction will not eradicate artificial intelligence. Some failing startups will have merely capitalised on the investment wave, destined to collapse, yet we will also witness the downfall of promising ventures with sound ideas, capable teams, and merely unfortunate timing.
The most crucial lesson these entrepreneurs will learn is resilience. The dot-com crash did not terminate the internet, and any AI downturn will not extinguish AI. Instead, it will forge a new generation of entrepreneurs equipped with hard-earned wisdom, prepared to re-enter the fray when conditions are favourable.
Fred Plais serves as the CEO of Upspun.
