There’s one case where it gets super popular, or attacked by spammers/bots, and the site becomes more effort to operate publicly than you want to do for free.

In another case your girlfriend stops using it, as well as some family, and the continued maintenance over the years feels like it isn’t worth it (because there is always maintenance.)

Either way you might want to shut it down, and promising “forever” to users feels problematic when you’re not being finically incentivised to keep it running. With the financial incentives it’s much easier to pass the burden on to someone else if you can’t run it for whatever reason too.

I built Sandclaw (https://github.com/qix/sandclaw) to test an idea of having a human in the loop on any write path that the agent might do. The core agent has access to read anything, but other than to the LLM provider that information is securely firewalled off.

Over time I can loosen the firewall, and allow safer actions (i.e. adding a task to my todo list is always safe), and perhaps build smarter checks to allow more autonomy. But the `DEFAULT DENY` to sending/deleting emails, browsing untrusted websites, or running shell commands.

Is there any project that actually does this? I've been searching but it looks like all the options focus on letting the agent handle when to ask, which feels... problematic.

Comments URL: https://news.ycombinator.com/item?id=47557998

Points: 1

# Comments: 1

Comments URL: https://news.ycombinator.com/item?id=47280270

Points: 2

# Comments: 0

    Current implementation has the following limitations:
      Maximum object size: 65534 keys
      The order of object keys is not preserved
      ...

    These limitations may be lifted by using more bytes to store offset pointers and counts on binary level. Though it's hard to imagine a real application which would need that.

Or perhaps some type of shrinkflation, just offering less services at the same price.

The dictionary allows for better compression without needing a large amount of data, and sending every client the same compressed binary data saves a lot of CPU time in compression. Streams, usually, require running the compression for each client.

Some people are addicted to the product. It's not even that they like it any more, and they can see the harm it causes, but they also can't imagine their lives without it.

I think it's a vital move by the government, but I also think it's too late.

https://www.wheremoneygo.com barely has a database, and bedsides protecting the Google access keys has minimal privacy/security risk.

The main risk, and something I've only seen once, is a corrupted database on the Google Sheets side. Since I don't want to keep backups of peoples sheets, if something goes wrong on the Google-side the entire database could be lost.

With hand written (not arbitrary) rules it's easier to understand the intent of the attacker and build a system that they can't work around because we're blocking them at their source of income. Sure they can figure out how to post messages but unless they can include their link/payload/etc it's not worth their time.

Machine learning defences are definitely a part of what we did, but they're slower to respond to attacks and generally easier to work around.

Proud of what we built at Smyte, and hoping it can find another live outside of Twitter [1] https://docs.sumatra.ai/scowl/

Version I'm pushing up right now has a

  LET ProfanityFilterEnabled := false;

In the meantime the Wikipedia demo should be working, although it is far less interesting (much less data so not much spam popping up.)

Comments URL: https://news.ycombinator.com/item?id=34696434

Points: 2

# Comments: 0

Unfortunately it looks like the Twitter event feed is temporarily down (they're blocking it, possibly as part of shutting down the API on Thursday). I have a cache of events from a little earlier that I'm going to try play through the stream.

Comments URL: https://news.ycombinator.com/item?id=34680269

Points: 200

# Comments: 44

The employees earning minimum wage, and who likely don't hold any stock, won't see any of it.

The buyer would might have been willing to pay $15,000+, but wanted to anchor (and start low) so went $4,500. The real max was introduced when the seller gave the $15,000 number... buyers never want to or will reveal their number (i.e. the $7500 was a lie.)

The metadata in the NFT can (and in my opinion _should_) point to an ipfs:// style URL. The websites displaying the NFT would have to use some redirector, but the actual token would have a URL that anyone could host (known as "pinning") on the IPFS network.

Right now it's owned/stored by the artist and someone buying the piece from me would be able to get it delivered directly from Beeple. That's a new level of unboxed where the piece of art is basically unseen at this point in time and can easily be guaranteed original.

[1] https://niftygateway.com/itemdetail/primary/0xd92e44ac213b9e...