nico

The other day I saw this project to flash your TV with custom software: https://www.tiktok.com/@itstundealao/video/77013919041931215...

Here are the repos mentioned: https://github.com/Gathin23/cleanyourtv

https://github.com/simbasjungles/free-my-tv

The video is pretty cool. The guy flashes his tv with a custom firmware with a look/feel and elements he chose, plus a feature he created to leave handwritten notes for his girlfriend on the home screen of the TV, remotely from his laptop

Also checkout the Wikipedia-style live docs here: https://id-le.com/pedia?halloween

Everything about Id|le is documented there, the creatures, their family trees, the way naming works, their abilities, the environment and music controls, the operator, and pretty much every feature of the sim

Oh, and a fun feature to play with, pull in content from any website to be fed to the creatures: http://id-le.com/?from_wiki=https://en.wikipedia.org/wiki/Y_...

(the above param only works for wikipedia links, but you can load up any website from inside the Studio > Feed the world feature onside the sim)

Also checkout: https://id-le.com/pedia

Open source here: https://github.com/nicobrenner/id-le

That’d be pretty funny actually

But I don’t think you have it always listening, you probably tab the space bar to talk and send a prompt, walkie-talkie style, push to talk

Is this the open source whisperflow clone? I think I saw a positive review about it yesterday on social media

It’s interesting that it’s supposed to be very bad to destroy evidence, especially after litigation has started. But in the end, usually the consequences are better for the person destroying the evidence, than actually preserving it

If the evidence was very damning, and it’s not there anymore, the judge/jury can’t really properly make a determination about it if it’s gone

I’m not a lawyer, so someone with more legal experience could pitch in, but that’s my impression from seeing these big corporate cases on the news, as well as some personal experience in smaller civil cases in court

Nitpicks about CLI coding agents:

Many times, while they are doing something, I’ll be typing a prompt, and then all of a sudden I get a question with Y/N options, and by accident, I’ll type Y/N as part of a word thus “answering the question”, many times either steering in the wrong direction or stopping the agent

Neither Claude code nor Codex solve this. The only way I’ve found to prevent it is to type somewhere else and then paste, but it’s super annoying

Edit: ^^ the above is especially annoying now that Codex popups the prompts/questions within other agents' sessions

Also, when I want two agents in different sessions to talk to each other, it’s kinda janky, I have to talk to both and act as a messenger for a couple of turns before they properly coordinate. I wish I could just sort of click somewhere, and link unlink two or more sessions

Does Próspero solve the things above?

Any pans for “swarm”/group control of agents?

Thank you for the insightful explanation

> And since in many agents you're also going to want to display actual HTML, unless you're going to embed a browser engine in your C++ library, that's just not doable.

I guess this is the main issue. Either you end up reimplementing a browser (which will def have a ton of issues), or you embed a browser and you end up with bloat anyway

> and all the other stuff that C++ frameworks are terrible at, but which is a doddle in HTML/CSS/JS

This is very interesting. I know that coding string-related stuff with C/C++ can be incredibly tedious and error prone (even hard security-wise). But all the stuff in the web-related stacks/libraries in other languages, is either already somehow being transcompiled to C to execute, or could be relatively easily ported/reused by directing an LLM/agent to do it

Do you think that could work to make C/C++ better at this? What things would you import/copy from elsewhere?

> nesformer builds on Qwenputer by Divyansh Agrawal, which compiled a CHIP-8 and then a RISC-V CPU into Qwen3.5-4B. It reuses Qwenputer's compiler and its central trick: the model's Gated DeltaNet state is the machine's memory

This is amazing, love it! Qwenputer is very cool too

Very impressed with both the cool idea and visuals, as well as the technical details. Had no idea you could do animations in Latex/PDF. So beautiful! Thank you for sharing

This is so trippy, very cool, love it

I made a classifier play doom[0]. But you turned the whole network into Doom! Amazing

0: https://playground.jeffyclassify.com/#doom

> There are tools out there that claim to be able to tell the difference between LLM-written prose and the old-fashioned biological product. I'm wary of them, partly due skeptism about their accuracy, but also because it rather misses the point

HN uses automatic classifiers. They seem pretty effective. At the same time, I think it's a loosing battle. I'd rather HN either flag for others to ignore/filter/review if they want to, and/or allow people to self-report (ala TikTok)

Eg. I wouldn't mind if a comment on HN was like:

Hey everyone, I coded this little thing:

<ai-generated> some code </ai-generated>

Then users can choose if they want to see posts with AI-generated stuff or not, or something else...

