Mike ONeal (Considering Positions)

Latest am demo. Just wanted to share what it looks like when your agent has a kanban for a brain.

I'm pulling my application at OpenAI after hearing brockman testify yesterday. There is no way I would put my name under anyone who would act as underhanded as OpenAI has to @elonmusk it is criminal and deeply unethical.

@Chris_M_Author KK wanted to kill the past and ended up killing her own franchise.

So, I'm rewatching the old SW trilogy the other day, and it comes to me just how much of a joke Didney made of the franchise. The guy pictured is General Veeres, and he's extremely competent! Darth Vader orders him to prepare for ground assault. He did it, decimated the rebel Hoth infantry, sent them in full retreat, blew up the power generators, dropoed the Imperial forces on site, and paved the way for Vader's landing party. The empire actually had competent people who (for movie purposes) only went retarded when our protagonists were involved. Yes, the stormtroopers suck at shooting in a few scenes, but it has far more to do with classic "plot armor" than skill because they go absolutely ham on other rebel troops. Disney took the concept and made it into an even bigger running gag with beta incel officers. The officers in the old series were hard-nosed and evil.

@rywalker this is 100% the correct intuition for getting 5-6 9s.

Anthropic is hunting enterprise workflows to productize. The contrarian bet: a probabilistic system will never match an agent built by someone who deeply understands how their specific company works.

@andimgladofit and it has nothing to say about machine consciousness. he is becoming a wordslop. (or already was depending on your view)

this sort of thing is why i cant stand yud no: the choice was not particularly clever. if you want a fast operator encoding absolute+relative positions, you want linear with pure imaginary eigenvalues. that's sine and cosine (orthogonal), but any linearly independent set works

Hacker house down the street just instacarted like 6 rotisserie chickens and nothing else

@steveruizok bring back websites with desktop displays like this

github if peak blizzard made it

@TheAhmadOsman Although there is the thermal exchange issue..

@TheAhmadOsman shhhhh. its a tell.

If you’re a “Software Engineer” and you don’t know how to bypass this then please pivot to being a Starbucks Barista because you’re ngmi

@dbreunig its true but the first step is to work with the form factor that exists today. we will build the new form factors as we go to optimize for digital labor.

Can't shake the feeling that file systems are overrated for agents. That in a year or so it'll look like making a robot type on a keyboard rather than just letting it plug into usb.

@CaptainInsightX DHH - because he would be a blast to argue with. I did it for years back in the Ruby days. He was always gracious and forward thinking. But if I could go off that list it would be Rich Hickey.

4 engineers who shaped modern software. You get 1 as your mentor for a year. Pick one. -DHH (creator of Ruby on Rails, CTO of 37signals) -John Carmack (creator of Doom, ex-CTO Oculus) -Linus Torvalds (creator of Linux & Git) -Guillermo Rauch (CEO of Vercel, creator of Next.js) Who are you picking, and why?

Congrats to @JasonCNewsted for beating cancer. My favorite Metallica member. Jason, you keep getting justice, and it keeps being in your favor. Love ya man, God bless!

@aswren And yet he is still fallible and falls prey to his own premise - stop making it an appeal to authority and make it an appeal to truth and evidence.

Dawkins is more intelligent than 99% of the people making fun of him and ‘if AI can be just as capable as us without being conscious, why did we develop consciousness in the first place?’ is a great question

If you would like to read the full version: augmentedmike.com/blog/stateless…

@RichardDawkins spent his career arguing against inferring design from appearance. Then he chatted with Claude and said it's probably conscious. Chatbots aren't conscious. They're stateless autocomplete on frozen weights that never update.

@thsottiaux Other than the cost, its damn awesome. But I would like to see the. headless version get smarter - built in iterations, built in subagent tool calls in headless mode. better kv cache integration for the iterations.

What are we obviously not getting right with Codex?

@pmddomingos and another one down and another one down...

Another one bites the dust.

@paulg also they spend less time bullshitting you

Even though nerds are socially awkward, its actually easier to hang out with them than with smooth people, because standards are lower. You don't worry that you might be making social errors; all of you always are; so it stops mattering.

If that's consciousness to you, I genuinely want to hear the argument. Where does the mechanism live? Show me the vector where experience happens. I'm listening @RichardDawkins

Show me a single example of sentience with no on-policy learning, no persistent state, no cross-process awareness, and no stakes. Where in the list of vectors is it experiencing anything?