choxi

But how would it know which weights to update? It has 3.5 and 3.8 weights memorized in its own weights and can somehow figure out which ones are responsible for different skills just in its own reasoning trace? (Since it didn’t use any external tools)?

That’s pretty crazy because, as I said in another comment, I just asked ChatGPT how many r’s are in blackberry and it said 3. I know it’s not Astra and the app doesn’t tell me which model it is, but I’m still very confused about how a model can edit neural net weights but not count letters in a reasoning trace.

Edit: Oh I misread, it didn’t use tool calls for the first part but it did have a custom harness for the part where it edited a models weights. Still impressive but not existentially confusing anymore.

I just asked ChatGPT how many r’s are in blackberry and it said 3. I’m so confused about the range of AI capabilities

Could you use Astra to create specialized models? Like if I want a model that just writes React, could Astra pull out the subnetworks that know how to write React code?

A domain search app called NameTheory: https://apps.apple.com/us/app/nametheory-domain-search/id680...

It uses Apple Intelligence to help brainstorm domain name ideas. I’m interested in seeing what small local models can do as they catch up to larger ones, this is my first iOS app so I started with something simple that scratches my own itch.

There are some other interesting dynamics in UX that aren't just cyclical trends:

- Natural language interfaces. You can now communicate with your computer with language, voice or text. There are some situations where this is an improvement, but others where it's not. It'll be interesting to see how interfaces are designed to combine the best of both worlds.

- Adaptive interfaces. The UI/UX of the last period of computing is largely a solved problem. There are standard UX solutions for most types of problems. It's also become significantly easier to build these interfaces, and LLMs are pretty good at writing basic declarative UIs. I think the bar will be raised such that users expect their interfaces to adapt to them, instead of a one-size-fits-all solution.

- Immersive interfaces. This might be similar to "dimensional" but more about the actual UX instead of just how 3D the buttons and icons are. I think using 3 dimensions is a natural solution for expressing higher information density. VR and AR will eventually catch on in some form.

Matz said he designed Ruby to optimize for developer happiness, it’s just a core principle of the language since it was created

Is this AGI?

I’ll know Apple is really in trouble when people stop complaining about each new product announcement

Maybe we’re all just in someone’s evolutionary chip designer

It’s weird in semiconductor physics too because the electrons flow uphill through voltage potentials

But why don’t they? Does anyone know what all these administrators do?

I’ve heard the theory that more regulation leads to more admin needs but I don’t think higher education has been increasingly regulated for decades.

If you just want React+Rails, the rails generator command comes with a bunch of options to set that up for you, including setting up and configuring: React/Vue/etc, a bundler like vite, typescript, tailwind.

It looks like inertia has additional features though.

Aren’t those examples deterministic? That’s the most interesting aspect of chaotic systems to me, they’re deterministic but still not predictable

I like SwiftUI, reactive frameworks just make sense to me for UIs. It also interfaces with UIKit in a pretty nice way so you don’t have to commit to one or the other. I’m a newer iOS dev though, I’ve heard many complaints from more experienced devs.

What do you mean? I thought the US dollar was fixed to gold between WWII and 1971:

> On 15 August 1971, the United States ended the convertibility of the US dollar to gold, effectively bringing the Bretton Woods system to an end and rendering the dollar a fiat currency

https://en.wikipedia.org/wiki/Bretton_Woods_system

Has anyone looked at competitor companies and compared the ones with WFH and the ones without it? I’m curious what that data looks like

I developed sleep apnea from poor posture and was able to fix it by correcting my posture. Developing a hunch or anterior pelvic tilt is especially a risk for desk workers and makes breathing harder, a little bit of yoga or posture correction exercises can save you a lot of trouble.

based on the naming conventions established by OpenAI and StabilityAI, this may be the most dangerous AI company yet

Claude does bullet lists a lot too, I think people just like bullet lists over blobs of text

> The prospects for such an enterprise have been enhanced by the relatively recent discovery that the moon harbors water in the form of ice, which could be used not only for drinking and irrigating crops but also for producing oxygen and making rocket fuel for future flights.

That’s really the reason?

Location: Chicago

Remote: Yes

Willing to relocate: No

Technologies: Ruby/Rails, TypeScript/React/Node, Go, AWS/Terraform

Résumé/CV: https://www.linkedin.com/in/choxi/

Email: roshan dot choxi at gmail

Hi, I'm Roshan and I'm a software engineer with 13 years of experience building creative solutions and solving scaling problems at fast growing startups. I've started my own company, worked with the Heroku founders on Xeroc PARC-style research projects, and most recently left a YC company (PopSQL) when we were acquired. I'm looking for early or middle stage startups working on interesting and ambitious problems!

Oh apologies, thanks for clarifying

Wasn’t Emad working at a hedge fund before this? I’m surprised the article glosses over the fact that he had 0 experience in the tech industry before starting one of the most ambitious tech projects ever

Or Google or Amazon, how are those not way more blatant antitrust targets?

There are dozens of other popular (group) messaging apps

You might be interested in Randall Carlson, he believes the megafauna extinction was tied to a global cataclysm and has speculated that it and its aftermath essentially traumatized human societies in a way that still persists today

Why do they do this instead of just maintaining the correct usage? The redefining of the word “literal” to mean “potentially not literal” really grinds my gears.

Has anyone tried doing that without LLMs? Could you make a GitHub repo under MIT license and just let people add a list of ideas to serve as prior art?

Interesting, thank you!

Any suggestions for creating training data? Did you just manually create your own dataset or did you use any synthetic methods?