ShinyLeftPad
ShinyLeftPad on Hacker News
> We added two sentences to Claude Code’s prompt on Opus 5.5 to steer it away from recreating someone else’s specific game because using Claude to infringe IP is against our usage policy.
By using an LLMs you are already doing it.
> how is the advent of inorganic online content impacting the scaling of pretraining?
Thinking too hard about this question is incompatible with increasing shareholder value.
Most people I know. Myself included.
When I realized like half of my bookmarks don't resolve I realized web is ephemeral. just look it up, if it doesn't come up then it's gone.
But also a cause of cancer.
Why stressing bones leads to repair processes but stressing other parts lead to inflammation?
Who bookmarks things these days?
There's so many dead domains though.
yep, nuclear weapons are not available to everybody
or if everyone realizes the same datacenters go bust.
> as if countless examples didn't prove that llms don't actually distinguish between your instructions and the data they read at your request.
That's what I wrote
You are under an illusion that prompt is treated as a different kind of input by the model, than data it reads.
You can build a 0$ fee system, but how are you going to finance its support, KYC to prevent mad ransom money laundering, etc?
Ironically then foreign adversaries target and hack you via wallet backdoors and such. And because it's such an "unseizable" currency, you don't get it back:)
It does sound like that now that I think about it.
Yep exactly. Broken from design stage. So that engineer were talking about is... Satoshi!?
> A native home for Claude Code, Codex and every coding agent. Every pixel is a file your agent can change. Close it, they keep working.
First two sentences give zero information what it does. Third one is useful but i think many harnesses do it?
Continuing:
> A new kind of agent workspace. Native and terminal first, with nothing between you and your agents. Every part of it is a plain folder your agent can rewrite, and every agent keeps running when the window is gone.
Okay, more of the same. Let's press on.
> Your agents run agents.
And hear back.
> An agent hands work to helpers and keeps going. When they finish, their reports land in its input. No copy-paste between windows: herdr-core carries the messages.
Are there modern harnesses that don't do subagents/subthreads/whatever the same way?
---
There are some interesting bits but the vibed copy is for me offputting and a sign even the author might not understand what he made.
If you follow that trail far enough it's whoever designed cryptocurrency, right? Because banks can and do reverse transactions as last resort.
Less than half is frozen then?
No, it's exactly why that works.
Criminals used to meet in person to discuss crime (because mail/phones can be intercepted) and meet in person to move cash. It was a hard limiting factor on what they can get away with. With e2e messengers and anonymous crypto both of those "issues" are fixed.
Do you know anyone who uses it for legit purposes? I don't. I bet you ban a coin like that and you kill orders of magnitude more crime revenue compared to .001% of honest users using it to send money home and save a few dollars on the transaction.
> the fact that the author, who clearly does want to socialise, would benefit from doing something outside of their tech bubble
I am not in SV and i don't have a single irl tech friend. OK, maybe just one.
Most of my friends would agree with me and with article author. One of them goes further and doesn't even have a smartphone.
IME techies very rarely have these concerns.
So I'm not sure who is in a tech bubble between me and all those people in this thread saying how the logic is illegitimate because the author doesn't go out (which I guess is a "fact" because they all know him personally lol)
compared to a future where the tech they use to wield this sort of power is not allowed to exist. or even to status quo.
they tout UBI because it's one more way to control.
Like cash except you can use it at crazy scale, so worse than cash when crime is considered.
Does a trading bot constitute a consumer? This is the same thing. Who launched the bot is the consumer.
Again the status quo continuing is the good outcome here. UBI is a means of control, not some utopian communism.
aren't you supposed to be legal resident in the country where you're applying for your visa...
Well I don't have the problem it's trying to solve so I don't need to. But among people who install it, how many will? And remember to do it again before every upgrade?
Here's prompt injection via a harness for example:
https://www.promptarmor.com/resources/google-antigravity-exf...took me about 30 seconds to find. there's tons of them. of course it's an arms race, the models get better the harnesses get better the attackers get better but again: for the model itself, your prompt and third party data it reads are the same thing (input). as long as it is true, the attack will always be possible.
It's so funny to read these, as if countless examples didn't prove that llms don't actually distinguish between your instructions and the data they read at your request.
Do you think a malicious repo author won't think of hiding a suitable prompt within?