picardo

No thanks. If I have to listen to Mr. Meeseeks' screechy voice all day, I'd shoot myself.

I don't know why I didn't of this before. Thanks for the suggestion.

I tried it with Claude Code for a while but lack of WebSearch tool became a dealbreaker for me. Does anyone know of they will provide support for it?

They could have saved themselves 3MB by converting it to AVIF.

For most static UI surfaces, I probably wouldn't use it, but I can see a use case in this for testing generative UI workloads.

I'm curious what a "stripped down version" of Github can offer in terms of functionality that Github does not? Is it not simpler to have the agents register as Github repos since the infrastructure is already in place?

The key idea is not running constantly, but being always on, and being able to react to external events, not just your chat input. So you can set a claw up to do something every time you get a call.

Is AssemblyAI or Deepgram compatible with OpenAI Realtime API, esp. around voice activity detection and turn taking? How do you implement those?

This reads like a classic case of bureaucratic mission creep. When the budget depends on showing numbers, and the definition of a "threat" is loose enough, the agency naturally evolves into a political tool. The lack of uniform identification and masks just removes the final psychological barrier of accountability for the individual agents.

I'm curious how the unit economics actually play out here compared to traditional search. With Google, the compute cost to serve a query is negligible, so even low-CPM ads are profitable.

With an LLM, the inference cost per query is orders of magnitude higher. Unless thy have a way to command significantly higher CPMs -- perhaps by arguing intent signal is stringer in a conversation than a keyword search -- it feels like a difficult margin to sustain.

This man is a legend. He's been working on making a better marble machine for more than a decade. The only payoff is the satisfaction of listening to the synchronized syncopation of falling marbles.

Here is a more recent video: https://www.youtube.com/watch?v=k8NXF2rtaEg

Happy thanksgiving all. In an era where algorithms on other platforms seem optimized for outrage and engagement bait, I'm grateful for HN's optimization for curiosity. It's one of the few places left where I can open a thread on a topic I disagree with and actually expect to have my mind changes -- or at least understand the opposing view better -- by the top comment.

You're right about that, but that wasn't common practice at the time. We learned about side-effects from Elm and Flux.

Isn't this just how the DOM works? Data flows down through attributes and properties; events bubble up?

That's right, but this communication pattern causes serious complexity. Imagine trying to find out what triggered a state change. You would have to listen to every event source to find out. With Flux, all state changes were mediated by the reducer in the store. It made things a lot simpler.

Didn't they call it Flux rather than Flow?

Ah, you may be right. It's been a long time.

For massive apps with 1,000 components on the same page, maybe React's complexity is justified. But what the other 99% of apps?

The number of components is not the only yardstick of complexity. Most of the complexity in building a UI comes from state management and how state changes are propagated across the store and the UI.

I worked with Backbone for many years, and I can distinctly recall the hours of frustration I had debugging a UI because it was freezing due to cascading state changes. That was because we were using Backbone Store, which had bidirectional data flow, and when one updated the store, it would trigger a change to the UI, which would change the state store, which would change the UI, etc.

You could argue that the real innovation of React was "unidirectional data flow," but React team made Flux architecture central to the framework, making it easier to adopt good practices, whereas Backbone remained store agnostic and even encouraged Backbone Store which used the observer pattern for many years. I think you should choose a framework that allows you to fall into the Pit of Success, and React was that framework at the time, and for my money, it still is.

This should be higher. While the research question is interesting, the sample size makes the conclusion highly suspect. I'd like to see more research on this.

Thank you for your hard work. I gripe about it sometimes but it’s still the most compelling learning experience I’ve ever come across. I hope it keeps getting better and better.

Video explainers, for me, are better than the audio overviews.

I use NotebookLM everyday. The simplicity of the design is much appreciated. However, there are real issues scaling the design and keeping it user friendly as the team keeps adding new features.

The most recent example of this is with the addition of 2 new capabilities (Flashcards and Quiz), "Artifacts Button Container" now has 6 large buttons, and is 328px in height! There are users who are accessing the site from small screen devices in India and they have been asking for help on Discord forums because they cannot see their notes anymore. So I had to create a Tampermonkey script to let users collapse it.[0] I heard the team is fixing that soon, but they should have done more testing before releasing it.

There are other issues like this that I've fixed with scripts. The strangest one is the "notes." Why force the users read a 2000 word essay in a 360px sidebar? So I wrote a script I wrote to let you pop it into full screen mode.[1]

Another example is the chat input field. The follow up questions are hardly usable at all. And they're not stable after you select them.

I can go on all day, but I think it's better to fix things than to complain.

[0] https://gist.github.com/volkanunsal/94db50629cad816eca84c836...

[1] https://gist.github.com/volkanunsal/fded9124d62422c0d2672b8a...

I'm regretting turning off my Ad Blocker. Do not click on the ads, folks -- especially if you're at work.

Last time I checked the context window for the embedded Gemini Nano was 1024 tokens. I hope they have reconsidered that limitation.

I'm excited by this multi-media approach. I've been avid self-learner for years, and I've often found the textbook format too dry, and forbidding, but ever since I started using NotebookLM, I'm diving into textbooks more and more. There is genuine value in creating a new format that meets learners where they are at.

Well, at least the growth mindset study is not fully debunked yet. It's basically a modern interpretation of what we've known to be true about self-fulfilling prophecies. If you tell children they are can be smart and competent if they work hard, then they will work hard and become smart and competent. This should be a given.

For context: https://chatgpt.com/share/68bc71b4-68f4-8006-b462-cf32f61e7e...

haha, I believe you!

131,484 people sought mental counseling from Claude in this time period. To put it into context, that's more than the number of psychologists in the US. This number points to the scale and impact of AI systems in making mental health more accessible to all.

I do that sometimes, but I get peer pressured into drinking alcohol if I socialize. You can't win.

Agree with you about the benefits of avoiding alcohol.

Oh snap.

Is it normal for people to be drinking coffee so late in the day?