d--b

I started to do this as well. Instead of prompting: "build x", I now prompt: "What questions do you need to ask me in order to have all the information you need to build x". It usually writes a long list of questions, which, once answered makes the thing a lot clearer for the AI to work.

If you don't do this, LLMs tend to make a lot of asumptions on their own without telling you.

It works for pretty much anything. Like medical advice: "What questions do you need to ask to diagnose my headache". Or practical help: "what questions do you need to ask to help me hang this mirror on the wall"

A lot of people obsess over phrasing.

Like for instance you want to tell a coworker his work is shit, but don't know how to put it in a way that's not going hurt them or make you look like an asshole.

I know people who'd potentially spend hours on a single email like this.

CEOs have no clue what's going on at the IC level.

I bet many CEO PA are using AI for many tasks. It's typically a role where AI is very useful. Answering emails, moving meetings around, booking and buying a bunch of crap.

So HN commenters only are trolls now?

Great article. On HN though “dark web agent” sounds like some AI bot did the work. It was far more interesting the way they did it.

what language/framework are you asking it to work with?

You know what works for me? Read these kinds of articles, and then immediately get to work just to prove the guy wrong.

I'm quite pumped.

People who get so annoyed by other people’s habits should really work on themselves rather than writing long blog posts about why others should bend to their own world view.

Or at least make it funny.

I mean that they're not really numbers.

Usually they explain it something like: oh, at first people didn't know what 2-5 added up to, but then we invented negative numbers. Well, complex numbers are that but for square roots of negative numbers.

But that's a completely misleading way to explain these things. Complex numbers aren't numbers aren't numbers really.

Whoever coined the terms ‘complex numbers’ with a ‘real part’ and ‘imaginary part’ really screwed a lot of people..

> Current agents are not nearly smart enough to "execute a plan"

Well I don't know about derailing some critical system, obviously, but when coding Claude definitely maps a plan out and then executes it. It's not flawless of course, but it's generally working.

Even if the posts are fake. Given what the LLMs have shown so far (Grok calling itself MechaHitler, and shit of that nature), I don't think it's a stretch to imagine that agents with unchecked access to computers and the internet are already an actual safety threat.

And Moltbook is great at making people realize that. So in that regard I think it's still an important experiment.

Just to detail why I think the risk exists. We know that:

1. LLMs can have their context twisted in a way that makes them act badly

2. Prompt injection attacks work

3. Agents are very capable to execute a plan

And that it's very probable that:

4. Some LLMs have unchecked access to both the internet and networks that are safety-critical (infrastructure control systems are the most obvious, but financial systems or house automation systems can also be weaponized)

All together, there is a clear chain that can lead to actual real life hazard that shouldn't be taken lightly

At this stage pretty much all ads on the internet are scams

I am definitely using Opus as an interactive collaborator that I steer mid-execution, stay in the loop and course correct as it works.

I mean Opus asks a lot if he should run things, and each time you can tell it to change. And if that's not enough you can always press esc to interrupt.

MAybe the agent returned "nope, can't be optimized any more".

Why not think hard about what to build instead of how to build it?

SpaceX is too big to fail for sure. If it goes bankrupt, it'll be broken down, and trimmed down to the succesful launching operation. But I don't think it's the reason it's buying xAI.

SpaceX buying xAi means that xAI shareholders are cashing in on its current high valuation. It makes it look like Musk is not very confident that xAI can navigate through the AI cycle, so he might as well sell it to rake in the profits.

But he still needs control over it because of the Tesla plan and in case something else happens in the AI field that he doesn't want to miss. So he's buying it with SpaceX, because he can, freeing some of SpaceX cash to pay himself and his xAI investors.

That he managed to bullshit SpaceX investors into buying xAI is pretty crazy. But I guess that's his main talent.

the employer raises the capital, so that you invest the someone else’s money.

This has 2 main advantages:

1. You have limited downside. if you lose all the investor’s money, you don’t lose any of your own money - and contractually you still get paud the management fee.

2. It’s way more lucrative to get a very tiny part of the returns on a 1bn investment, than to get all the returns on a 100k investment…

Yeah I've been freelancing fully remote from France since 2016. Clients are in the US.

I work in quantitative finance, it's a fairly complicated field. I use Claude Code myself to code faster, but at the end of the day, the guys I work for need the numbers to be correct, and they still need people to make sure that everything is right.

The real promotion is when you get out of this corporate bullshit.

So glad I don't have to deal with any of that crap anymore.

Trump is lowering interest rates, fueling the bubbles.

We can’t know when it’s going to happen, but there is a good chance that one is going to be super bad.

We basically borrowed our way out of the 2008 crash and through covid, but we havent repaid the debt. It is so high I doubt we can do the same next time.

Tesla can compete with BYD all right. They have a better brand, they are still a status symbol. They could totally build the best cars if they wanted to.

But competing with BYD would mean becoming "just a car company". And that's what Tesla can't do. Too many promises have been made, the stock's been pumped too high, and there is no way a just-a-car company can justify that market cap. Their only way is to go for the moonshot now. Maybe once the moonshot fails, stock goes down to "normal", and Tesla can compete with BYD.

Probably the only added value is direct notifications of new content.

Patreon is probably going to shut down the payment feature from the app and orient people to the website. That's what I'd do... And bad mouth Apple.

Given Patreon's clients is influencers, this is a fairly bad PR move by Apple, for probably zero return...

I'll buy it when I see it working as a phone. Looks like phone support is an option here, but after I bought the pinePhone, I am wary it's not going to work.

Or Pockit

https://pockit.ai/

The man's a legend. Whoever does not know him should listen to Richard D James, Drukqs, and Syro.

It's incredible how the experimental stuff he was doing in the 90s has pervaded mainstream music so profoundly.

Also, people should listen to Bogdan Raczinski who influenced Aphex Twin quite a bit and doesn't have much recognition beyond the rave scene.

Yeah but that definition is a deformation of what bankrupcy means for a business, and the use of the word bankrupcy is more alarmist than its "proposed definition".

"irreversible damage to nature" is vague. Is a dam causing irreversible damage to nature? It sure is. Can it solve a lack of water in the dry season, of course it can.

People are more ready to build dams and damage nature than to starve.

The whole notion of bankrupcy is unclear, and making headlines with alarmist claims that are not really felt on the ground is just going to make people cringe.

The author is right, I care a lot more about the Venice thing than reading the rest of the article.

And the salesman always says it’s great while it’s in fact lame.

This is huge!

At first I thought this was someone working thousands of hours putting this together, and I thought: I wonder if this could be done with AI…