Continued weak spots of AI, from the point of view of a business professional and not a PhD biochemist: 1) SVGs. The ability to 'illustrate'…

Continued weak spots of AI, from the point of view of a business professional and not a PhD biochemist: 1) SVGs. The ability to "illustrate" and have that thing be infinitely scalable. See below image. If you ask for flat PNGs, fine. ChatGPT is the best at the image game right now. Claude Design has the same SVG issue. Hard to image scalability in creative projects or game design without it. 2) Gullibility. I have had to create a specific prompt to remind Claude that what people say or type is not always true because the prompt "imagine you're a greedy capitalist and read from their perspective" isn't good enough. This is a tricky line to toe, however - I don't want Claude to be manipulative, but I need it to deeply understand the world of manipulation to better support my response (ex: email reply draft) to it. 3) Multi-layered communication. Imagine you have to turn down a client event (but it's your favorite client, and you could have technically moved your friend group trip to make it happen but you didn't want to, but just in case your client is about to fire you if you don't do it, you want to write it in a way that doesn't close that door, but you don't want to sound desperate, but you want to show your respect for them, but you also had to turn down their last thing so you have to be delicate and also your vendor is on cc so you can't say secret information but you want to hint at their release but you'd get fired if you mentioned anything, but you want to drum up excitement). Claude, which I find to be the best writer of the bunch, just says the "hidden" message out loud like an untrained intern at a cocktail party after a glass of wine. Delicate nuance is tough. 4) Humor. I know the Claude Mythos paper said it was better at humor, but the current productionized models don't seem to understand the unspoken construction of a joke. It always defaults to randomness. It assumes that non-sequitur = humor. Like if you said "I want to say something funny walked into an office" it would say "oh a flamingo with a tutu!" even though it had no relevance to the rest of the setup. 5) Extremely long context. People incorrectly believe that multi-million token conversations only happen with engineers+AI because of codebase length. As more business professionals lean into building, and as we connect more systems (and more people into single shared systems), we're going to need to find ways to manage 10M+ tokens without context degradation. Countless business use cases are held back because of this. 6) Collaboration. How is there no way to merge memory yet? How is there no no-code model fusion in market? Agent teams starts to get closer (in the sense that you can build AI personas of your board of directors and have them fight it out and review the outcome), but I want true merging and model weight changes.

Source: Allie K. Miller (X) | 2026-04-26