Over the weekend, I made a game. It's heavily dialogue based. I wanted a screen and game content that centers around this gameplay. It's not a classic VN, more customized like Ace Attorney.
It's a free game so I need an intro that hooks people in the first 30 seconds. What examples are good for this? Since there's few games like this, I look into movies. I don't want that scene from Inglourious Basterds, that's too fresh. There's a lot of good stuff in classic film, before the 80s that nobody has seen. These movies had to carry themselves on content because they didn't have special effects.
And that's where DeepResearch comes in.
I'm looking for specific films that fit the tutorial of the game that I'm trying to build. Without going into details, there's 4 mechanics I want to squeeze into the tutorial level. I'm looking for specific scenes that capture these mechanics.
One intent is "How do I get players to not skip dialogue like they would with a VN?" AI's interpretation of this is "scripts that encourage impatience". I didn't even know what question to ask, and it helped to rephrase my intent into the right questions then find examples of those.
It wrote a 6000 word essay, plenty of citations, some with clips. I don't have to spend a couple nights watching Psycho or The Night of the Hunter. I can just jump straight to the clips I want. I can skim 24 different movies in the same two nights.
Then isn’t this a terrible place to ask your question? The echo chamber is just going to reinforce what you believe to be true.
Do you think the vibe coders getting a lot done are going to be sitting here arguing with the Never-AI-stans of Hacker News?
I would love an intelligent agent I can converse with via both voice and text that has access to all my information, privacy be damned.
Allow me to introduce you to the first step fallacy. https://thebullshitmachines.com/lesson-16-the-first-step-fal...
I like to use the analogy of building a ladder to get to the moon. The first step fallacy says that building a tall ladder is just the first step to building a ladder tall enough to reach the moon.
What I'm talking about isn't attainable via LLMs as far as I can tell, or at least, LLMs alone.
In other words, AI is not just about LLMs.
Software applications with more intelligent interfaces get us closer to a bicycle for the mind. LLMs are already a part of some of those.
But I think we are in the early stages and we really do not know what is going to work or not. People are hyping their own stuff to get funding or customers.
I am taking things slow. Trying out stuff like PydanticAI and simple mcp stuff.