With all acknowledgement that I don’t know what the hell will happen
I really think it depends on the nature of the software they are building. We were playing mexican dominos a couple nights ago, and we ran out of scorepads.
I told claude to build me one as a web page I could share around. about 90 seconds later I had an html file I could share that was a perfect scorepad for 6 players, would track rounds, total scores etc, and looked good
It’s just really really good at cookie cutter tasks to build a quick app or something. Its less good at systemic software systems with many of millions of lines of code. I’m inferring that from its struggles with my project that is 100,000 lines at most, but of course we just saw an enormous leap with opus 4.6. I see it as a breakthrough, and since we have had one we could get another.
So I still think the larger and more complicated the system not only the harder for AI to replicate, but the greater the risk to swapping out. Every SAP implementation I read about in company PR’s is a nightmare.
So the bigger, more integrated, and risk sensitive the less likely it gets swapped out. If it costs me 0.2% of my revenues for a system that runs the whole organization I’m just not going to roll my own. There are more important things to focus energy on. I can’t see myself as an investor being pleased with an earnings PR emphasizing managements focus will be on vibe coding their auto parts inventory system.
But I agree some of this work will happen. Expensive and not overly complicated niche systems say an inventory management tool geared towards iphone repair shops is in trouble. I think what you will see a lot of is custom software for business tasks that are highly specific or niche where nobody would build this thing for general use because the market is too small and nobody would build if for themselves because the cost is too high.
Software backed by proprietary data is still valuable. META isnt valuable because of their uniquely amazing social media software platforms. Its all that network, data, and way to monetize it.
Really I just see investors assuming an AI takeover in places where I think the outcomes are still fuzzy. Of course many of these software companies are still expensive even with the haircuts. But I am serious about the quality of this particular model. It is impressive enough that a path to truly autonomous workers seems more plausible.
Until one makes a catastrophic mistake somewhere and AI related companies drop 40% in one day.
AI
Market Data 
Markets 
