manbash

This somewhat of the equivalent of "quitting cold turkey", in the sense that you remove the temptation from your reach.

The problem is that it's just much easier to un-quit and run the LLM in the same laptop you work on.

It's just so very tempting.

This is very cool. I couldn't find it it already supports duets. I assume it won't.

Copyleft was intended as a principle to keep the software free (as in 'freedom'). Proposing to lock out certain areas of the codebase is directly opposite to this principle.

Yeah, it really wasn't about the school.

Folks here can decide for themselves whether to check green accounts' "Show HN" these days. We are all aware of AI slop and creep in all shape and form.

Moderation is already taxing as it is.

I think it's a nice break from vibe-coding. It feels like a good direction in terms of use cases for LLM.

I was reviewing a confluence page which was reviewed my many stakeholders. Something bugged me to an excruciating extent about the content of the page: something was off at some parts of the text. I inspected the page and it turns out the font color somehow changed to a mildly grayer color than black. This was likely due to a copy-and-paste that has gone wrong.

Wait, I am curious how Gemini "pushes" devs to GCP.

This paper seems like something I would like to keep as reference, especially with its approach to employ STRIDE and present very elaborate DFD (which you don't see as often in my experience).

The DFD in 7.1 is quite impressive.

"software developer" is pretty broad. Here this is specifically B2C (business to customer) applications. I only assume that you haven't been in this market sector, otherwise you would've been more familiar with GDPR and all the concerns that prompted it.

There was a time where the Internet was the wild west and you could've easily been personally targeted and exploited. Businesses sold your data to whoever.

Even today, if you decide to accept all cookies, you're safer than what you used to be.

Rejecting the non-essential cookies puts you in the safest spot from bad actors.

Respectfully, what was the purpose of this comment, really?

And I also disagree, your suggestion is not easier. The & operator is quite intuitive as it is, and conveys the intention.

> i had two options: cut gradually over months or years as this shift plays out, or be honest about where we are and act on it now. i chose the latter.

Making it seem like there's only going to be one mass layoff round. There will be another one, you can be certain.

Yup don't expect up-to-date practices and always come with the expectations that your security will be flawed.

It's nice that they explain the "what" (...it is doing) but not the "why". Who is going to use it and for what reasons?

Also, if it's essentially a sort of metadata, can't the output generated image be replicated (e.g. screenshot) and thus stripped of any such data?

RE agents are really interesting!

Too bad the author didn't really share the agents they were using so we can't really test this ourselves.

No, it's not. You assume it was stolen, while it wasn't implied in that comment. They simply stated that the data have been put publicly at some point or another.

Exciting! It doesn't imply that we should now sprinkle the new annotations everywhere. We still should keep working with proper iterators and robust data structures, and those would need to add such annotations.

Chiming in: indeed, the response is very dry, and raises suspicion. It states the obvious facts and then makes a boring statement, lacking any nuance of the conversation. It is not argumentative in the very least.

Time has been spend, yes. But the topic at hand is after so much time the conclusion to abandon this path is justified.

The list of issues does not seem to stem from whether they had used this build tool (CMake) nor others (nor official build environments).

Why laptops specifically (and not also desktops)?

AI and LLM specifically can't and mustn't be allowed to publically criticize, even if they may coincidetally had done so with good reasons (which they obviously don't in this case).

Letting an LLM let loose in such a manner that strikes fear in anyone who it crosses paths with must be considered as harassment, even in the legal sense, and must be treated as such.

The models generalized "understanding" and "reasoning" is the real myth that makes us take a step back and offload the process deterministic computing and harnesses.

A bit off topic, but I am curious about your experience with Structurizer. My experience wasn't so positive (albeit the general incentive that it's open-source and local).

I have worked with Wayland + Flameshot for the past 2-3 years and it seems to work fine as far as I can tell.

Oh, definitely. Also, they end up getting stuck in a loop, adding and removing the same code endlessly.

And then someone comes and "improves" their agent with additional "do not repeat yourself" prompts scattered all over the place, to no avail.

"Asinine" describes my experience perfectly.

Aw shucks I was almost pulled for another round of Star Fox 64. Unfortunately it's not there.

Ah, those days, where you would slice your designs and export them to tables.

How else can you take responsibility if you don't make it public? You can't have integrity if you hide away your faults.

Under "Programmatic Tool Calling"

> The challenge

> Traditional tool calling creates two fundamental problems as workflows become more complex:

> Context pollution from intermediate results: When Claude analyzes a 10MB log file for error patterns, the entire file enters its context window, even though Claude only needs a summary of error frequencies. When fetching customer data across multiple tables, every record accumulates in context regardless of relevance. These intermediate results consume massive token budgets and can push important information out of the context window entirely.

> Inference overhead and manual synthesis: Each tool call requires a full model inference pass. After receiving results, Claude must "eyeball" the data to extract relevant information, reason about how pieces fit together, and decide what to do next—all through natural language processing. A five tool workflow means five inference passes plus Claude parsing each result, comparing values, and synthesizing conclusions. This is both slow and error-prone.

Basically, instead of Claude trying to, e.g., process data by using inference from its own context, it would offload to some program it specifically writes. Up until today we've seen Claude running user-written programs. This new paradigm allows it the freedom to create a program it finds suitable in order to perform the task, and then run it (within confines of a sandbox) and retrieve the result it needs.

> I speculate that within a few months, the communities will have settled on a single dominant one.

The solutions on the roadmap are not centralized as GitHub. There is a real initiative to promote federation so we would not need to rely on one entity.