Stop reading HN and read the book now. You’re welcome. ;)

The Gibson! Very cool. did you make this?

I made a Tron lightcycle game: https://new.af/tron

Now that AI accelerates dev so much, I suspect we'll get to see a lot of cool throwbacks.

> Now that AI accelerates dev so much

Then where's the massive explosion of software?

While we're at it, I remember an atari st game that was like the tron lightcycle as "Trek4".

Did it really exist ?

Was it maybe called Surround? Try looking up Atari catalog # CX2641 and see if it brings back any more memories :)

Hmm I had an ST. I think I vaguely remember this. Was it a PD game or commercial?

Only thing I can remember is a blue 3½ floppy with "trek4" handwritten on it so it must have been "found" by my relative that owned the st at the time.

That runs super smoothly!

Thanks! working on multiplayer now...

Obligatory sshtron reference: https://github.com/zachlatta/sshtron

+1 for explorers

Manus has been the best agent for turning text into work --useable slides, code, extracting data from websites, etc. that I've seen. There are better tools for specific cases like coding, but for one tool that could handle agentic workflows with minimal oversight and configuration, it's the best.

Hope Meta doesn't hose it.

Greatest products with such an impact on people should not be behind closed doors. And delegating complex tasks to AI is clearly what's next.

That's one of the reasons I'm building out in the open:

- https://github.com/codename-co/devs - https://devs.new/

It's not ready for orchestration yet, but most fundamental layers are already working great.

Create your agents using the LLMs of your choosing, directly from your smartphone of you want full privacy, and with no ads, no paywall, no sign up required.

Manus was ahead of its time. But the directions are parting ways.

Bummer. Manus was the best actual agent for my money. I literally have it working for me right now so I can goof off on HN… no joke.

What is it doing for you that other agents like GPT/Claude wouldn't do?

We are a sandbox provider company and we have a manus like agent deployed to "showcase" our capabilities. You can build one too -- maybe we will open-source it. For now, you can try it for free at https://showcase.instavm.io/

What is your agent doing?

If you’re looking for a (non-meta-owned) alternative, check out our startup Tasklet (tasklet.ai)

I just bought this vintage magazine on eBay. It’s too good not to.

There are a few more out there if you want to thumb-through the deal thing

I honestly think that if you were to show the tech we have today, to someone at OpenAI back in 2015, they would say “we did it!!”

Outside of robotics / embodied AI, SOTA models have already achieved Sci-Fi level capability.

Such a great book… should be required reading for anyone managing engineers.

Interesting, I hadn’t see the Knowledge Navigator before. I would argue that we’re very close to the capabilities shown in that video.

Isn’t this already that? A new business model? Something like OpenAI’s search or Perplexity can run on its own index and not be influenced by Google’s ranking, ads, etc.

In areas where there is a simple objective truth, like finding the offset for the wheels on a 2008 BMW M3, we have had this capability for some time with Perplexity. The LLMs successfully cuts through the sea of SEO/SEM and forum nonsense and delivers the answer.

In areas where the truth is more subjective, like what is the best biscuit restaurant in downtown Nashville, the system could easily learn your preferences and deliver info suited to your biases.

In areas where “the science” is debated, the LLM can show both sides.

I think this is the beginning of the new model.

Complete with milkdrop and skins.. love it