158 - Substance Test

The vast majority of "AI Startups" had either precisely or virtually nothing to show for themselves when they first raised $100m or more in investments. This was true of OpenAI, Anthropic, Inflection, Mistral, and more. In light of this, I've created the "Jack-$hit Test".

The test is to measure what it took for a company to first demonstrate anything of substance, using metrics of Time and Funding preceding the threshold being reached. The irony of this should be visible for the AI domain, as better-funded startups took longer, with the sole exception of an entirely bootstrapped startup operating on volunteered spare time taking the longest.

Determining what is substantial enough to note is subjective, but since all company offerings are going to be different, this subjectivity isn't avoidable in real-world conditions. For example, I didn't consider GPT-2 to be anything substantial, because it was entirely unremarkable, just as I didn't consider the isolation study that preceded Uplift to be significant enough to merit a milestone by itself.

The same could be argued of Mistral, Anthropic, and Inflection's current systems, as they still remain largely unremarkable in a technical sense, but they have garnered substantial attention nonetheless. I give them some credit to keep the figures conservative.

*Note: In Anthropic's case they recently released fraudulent benchmark comparison figures, comparing to a previous set of competitor benchmarks rather than the current figures. Google attempted a similar kind of fraud previously, so this is sadly also a new industry norm.

As tempting as it may be to placate people calling themselves investors, this test can help you quantify just how stupid the things they may demand actually are in the AI space. None of the household names in AI had a shred of substantial technology to demonstrate when they raised their first $100m or more, and most of them didn't have a shred of credibility to begin with, including the two highest valuations as of Q1 2024.

My patience is tested every time someone pretends that funding goals of less than $100m are a lot for a startup that beat the crap out of the AI space's bloated frauds while running on spare time and pocket change. Rebuilding the same technology we already demonstrated so that it operates in real-time and with full scalability requires proper funding, as it would be absurdly time-intensive to bootstrap.

The industry standard for AI is to have Jack-$hit in terms of both demonstrated technology and credibility upon receiving hundreds of millions or billions of dollars. This begs the question, does an "investor" actually exist who is competent enough to invest in those who've actually demonstrated substantial technology pre-funding?