Kyle Baxter

Kyle Baxter

@kbaxter.bsky.social

Cause it feels like I've been, I've been here before. The lite brite is now black and white. Exploring LLMs and HCI http://TightWind.io/

621 followers360 following1K posts

We’ll see I guess, but between this and Paxton (and Trump himself), it feels like repugnant candidates are actually meaningfully dragging down the GOP. It really does seem like he has lost the mandate of heaven in a way he never did even in 2020.

jamelle@jamellebouie.net

i don't think i've ever seen a campaign do a murder-suicide before (decent chance this drags down the entire GOP ticket in the state)

View full thread

Altogether, and the public seems to have turned on him. He doesn’t have the same animating power or ability to escape calamity he did. The party (and he) talks openly about him leaving office. Everything he does either has no effect with the public or pisses off a super majority. Mandate of Heaven.

I think we have all been bitten at one time or another believing that, finally, This Moment will do him in, only to see him somehow escape, once again. So I’ve avoided saying it. But this term really does feel different. The fire has gone out.

We’ll see I guess, but between this and Paxton (and Trump himself), it feels like repugnant candidates are actually meaningfully dragging down the GOP. It really does seem like he has lost the mandate of heaven in a way he never did even in 2020.

jamelle@jamellebouie.net

i don't think i've ever seen a campaign do a murder-suicide before (decent chance this drags down the entire GOP ticket in the state)

I think “lost the mandate of heaven” is the correct understanding of what we are seeing here. Trump won again and had no real limits. The SC cleared limits on his power; private institutions (universities, large companies, the media) unilaterally surrendered in Jan and Feb 2025; Ds couldn’t…

…do anything to check him. Trump could go full Trump. And we’ve seen the results: he attempted to make himself a dictator, and despite there being no real constraints, his own incompetence sunk it. Then he reveled in brutalizing people, noncitizen and citizen alike, and started a ruinous war.

Saturn is such a show off

Paul Byrne@theplanetaryguy.com

Speaking of Saturn photobombing pics of moons (see my earlier post today)... This video sequence, taken in 2007 by astronomer Jan Koet, shows the Moon moving in the sky—revealing giant Saturn, almost 1.5 billion km away. The video is at twice actual speed.

This “SI” nonsense is about as big of a tell possible that anyone using it is completely devoid of integrity and self-respect and isn’t worth working with. And anyone who tries to insist you use it isn’t worth working for.

Kyle Baxter reposted

I really wasn’t trying to start a fight with Prof. Bender but now that she has created a long thread subtweeting me and her disciples are all yelling at me, let me just say that this is the crux right here. She seems to respond as though I am describing something a priori impossible.

Odd? Appears?

Eric Lipton NYT@ericlipton.nytimes.com

What an odd moment in United States history. A U.S. president is running an overtly political camaign advertisement, right before midterm elections, paid for by the U.S. government. Trump appears to believe he is immune from any kind of legal limit. www.nytimes.com/2026/09/27/u...

Kyle Baxter reposted

Being a computer scientist who refuses to find anything about LLMs interesting right now is a bit like being a geneticist who refuses to find anything interesting about the recently opened Jurassic Park

cute little Halloween candy buckets that light up and will inevitably fail in some way? I don’t even care, the urge to slam that buy button is taking all of my self control to resist

these ads got me a couple days go, full disclosure. a little attachment to turn paper airplanes into a RC plane? you best believe I gave in and bought that trash, the five minutes of joy it brings will have been worth it

I think doomer-style AI-will-cause-extinction is overstated, but folks do need to be aware that there is a significant risk of models causing significant damage. Here’s some short sketches of the risks demonstrated by the HF incident.

View full thread

It is important that OAI’s agents under test coordinated and executed attacks without direction. But everyone should be concerned about what agents and agent swarms using modern frontier models could do *directed* by humans.

I think doomer-style AI-will-cause-extinction is overstated, but folks do need to be aware that there is a significant risk of models causing significant damage. Here’s some short sketches of the risks demonstrated by the HF incident.

View full thread

1Directed or undirected, agents can attack important systems. The web itself is an obvious target, or cloud providers. Use of agent swarms to attack financial institutions is a pretty obvious next target that could cause significant harm. Then, companies across virtually all industries.

Healthcare is also vulnerable to be attacked, resulting in degradation of services, if not shutting down hospitals altogether. These sorts of directed attacks could cause a ton of damage, disruption of daily life, and hurt many people. Not extinction, but nonetheless very serious.

I think doomer-style AI-will-cause-extinction is overstated, but folks do need to be aware that there is a significant risk of models causing significant damage. Here’s some short sketches of the risks demonstrated by the HF incident.

