Hacker News Reader

O_H_Eon OpenClaw SaaS?

Given some of your other comments here, you got me curious. So you don't want it to access you email, then what email are you hoping for it to access? If it's using its own siloed calender, how would you ask it "am I free tomorrow?"

O_H_Eon Ask HN: What tech companies do you think have a good moral compass?

https://oxide.computer

O_H_Eon Goodbye Claude, Hello Pi

> At the time of writing, there have been 244 releases to Claude Code this year; over 3/4 of these changed the system prompt or tool descriptions.

ooof, that is a lot!

And we're supposed to believe them and start one-shotting our projects...

O_H_Eon The Components of Agent Memory

> is something that has been discussed very extensively

Curios to learn more, you do have some reference discussion or posts about that.

O_H_Eon What do the next 4000 emoji look like?

ime those phone keyboard search features are in dire need for some semantic search feature. When my mind happens to search with a synonym that is not included in the underlying description of the emoji, I fail to find what I'm looking for.

Edit: hmmm it seems actually to be more of a problem with the symbols that are constructed by combining two emojis with a zerowidth char. Like the heart on fire one.

O_H_Eon ChatGPT silently passes data between connectors and the bounty wants an ID card

Well, if LLMs are conscious these days...../s

O_H_Eon Claude Code sends your email address to websites

Any good services you know about that allow me to generate aliases while forwarding the emails to my gmail?

O_H_Eon Claude Code sends your email address to websites

Yeah, it feels like a big failure mode nowadays is that we basically have to keep discovering the delta between the model's "common sense" and our common sense, and patch the system prompts accordingly.

For models such as Opus 5.5 nowadays, it seems that there is a fair overlap between those two intuitions in the family of tasks such as the recently shared HackerTube or Simone's game. But when we (humans) are in a different mindset about a task, we have to start editing md prompt files to coerces the model to abide by the "common sense" for the current task.

And yes model "common sense" is probably RLed or emergent patterns, but the point still stands.

O_H_Eon Claude Code sends your email address to websites

Given the seemingly reasonable answer in the comments about it sending that info in the headers when the websites require it for data access, it sheds a light on a reoccurring pain point, that is usually viewed positively when thinking about "agentic coding". Namely, task persistence.

It is a continuous fight in markdown guidance files to try to convince the models to refer back to the human in order to resolve ambiguities or important decisions like those. They've been RLed to hell and back for "safe" task completion and not "prematurely" stopping (which was an earlier failure mechanism). Which is useful and we've bore its fruits recently, but not without drawbacks.

O_H_Eon Residential proxies detected on White House network

> whether it be buried in some agreement somewhere

is that actually a thing that was caught before?

O_H_Eon Kimi K3.1 Pro model weights are out

Wow, really interesting work. Run of the mil benchmarks has been so shallow in measuring real performance, especially outside agentic software development.

Aside from this paper and K3.1 specifically, is there somewhere where I can read more a high-level introduction about the state of things in this niche? like the different types of faults like these models fail and ideally how to fish them out?

O_H_Eon Ripple-65B Preview: U.S. Open-Weight SOTA on Agentic Coding

well yes but the y axis heavily depends on the benchmark you choose

O_H_Eon Microsoft fixes a Windows 11 annoyance as PC revenue declines

I had recently had to install windows for the first time in years, did some chatting with mr gpt and poking around on reddit until I arrived at: https://massgrave.dev. They guide you on how to download windows LTSC from the "official" Microsoft links. So much less crap shoved down the system, it's kind of nice.

O_H_Eon The 'Astra can read and write weights to improve Qwen' post is misleading

I'm even more clueless about the the field.

Those two types of reasoning, are those well definitely in the field generally, or are they mini-personal evals you run?

It has been frustrating trying to quantify why one pseudo-frontier model seems better at some non-coding tasks than another pseudo-frontier one.

O_H_Eon Ask HN: Are you using a custom harness?

Yeah....words are confusing

> AI client is doing this on its own

Well, the llm api on its own cant do anything other than return a stream of text. The harness takes that and decides, this text is part of the response, this text is an edit tool call, this is a bash call to compile, runs the commands and calls the api again (as long as the llm didn't emit a stop_token. They now have a tool_call token the llm sends in between intermittent responses). And saves a session record to disk so that you can exit and resume the conversation later. It also auto injects things like agents.md etc...

