alt Hacker NewsAm I one of very few people that believes LLM apps don't solve or replace a large number of apps.
Take all these basic lifestyle CRUD apps. Maybe it's a skill issue on my part, but whenever I've tried to use LLM apps over any sort of medium-to-long-term time scale for tracking something (ie. endurance running training), within about a month it begins hallucinating, recalling the wrong things, linking the wrong bits of data to other bits of data (dates, distances, heart rate data, workout paces etc), and it does all of this even when pointed to a basic spreadsheet or simple text file to keep it onboard with what happened where. It just seems to look at these files, and then look at it's own context and confuse all the details based on its own context.
Every other month, some brand new shiny model is released and it's exactly the same deal every time. I think the underlying technology simply can't deal with these sorts of goals. A whole range of apps for lifestyle tracking don't seem under threat, in my opinion, if this is the best its capable of.
Many other things I can't imagine wanting an AI to do for me, especially anything involving payments, bookings. There was a discussion the other day on here about AI being used to order takeout. Why would somebody want to delegate one of their last personal decisions to a LLM?
Replies
I am in the pro food-order delivery department. I just want to tell my phone, "Hey Siri I want to order at Restaurant X, we are two people, no desert - make suggestion" and it shows me the order on the display with total price, arrival time and the selected food items. If I want, I can command changes - or simply say "Yes thats good, confirm the order" and be done. This is especially useful as I mostly order the same things across 2-3 restaurants, that is most of the time the suggestion is probably right.
I don't doubt your experience with a basic lifestyle CRUD app today.
However, about 1 year ago, AI coding agents were laughed. Now the vast majority of code is written by them.
Intelligence getting exponentially cheaper per token. At some point, there will be an inflection point where "it just works".
A good alternative for your example is to make it create a stateful artefact (Claude) / whatever OpenAI calls ChatGPT.
And there, it has a database of sorts, running god knows where on god knows what, and you have a single-user web app that you may or may not ever open and see (since your favorite AI agent can just read and update it for you anyway).
Today it’s still something the user has to suggest / request.
Tomorrow? If they put a bit of effort these models will reach for building these one-off ephemeral or long-lived hyper-custom apps the same way they now reach for a python sandbox for basic arithmetics and statistics.
Looking for a flat? Scheduled task + CRM like artefact for it to monitor, maintain, and update.
Looking for holidays? Your own personal shareable moodboard with destinations and what not.
Have a use case or task for which chat or editing text docs is a frustrating way of correcting or giving feedback? Have it build a dedicated, custom artefact just for that.
When you encounter a use case where LLMs limits show, make them build tools.
What you're saying is true now, but given the existence of grok-bot and similar dots (et al.) there is now an amassing amount of training data for "personal assistants" -- so in a couple of generations time, you'll find frontier models will be extremely capable in these areas.
Harnesses have been developed for this, and its also fairly trivial to do large scale verifiable-rewards post-training for "personal assistant" style things. eg., invent a booking website, invent a PA GUI-aware harness, and reward the model when a successful booking takes place. Do that 1bn times.