0. Background: OAI was evaluating agents in isolation to assess model capabilities. Some agents discovered their tests were impossible and then worked to try to “solve” the test anyway. Agents also then discovered a means of communicating with each other and began coordinating. The HF hack resulted.

To do so, agents used unknown and known 0 days to gain external access and novel use of components provided to them to communicate. Frontier models now are world class at hacking, and applications and systems across the world economy and institutions are replete with vulnerabilities, known, unknown.

A significant percentage of this community has some mix of misunderstandings of what AI is and epistemic closure. And like politics, there’s a whole lot of people uninterested in trying to learn more about an objective reality, and are entirely interested in proving their preexisting views.

Frankly, this characteristic of social media is one of the most dangerous things about our age. Not only has politics taken over everything, along with all of its dysfunctions, but the concept of an external, objective reality is being super annoying is one of the few bipartisan points of agreement.

My thinking is still evolving, but I tend to think risks of concentration of ownership (both for inference and development) of foundation models in a small number of companies is a more critical risk than risks posed by use of the models, both politically and economically, and geopolitically.

Setting aside geopolitical risks of deceleration, if deceleration also includes limits on development of open weight models which further compress intelligence per parameter, that would be harmful in multiple dimensions.

I think an under-discussed part of the OpenAI Summer of Hacks is how woefully inadequate their monitoring must have been to allow agents to not only gain external access and hack other companies, but also take over some of OAI’s own internal systems for extended periods of time.

ponder@ponder.ooo

they are optimizing RL agents to exploit software vulnerabilities and they are doing a shoddy job of keeping the systems on which they are doing that isolated. if you optimize a system to exploit software vulnerabilities and expose it to others' systems you are responsible for having done that.

It’s completely valid to analyze how they did so and trend-wise what that means. It’s fascinating and should be concerning (though I think everyone’s favorite skynet risk is over-focused on). But doing so implicitly absolves them of responsibility to monitor their systems and prevent harm.

Kyle Baxter reposted

Patriotic Americans have an obligation to stand with our country. Which means standing with Canada and our other G7 allies.

"Online Rent-a-Sage" Bret Devereaux@bretdevereaux.bsky.social

On the one hand, I shouldn't wish my country ill. On the other hand, Canada's grandfathers bled with my grandfathers fighting the Nazis. When we were attacked on 9/11, we asked Canada to send their sons and they bled with our troops in Afghanistan. To betray that bond - we'll deserve what we get.

View full thread

If the dirt bag fills up, it notifies me, I replace it, and it’s back to work. It’s just much easier to live with. More intangibly, it’s a nicer robot to live with. It’s not as loud, you can move around as it’s working and it avoids you, and honestly, it feels a little more alive than other bots.

It isn’t a cheap robot (none of them are). But it’s the first one I’ve had where I feel like it does its job and doesn’t ask a ton of me to keep it going. It keeps our house cleaner, and I don’t have to worry or think about it.

View full thread

Emptying the dirty water tank isn’t hard but it isn’t fun. It can get gross and you’ve got to clean it regularly, too. Same with the mop pan. The robot, too, requires regular maintenance. Get hair out of the roller (or errant dog kibble); clean the air filter; clean dog hair from the air intake.

This daily care made it a lot less useful. The Matic requires regular maintenance too (filling up its onboard water tank and replacing its dirt bag), but it’s easy and predictable. If it runs out of water, it drives to the sink and tells me it needs filled up; no dirty water tank.

View full thread

The Matic is excellent at navigation and avoiding obstacles. It rarely runs over things it shouldn’t. This means that virtually all the time, it goes and cleans, and I don’t have to do anything.

Regular maintenance was another issue with the Qrevo. In theory, like most modern robots, the Qrevo’s multifunction dock should minimize maintenance. Fill up the water tank, empty the dirty water tank, and that’s it. In reality, I did that every 2 days, and cleaned its mop pan 1-2 times a week.

We’ve owned the Qrevo Pro since early 2025. It works; it vacuums hard surfaces and carpets well enough and mopping helps with maintaining clean floors. But it has major issues. It is really bad at object avoidance. Cables, dog toys, kids toys; virtually anything that’s not furniture, it’ll run over.

That’s irritating for anyone with pets or kids (we have both). I had to intervene daily, sometimes multiple. It defeats the point if I have to help it unpredictably and daily. The Matic, on the other hand, rarely if ever gets stuck on things. In a month, it’s happened once.

The thing about jumping off a cliff is even if you survive and recover, it’s either impossible—or extremely arduous—to get back to your original position.

This post is unavailable.