Subject: Re: Constellation Software - CNSWF
Thanks for the link and thoughts -- I mostly lurk but always like your posts.

As for lower barriers to entry, Sven Carlin asked Gemini to do a "deep dive" on the best bear case (https://www.youtube.com/watch?...), and it came up with the following scenario: An AI-savvy startup creates a free AI-based "vampire" interface for the software in question, which can make the user's job easier (e.g. answer phone calls) while quietly subsuming the underlying database. Eventually it can do the same job for no monthly fee, collecting only a small cut of transaction fees, at which point the user is happy to make the switch.

As a recent buyer of CSU, this does worry me a little. But not much. For example, what's my motivation to create this brilliant vampire app, if I'm making much less than CSU was from it, and moreover if someone *else* can come along and disrupt *me* just as easily? It's not like I can look forward to billions in revenue if I succeed. This scenario also seems to ignore some fundamental issues such as whether AI API calls will remain as cheap as they are now (the hardware/energy costs aren't insignificant even if the trained models get cheaper, and they're currently heavily subsidized by investors). And LLMs are still probabilistic and thus subject to bugs which can kill a business and require ongoing maintenance in any case. Maybe most importantly, this scenario will still take years to play out, during which CSU is assumed to sit around and let it happen.

Valuing CSU does seem tricky too, as you say. Are they a Berkshire, but focused on cigar butts and without the free float? Viewed that way, their value depends mostly on their investing acumen (and opportunities), I guess. Or are they an Amazon, which famously had essentially no earnings for most of their growth period, but is now worth $2T or so? I'm not an analyst, so I rely a lot on other investors I trust (Manlobbi and Sequoia in this case). I highly doubt they're the next BRK or AMZN, but I'll hold until I see concrete evidence that AI competitors are at least *trying*, let alone succeeding.