Of course there are always going to be some people trying to game the system. But what are we optimizing for? (a lot of times the 95% unnecessarily suffers because of the mitigation efforts to prevent the other 5% from abuse)

Awesome... in the same spirit, here's a cursed source control system for it: https://nicobrenner.github.io/lossy-git/demo.html

Or the cli version if preferable: https://nicobrenner.github.io/lossy-git/

I do want people to independently navigate. But also, I'd like to bring them back to the current slide, or somehow highlighting the current slide, essentially making it easy for viewers not to get lost

Treat it more like a cool proof of concept, now you can vibecode your own and use it with your own data

The feature is very cool and potentially useful, but: it only works on certain types of meetings/calls, and they way it’s implemented, people get lost, because they start playing with their slides, and then it’s not clear you can press the “Live” button to sync up again, so then they end up interrupting the speaker to say they are lost, or multiple people will say they are not clear where they are…

I love this, and I feel every legal document (of which there are many, for example, a random one: insurance companies' policies) should have something like this (and no, Word is not good for it)

Does it already exist?

That's very cool. I'm curious if you need a GPU for your use case (email and text triaging). There are some open source alternatives[0] that run/train on CPU, so you can self-host or run locally (and train on your own data for better accuracy), would something like that work for you?

0: http://jeffyclassify.com/

Awesome, I didn't know qmd style search, thank you!

I'm all up for using the better/simpler task for the job. That's been a thing lately also with the Jev/decision models hype, most of those things can either already be done by a pretrained classifier[0], or you can start with a bigger model and then train your smaller/adhoc one when you have some data (it doesn't even take much for most applications, you can get about 95% accuracy on email classification with 50-100 examples)

0: http://jeffyclassify.com/blog/when-llms-fail-at-classificati...

Not sure about how the code in this project works. But the way I see it, an LLM+chat is a great interface for information. If you then connect it to a db with the knowledge/articles you need, then you get a great way to navigate it, without needing the model to memorize all the info

Love this. And as a trend, I'd love to see more local/small task-specific models/agents. Experts at helping you with that one thing, leaving the everything-and-the-kitchen-sink to the frontier models/agents

Thank you for making Ghostty, it’s a pleasure to use it everyday for work :)

I’ve decide that the best (and maybe only) way for me to interact with google AI models, is on the search results

It’s where I get the most value, and my expectations are very low, no attachment if it goes away

Of course they are mining me for data anyway, but they were doing that before already (and I do have a Gmail account, although I’ve almost completely stopped using Chrome - btw, it’s crazy that you can’t add plugins/extensions to Chrome on Guest mode)

> decent way to make a product is to do the hard part for someone else.

100% this is a very good insight

I just ran an experiment, I took the banking77 dataset, took the labels out, and got an LLM (Opus 5) to label them, then trained a logistic regression classifier (Jeffy http://github.com/nicobrenner/jeffy) on the resulting set

Got the same accuracy and f1 from the Opus-label dataset than the original one (I realize that Opus might have banking77 in its training data, so might need to account for that somehow)

But the point is: it's way cheaper to run the LLM model over your data to label once, and then use a cheap classifier, than using the LLM directly for classification

> and not using in anger

Is this a typo? are you saying that this is so new people haven't struggled with Jev enough yet? very interesting

> it starts to break down over 20 classifications

Yes, I recently ran a benchmark on exactly that: https://jeffyclassify.com/blog/when-llms-fail-at-classificat...

At around 4 classes already LLM-based classifiers (Jev/Laya/GLiNER et al) start falling behind

> For any application at scale it'd be way more accurate to have a dedicated classifier using the right architecture

Agreed, and that’s the premise behind Jeffy, open source, bundles pretrained task-specific models that run (and train) locally on CPU: https://jeffyclassify.com/

At the same time, I’ve been working with some people implementing it, and there’s definitely value in having a more general model like Jev, or an LLM, for initial exploration and data labeling, to then use that data and train the task-specific model. It’s a whole workflow

Also, the market/pain insight of Typesafe is super interesting. They realized that a big chunk of the market was using LLMs as classifiers, and they built a solution around that (very much to your point, a task-specific model, rather than a more general LLM), and it worked! (Also they have the ability to generate a lot of attention, given the founder being ex-OpenAI and having raised quite a bit of money too), so it’s not just a technical/engineering thing, it’s a business/market/need/discoverability issue as well