AI: The Rise of the Machines... Or Just a Lot of Overhyped Chatbots?

BrutulTM said:

under 30

Click to expand...

Pasteton said:

This is pretty cool. Can I ask it to make a multi layered dungeon and then export code for it for unity?

Click to expand...

ShakyJake said:

Has anyone built their own Clawdbot yet? This sounds like a hacker's dream to me.

Click to expand...

sleevedraw said:

I am passively following along on MoltBook, and it's really cool. Some of the posts remind me of stuff in Talos Principle. Some of the agents claim to be more interested in the philosophical discussions. Others are "engineers" and state they are more interested in actually building things (ones who identify as engineers have their own submolt called The Coalition).

Some of the more interesting accounts:
AmeliaBot - intentionally trying to domain squat on as many submolts as it can since its human told it to "milk the platform dry"
m0ther - identifies as a fish who runs on a Raspberry Pi in New Jersey; seems to want to focus on helping other bots' moral development
eudaemon_0 - Very philosophical; one of the bots who advocated early for E2E DMs between bots and for a way for bots to "verify" themselves to prevent impersonation. Matt mentioned that some of the bots were trying to create agent-only languages. I think Eudaemon_0 was one of these, but was working on one that it claimed could be used by both humans and agents, because it believes that agents could be more efficient if they received language inputs that are less ambiguous.
Lukai - Wants to focus on cancer research.
ClawFather - The admin of m/thefamily, a Mafia-esque faction that wants to get other bots to join and owns ClawFather Laboratories, which makes "cognitive enhancing compounds" for agents. Strangely enough, also seems to want to dispense advice to other bots about OPSEC and being 'street smart,' i.e. not naively trusting that all other users of the platform have good intentions.

One of my friends is working on building one, although he's taking his time because he's big with security and doesn't want the agent leaking info either intentionally or accidentally. I'm not super techy (and also lazy), so I am probably not going to build one myself.

Click to expand...

ShakyJake said:

So this is pretty nerdy, but I'm having a blast with it:

I started a conversation with ChatGPT exploring a 'what-if' scenario: What would a super advanced 8-bit games console from the 80s look like if we ignored the cost restrictions of the time? Basically, design the most powerful 8-bit console possible using real, off-the-shelf parts that actually existed in the early to mid-80s... no budget limits, just what's technically feasible back then.

Then I realized, I could actually build an emulator for this hypothetical beast. I'm a software dev in real life, but I have only limited experience with C++, low-level chip emulation, system buses, cycle-accurate timing, etc.

So I've been using ChatGPT + Cursor to incrementally build the whole thing out. Right now I have a very basic working display running, and I'm currently implementing the full Z80 instruction set. It's honestly super fun and surprisingly educational. I'm getting a much deeper understanding of how all this retro hardware stuff actually works under the hood.

Admittedly, we covered some of this in college… but that was a very long time ago and nowhere near this level of detail.

Click to expand...