Kyle Baxter
banner
kbaxter.bsky.social
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.net/
Thanks! This is the same realization I’ve had over the last 1-2 months too, starting with realizing that agents (with the right combo of prompt and tools, like memory and filesystem) are generalizable (i.e. “deep agents,” but bad term). And then agent skills was like a bolt of lightning
November 22, 2025 at 11:23 PM
Gonna use social media for its real purpose for a second (post blurry ambiguous show photos)
September 20, 2025 at 7:00 AM
One of the problems with this “utopia,” as described by @tedunderwood.me (not endorsed by him!), is making decisions requires normative decisions. What economic system (even with abundance, we are still constrained by nature)? What is prioritized as a society? What is the social structure?
July 5, 2025 at 10:25 PM
'Were you not at the Council?' answered Frodo. 'Because we cannot use it, and what is done with it turns to evil.'
July 2, 2025 at 3:33 AM
The president regularly waddles out before the media and regurgitates statement after statement which have no basis in reality, aren’t even coherent as a statement, are profoundly evil, or all three, and the media just carries on as if there’s nothing remarkable worth digging into.
July 1, 2025 at 5:23 PM
I’d say this applies to him, but at this point, that would be flattery.
June 15, 2025 at 9:40 PM
I’m also really excited about Shortcuts.app having a shortcut for prompting an LLM (Apple’s or OpenAI’s, I believe). This is going to be really useful for a number of tasks. For me, automating processing meeting notes into tasks for Things.app is a big one.
June 11, 2025 at 3:56 AM
macOS Tahoe’s Spotlight is pushing all my text-plus-responsive-UI buttons using app intents, but I really want more UI, not just app intents. A mini email widget here below Spotlight would be super nice. www.macstories.net/news/macos-t...
June 11, 2025 at 3:10 AM
Sorta thinking that the headline for this story maybe should include the words “bribe” and “extortion,” but what do I know
May 29, 2025 at 1:35 AM
Moment of relaxation on a long holiday weekend. Cheers, all.
May 27, 2025 at 1:56 AM
it’s probably not going to be good for my time management that I can now create images representing whatever dumbass thought flits through my mind
May 15, 2025 at 5:49 AM
You can read this “analysis” and decide if you find it convincing or not.
May 11, 2025 at 2:08 PM
Thank you for sharing this. Quite provocative. Below is clearly true in some sense. Our use of LLMs have shown though that meaning organizes independent of agents, but each agent is like a mini world with meaning, and can try and change social meaning (intentionally or unintentionally).
May 8, 2025 at 3:18 PM
Interesting. Not so much that evals + human testing didn’t catch any sign of it, but rather human testing did, and no one picked up that might be a concern (or did and ignored it given positive eval and a-b results). Explicitly looking for contradiction among tests and analyzing them would be useful
May 3, 2025 at 3:51 PM
I wrote about this concern back in 2010. I was more concerned about the impact on culture and the individual. I didn’t realize it could be so profound and far reaching. tightwind.net/2010/09/the-...
April 15, 2025 at 3:44 PM
Every year that passes, I am more glad I chose to write out my position in Fall 2016, both as a detailed artifact of my thinking and because it continues to represent everything I believe, affirmatively and what MAGA is. I do wish I had been correct in my prediction tho tightwind.net/2016/10/afte...
April 15, 2025 at 2:06 AM
If this is allowed to stand, it is the nullification of the law and Constitution itself, and none of us will be safe.
April 1, 2025 at 2:43 AM
This is super cool: MCP tools for macOS. github.com/Dhravya/appl...
March 22, 2025 at 4:46 PM
This is such a beautiful, structured process for verifying an argument. Decompose into constituent premises, and verify each premise’s logic and factual groundedness. This is useful for any sort of QA task, from simple to complex.
March 15, 2025 at 7:47 PM
“Toward an AI co-scientist”: arxiv.org/abs/2502.18864
March 15, 2025 at 5:14 PM
“Arlington Cemetery Website Loses Pages on Black Veterans, Women and Civil War” www.nytimes.com/2025/03/14/u...
March 15, 2025 at 4:02 PM
Celebrating his continued rampage using the powers of the office, and completely invented powers he has declared, by declaring how terrible it is to “weaponize” government against opponents. By which he means the rule of law.
March 6, 2025 at 11:51 PM
This resonates. I’m intellectually interested in agents-as-orgs, but for implementing real tools useful to people, I’m dubious about value beyond Plan Act Verify roles. It feels like if you need to add another role to solve a problem, it’s mostly just adding a new capsule around the same problem.
March 6, 2025 at 6:10 PM
The party of death.

bsky.app/profile/anne...
February 27, 2025 at 12:55 AM
In so many words, this is DOJ saying that civil servants serve the president—the individual, or one individual anyway—and not the Constitution. That should send a shiver down everyone’s spines.
February 14, 2025 at 2:47 AM