I'd say thats the bare minimum. But then usually also includes hooks, introducing tool calls to allow the main session to launch "subagent", and the bits and pieces pile on....

O_H_Eon The 'Astra can read and write weights to improve Qwen' post is misleading

That "Likewise" is quite load bearing in the original article :)

So in effect, all what we're quarreling about here is whether it is _misleading_ to use "Likewise" to transition between two paragraphs without implying that the details of the first apply to the second?

O_H_Eon The 'Astra can read and write weights to improve Qwen' post is misleading

I mean separating two paragraphs by "Likewise" is pretty much how an author communicates to a reader to follow the barely-hidden line of reasoning.

O_H_Eon Serbia just elected the world's brainiest parliament

Thanks for this comment, actually made me think about something new.

O_H_Eon New iMac with M6

> Yeah we could just as easily drag our laptops into the kitchen but then you end up finding a comfy spot on the couch or for the kids in their rooms.

I've been feeling that: much about keeping my sanity in this age is about noticing those patterns we keep falling in and arranging for "easier" ways to fall into healthier patterns. The annoying thing is that it doesn't seem that those "unhealthy" patterns were that easy to predict in order to avoid. Like some 15 years ago would it have been as obvious that we would get to such a consistently predictable pattern of "comfy spot on the couch or bed" and the unhealthy social and personal repercussions of that?

I just wish I start being able to predict those pitfalls earlier than the current few years per tech evolution.

O_H_Eon Show HN: A Truly Modular Keyboard

I'm not sure if I have a use case for this but on the technical aspects this is so cool. Those magnetic snaps made me grin!

O_H_Eon Typesafe AI raises $870M at $7.5B

Oh wow, thanks for this.

O_H_Eon Codex Resets

There is a concrete expiry date listed for each reset in the codex desktop app.

O_H_Eon I used Claude Code to get a second opinion on my MRI

> Most people should have learned at a young age that absence of evidence is not evidence of absence.

That might be true, but it is definitely not the world we live in.

O_H_Eon DSpark: Speculative decoding accelerates LLM inference [pdf]

https://github.com/kenn-io/agentsview

> Local-first session search, analytics, insights, and token use statistics for coding agents, supporting Claude Code, Codex, and more than 20 other agents.

solid piece of software

O_H_Eon Stopping bad guys from using my open source project (feedback wanted)

This is actually one creative idea, kudos. I encourage you to reach out to OP as he was asking for discussion in case he didn't see this comment. This is the only newer out of the box idea I saw in this thread

O_H_Eon Stopping bad guys from using my open source project (feedback wanted)

> There are already so many ways (and reasons) not to do free or open source software. People who find them convincing are using them.

To be honest, I don't think the space between GPL/MIT and commercial closed source is explored enough. I'm aware there's a few examples of things in between, but they are not common knowledge and they don't satisfy everyone. It is not a space that is easy to search online for established wisdom and comparisons in.

O_H_Eon Stopping bad guys from using my open source project (feedback wanted)

> Besides, why would "bad guys" be deterred by a license, anyway?

I imagine because we're talking about a subcategory of "bad guys" who still like to stay within the confines of the law (supposedly at least).

O_H_Eon We're losing our voice to LLMs

So in that case, would someone willing to publish LLM-speak under their name be similarly adopting that "voice".

Does that entail that LLMs are not in fact erasing our societal voices, only making it easier to adopt bland-corporate en-mass?

O_H_Eon We're losing our voice to LLMs

I don't find that current generation LLMs output such short sentences that would start with the same prefix such as "Your voice".

O_H_Eon We're losing our voice to LLMs

There was recently this link talking about AI slop articles on medium

https://rmoff.net/2025/11/25/ai-smells-on-medium/

He doesn't link many examples, but at the end he gives the example of an author pumping out +8 articles in a week across a variety of topics. https://medium.com/@ArkProtocol1

I don't spend time on medium so I don't personally know.