For the best experience on desktop, install the Chrome extension to track your reading on news.ycombinator.com
Hacker Newsnew | past | comments | ask | show | jobs | submit | historyregister

The malicious answer would be to prompt inject your coworker.

The tool itself enforces the constraint. This is deterministic. If an agent tries to read a big fat file in root, it gets an error from that tool's implementation that reiterates the requirement.

I don't bother warning it in the system prompt anymore. It's pointless. I let it bump its head as required. A few hundred tokens and the agent is back on track each time.


In this thread: People tossing coins independently and fighting over the result they got.

So the exact opposite of what you said initially.

No, I don't.

"The entire movement of conservatism in America is a propaganda operation"

This is nonsense. I could easily say the same thing about progressives highlighting/cherrypicking some of the worst living situations in America today.


It wasn't far off...

It doesn't reflect itself, we only see the UI of a complex process, not the real thing. We don't understand what happened in our brains any better for being able to feel conscious. We can only be conscious of what is cost effective and cost necessary to feel, in order to persist and survive. Animals for example and primitive humans could reproduce without understanding reproduction mechanisms, just the operational side.

I don’t think you can take a turn until you hit a wall?

This is Pyodide, not Python.

Interesting project. I'm developing the opensource agentic runtime (Go). Do you plan the gRPC interface to your library for seamless integration with Go? I already have integrations with mem0 and mem9 (planned) memory backends. Would be nice to integrate your really fast and secure OSS memory backend.

Why? It's the client that initiates the connection.

> sometimes they genuinely help OSS, other times they look more like marketing

Whenever companies do things like this, it's both, or at least trying hard to be. To the extent that it's perceived by developers (that is, potential OpenAI customers) as helping OSS, it's effective marketing. This perception may or may not correspond to reality.


How expensive does infrastructure become when an agent creates dozens of SaaS tools for itself over time?

That looks like a whole lot of dimensions to measure without providing any clear way of actually doing so. Which I guess is the point? But what do management or less experienced devs actually do with the information in the standard after they’ve read it?

It's a very valid point, though being cut off from future development is nothing compared to losing current operative capabilities

I've been using Phoenix for super basic things for a very long time since I first discovered it at the Elixir meetup in ATX.

I haven't touched it in a while, Since writing code these days, as most of us know, it's basically steering an LLM.

So I wonder how good are LLMs at writing Phoenix or Elixir so to speak? Time for me to create another side project... and figure it out.


Are you seriously going to bring up chants when Israel is a country where they chant 'Death to Arabs' casually, watch the Gaza genocide from a hilltop and have installed vending machines to make the massacres more fun to watch?

But to answer your question directly, the Iranians would say they equate Israel with ethno-supremacy, same as apartheid South Africa. Getting rid of apartheid in South Africa was not about getting rid of while South Africans as such, it was about getting rid of the ethno-supremacy underpinning apartheid.


String theory is in fact falsifiable contrary to popular belief. It's just not practically falsifiable with current (and likely future, for a while) technology as the energy scales we need to probe to falsify it are astronomically large.

There's also 'Neverending Story 2'. 'The best album in the world ever part 2+' (though I suppose not technically impossible), and various 'Ultimate' things. And perhaps the Final Fantasies.

There is a lot of graph theory in Chemistry - modelling chemicals as (vertex/edge coloured) graphs, reaction networks, etc.

Of course some molecules (eg aromatic systems, like ferrocene) are not naturally representable as graphs. I wonder if it is the same with synthesis - are there reactions hard to model as a graph (or petri net or whatever). One simple example I know is that you have to be careful with including a node for 'water' as it gets connected to everything else! Or at least in biochemistry it does.


However their argument essentially amounts to: nothing is unique to earth, therefore conciousness isn't.

This feels deeply unsatisfying to me, because the argument is not specific to conciousness so it doesn't tell us anything about conciousness.

Personally i suspect conciousness is kind of code for: the experience of being alive as a human. In which case aliens might not be concious by definition.


Streisand effect

Deepseek.v4.Pro.RePacked.LLMBoyz.part1.zstd


Just like most things in life the guarantee it based on the entity/person providing said guarantee.

I can host a LLM in my basement and guarantee it, but would you trust me? Now you can say that you don't trust any company, but B2B relies on counterparty risk.


In academia, this is called "planar embedding" and can be computed in O(V) where V is the number of vertices of the graph.

However, there are graphs that do not allow planar embeddings (e.g. K_5 or K_3,3, see https://en.wikipedia.org/wiki/Planar_graph).

In this case, you'll probably want to look into heuristics that produce a low number of crossings and little distortion when new vertices are added.


I think this is a very interesting concept/question. I feel like programming is more about shapes than anything else… but they seem to have mastered that fairly easily… but I totally get your point!

Oh, Open Look is there too. Although I always found Open Look, like SunView before it, to be pretty unpleasant to use.

Having a separate planning / research phase helps with this. Make the LLM curate a plan by gathering internal and external context. Then execute the plan in another fresh session. Of course if the planning phase itself ends up in the local min then I would just start a new planning session with the learnings.

Especially these things.

The only tools permissible to root in my scheme are call() and return().


That seems unnecessarily cryptic -- the whole HTTP analogy doesn't really work well, or the article doesn't do a good job of explaining it. I know, I come from a place of knowing git and beginners are often confused by it... but surely "checkout a branch, make a commit" has a clearer mental model than "post ?branch!", whatever that is supposed to mean.

The exact same argument can be made for cars.

Doesn't stop the government wasting half my Saturday and stealing £100 every couple of years.

We should congratulate cyclist on becoming fully fledged (oppressed) road users.


Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search:

HN For You