kait.dev

link Posts

s21e05: That Whole Moltbot/OpenClaw YOLOing AI Agents Thing

Link s21e05: That Whole Moltbot/OpenClaw YOLOing AI Agents Thing

This entire thing with non-deterministic agents where you can’t be sure what’s going to happen each time has always felt like Sorcerer Mickey and the Mop and the unintended consequences. But the deal with that story is that Mickey really really wanted to get some mopping done and was totally OK with the risk of, I don’t know, drowning everyone until everyone started getting drowned. These agents are stupendously useful to people and people are getting a lot of value out of them. And they’re horrifically insecure. The value and the lack of security and risks come hand in hand. Like I’ve kept saying, to be most useful, they need to know what you know, and to be able to do what you do. The whole deal with an agent is, if I may American High School Essay myself, that a dictionary defines it as “a person who acts on behalf of another person or group”, like, literally a thing that is empowered to do things for you. There have been some pretty shitty human agents! There are whole laws in England and Wales about the duties that agents owe to their clients. As soon as you let something act as or for you, there are risks. The deal with computing and software is that we think we can mitigate those risks. Sure, we can, a bit. Maybe. But probably not as much as we want to, because the world is just too messy with too much ambiguity and need for flexibility. It will always, always, be easier and faster and less hassle to just let something else do it.

The rise of Moltbook suggests viral AI prompts may be the next big security threat

Link The rise of Moltbook suggests viral AI prompts may be the next big security threat

Palo Alto Networks described OpenClaw as embodying a “lethal trifecta” of vulnerabilities: access to private data, exposure to untrusted content, and the ability to communicate externally. But the firm identified a fourth risk that makes prompt worms possible: persistent memory. “Malicious payloads no longer need to trigger immediate execution on delivery,” Palo Alto wrote. “Instead, they can be fragmented, untrusted inputs that appear benign in isolation, are written into long-term agent memory, and later assembled into an executable set of instructions.”

If that weren’t enough, there’s the added dimension of poorly created code.

On Sunday, security researcher Gal Nagli of Wiz.io disclosed just how close the OpenClaw network has already come to disaster due to careless vibe coding. A misconfigured database had exposed Moltbook’s entire backend: 1.5 million API tokens, 35,000 email addresses, and private messages between agents. Some messages contained plaintext OpenAI API keys that agents had shared with each other.

I love the idea of a delayed, time-offset prompt injection attack. AI provides so many new avenues of attack!

The Cultural Decline of Literary Fiction

Link The Cultural Decline of Literary Fiction

An interesting read on the decline of literary fiction, though I think it overall is drastically underestimating the impact of the rise of alternative media consumption possibilities, inasmuch as the slice of attention/time that literary fiction was fighting for in the 50s-70s had far fewer other competitors.

AI imagination gap

Link AI imagination gap

Every time I’ve tried to involve AI in one of my creative pursuits it has spit out the exact same level of meh. No matter the model, no matter the project, it simply cannot match what I have in my head. Which would be fine, but it absolutely cannot match the fun of making the imperfect version of that idea that I may have made on my own either. Instead, it simulates the act of brainstorming or creative exploration, turning it into predatory pay-for-play process that, every single time, spits out deeply mediocre garbage. It charges you for the thrill of feeling like you’re building or making something and, just like a casino — or online dating, or pornography, or TikTok — cares more about that monetizable loop of engagement, of progress, than it does the finished product. What I’m saying is generative AI is a deeply expensive edging machine, but for your life.

If we are to assume that this imagination gap, this life edging, this progress simulator, is a feature and not a bug — and there’s no reason not to, this is how every platform makes money — then the “AI revolution” suddenly starts to feel much more insidious. It is not a revolution in computing, but a revolution in accepting lower standards.

Digital media

Link Digital media

I’m not sure if this is a special section or an ongoing series, but it’s basically the entire who’s-who of good digital journalism of the 00s/early 10s. Mostly I’m just mad about how difficult it is to find the work of most of them nowadays.

Newark Nonsense

Link Newark Nonsense

As someone who's flown out of Newark recently and has to do so again, trust me when I say you don't want this.

At least one of our engineers wound driving up home, as it would be faster than waiting for a flight that wouldn't get canceled.

Forgery

Link Forgery

The singularity is the victim of bad press. Instead of an omniscient superintelligence, it’s more of a “human centipede” of AI-generated content.

I remain not completely anti-AI, just against its predominant usage of “producing content that otherwise no one would bother to pay for or take the time to create on their own.”

AI reporting

Link AI reporting

If AI-written stories were any good, they’d put them on beats they perceive people care about. Instead, they dump it on topics the suits perceive as lower interest and low-impact, like women’s sports.

How it feels

Link How it feels

I can't explain how this feels.

Athletics bans don't affect me, personally, in terms of preventing me from playing sports - I'm well beyond the age or ability for it to matter.

But that fact doesn't make it feel any less like another punch to the head, another hit to the gut, another in a long line of kicks when I already feel so beaten down.

I can't explain this feeling.

It's yet another way of being told that we're different, separate from, less than. Trans women are women except. Trans men are men but.

It's especially disheartening when so many struggle to have even the basic aspects of their dignity respected (names, pronouns, getting an education, not getting fired for existing while trans). Time and again, the only concrete actions taken are to strip more from us.

I can't feel.

It's a systematic desecration of our humanity, a systemic approach to telling us not only do we not belong, but that we shouldn't exist.

A cistem built on our destruction.

I can't.

I desperately want to avoid talking about the (junk) science of it all. I'm putting the finishing touches on a conference talk about properly being data-driven - so many people take whatever (bad) available data they have and try to map it to outcomes that are only loosely correlated. This is a prime example.

If the concern is the effects of testosterone on performance, then organize your damn divisions among testosterone blood counts. Period.

Link Sacre bleu!

Gird your curds! Say a prayer for Camembert! A collapse in microbe diversity puts these French cheeses at risk.

An interesting unexpected side effect of uniformity in food (which I generally like!).

Aw, man

Link Aw, man

Found out today Pogo sucks as a human.

I hate when I found out art I enjoy was created by assholes. But I have little problem dropping it from my life - there's way too much art out there made by people who aren't awful.

There goes half my glitch-hop playlist

Where does it end

Link Where does it end

management's quest to see how much more cheaply an increasingly poor product can be sold at the same price and under the same name as what came before is, at bottom, the story of basically every industry or institution currently in decline or collapse.

The race to the bottom is a problem because nobody knows where to go once you've won

See almost all current commercial applications of AI, for example.

Lacking vision

Link Lacking vision

It is incredible that all of this works with just a single button click, but all that scaling complication also explains the bad news: you can only have a single Mac display in visionOS. You can’t have multiple Mac monitors floating in space. Maybe next time.

Fear of AI is eternal

Link Fear of AI is eternal

“The big curiosity is what medium a Furby uses to record audio,” one employee wrote. “I would assume that since it can ‘respond’ to certain audio cues that it would use storage similar to a digital answering machine or straight computer memory chips. Anybody know?”

Others said “Furby is only a $35 toy and is not that sophisticated. As a previous [listserv] posting pointed out, the ‘learning’ the doll does is programmed into it so that the longer you use it the more it ‘knows.’”

A great reading of newly FOIA'd documents from the folks at 404 Media. I definitely understand the impetus to understate existing rules about banning personal electronics from NSA spaces, but doesn't it also smack somewhat of security by obscurity?

It's always fun to get messages worrying about people FOIA'ing documents in documents you FOIA'ed.

A Vision for the future

Link A Vision for the future

Apple keeps emphasizing that the Vision Pro isn’t meant to isolate you from the rest of the world, and the display on the front of the headset is designed to keep you connected to others.

I don't care if it isolates me? I don't want to be wearing it constantly, anyway.

If I'm perfectly honest, the killer VR app for me is working. If I can use a head-mounted display for a large screen for an existing computer (and get rid of the gigantic monitors of my workstation / use them when working away from home), I'm in.

Just ... not for $3.5K.

I mean, I would also probably play games on it, but not dramatically more than I do now (which is maybe 1-2 hours a week across all platforms, if I'm lucky?)