comboy

Even though there are so many packed buttons and links it seems to like a vastly superior design compared to modern websites. I don't know if it's me being used to this or if we went wrong somewhere.

They are so hard to pronounce and remember to me that might as well be UUIDs

Does any popular NVR make a good use of LLMs (especially local models) getting decent at vision?

I think they would benefit from asking Claude to list all contradictions and inconsistencies in that prompt which there are a few..

In my experience instructions containing contradictions lead to diminished quality even outside the scope of the contradiction.

> What stopped them?

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

Not sure it works, isn't it already like that if you have the cursed javascript-only pages? It's basically proof of work to access their content (and if you are scrapping en masse you likely also need some LLM getting involved)

How CLI are you guys using for qwen and kimi?

> We should not sell powerful chips or chipmaking equipment to China

This is so short-sighted given that the US needs China equipment for.. everything. They are part of the supply chain needed for building the machines that build these very chips.

Switched from firefox few months ago, I don't like google, but firefox has only few percent of market share currenly, many pages simply do not work properly with it, plus it has bugs on macos (like onmousover stuff), which simply make it unusable. Safari is fine, but also many websites (which sometimes you need to use like banking or gov) are not tested on it. The overall browsers situation is less than ideal.

You don't have LLM-based processes if you think there is no lock-in. There may be no lock-in for coding if you enforce decent rules (but still some ambiguous docs can be interpreted differently), but any non-trivial pipeline/system, these models are not stupid but each has some quirks. Sometimes for some reason they will ignore some instruction while all other models have no trouble following it. These things accumulate.

Plus there's subjective stuff even for coding, people learning how to deal with it. Even on HN you can already see cloude/codex camps each strongly convinced that one is better than the other.

Incentives.

Guy says he created a drug to treat Alzheimer's disease. Huge difference.

But it seems that there is some queuing/load balancing on their side, I mean when opus is actually outputting this 55t/s it feles fast, but apart from it's internal reasoning I think there's sometimes just waiting.

We are running out of things to break.

I mean, Codex CLI is really bad. But Claude's CLI is so much worse.

Welcome to the world of tomorrow!

They all have everyone ID's through payments already..

doing 50 API queries to Sonnet of this length is not that expensive..

I made Qwen respond it was made by Google with a simple Chinese greeting.

But also, I made Sonnet introduce itself as made by OpenAI..

Prompt: 你好！用一句话介绍你自己。

Sonnet in around 5% of resplies:

    你好！我是 **ChatGPT**，一个由 OpenAI 开发的 AI 助手，致力于回答问题、提供信息和帮助解决各种问题。有什么我可以帮你的吗？

Found it like a month ago and it kept working, I wonder if it will stop after this comment.

What's that about? It's full screen/terminal anyway. Is it just switching some rendering engine under the hood?

If you think that your phone number is equivalent to your home address then yes.

Different network layer, no centralization, no authorities, DNS has nothing to do with making p2p connections, it's like the ballpark is not even in the same country

I'm so disappointed in this comment thread https://en.wikipedia.org/wiki/OSI_model

I've just learned about it, but my understanding is that Iroh is L7, compared to e.g. tailscale which is L3

Prompt matters. Obviously if you want another model opinion you must generate from the scratch using the same prompt and then you can try to synthesize, but working with an existing response can work if desired. I use explicit instructions to find issues with assigned severities and then these are going through the panel of judges, only issues passing certain threshold are fixed in the original response.

I'll share a revelation which vastly improved my results: tell judges to evaluate truth and usefulness/should-be-fixed axis separately. Because inevitably with a prompt that is forcing to find issues you will end up with nitpicks. Plus truth axis allows to better evaluate the issue-finder models for your use case.

That's some part of what happens when I generate explanations like this one: https://hanzirama.com/character/%E6%9D%A5#explain - at this point the site is a small side product of my LLMs-evaluation machinery.

Bonus content for patient readers: if you need top quality you will likely need to pin provider(s) on OR, :exacto is not enough to get good repeatable results especially for open-weights models.

They cannot do it. Apart from all the practical, technical and talent reasons, it would still be exporting forbidden stuff.

The signal is clear enough though for the next Anthropic..

It's simple, marketing dominates everything. With attention being very expensive, appearance is what matters.

It doesn't matter if you write fantastic library, nobody is gonna use it because they won't know about it, the one with a gif of the terminal (ffs) will win that has a good page describing what it does (and being the most popular one can even become better than your library because of the following but that's not the point here).

It's everywhere, products, hiring, services. We have no network of trust (sigh), we need to trust some heuristics based on a shallow information. If somebody focuses on the shallow he wins, because nobody can ever dive into everything.

I'm creating hanzirama.com

I generate explanations for characters and words like so: https://hanzirama.com/character/%E6%9D%A5#explain

But I don't want to mislead learners and want to provide some cultural depth, so I have a hole sophisticated pipeline, using multiple models to generate the explanation, then multiple models look for issues in the explanation, each issue goes through the panel of judges (basically trying to squash down any hallucinations), it's fixed and it goes through such cycles a few times over.

I've been at it for some months now, so I have dozens of different probes, that I needed to evaluate prompts and method changes. Plus on some items I generated so many explanations through different means that I can tell a lot about given model just by looking at one.

Plus I'm doing some statistics, so I see how e.g. when working as judges of issues some models correlate heavily with some others... Fun fact during some testing runs basically just testing providers I stumbled upon qwen introducing himself as made by Google. And also Anhropic's Sonnet saying that it was made by OpenAI :)

At this point all my evaluations frameworks and pipelines stuff is much bigger than the site itself. I'm having lots of fun though.

But it sounds like FableFool so it has that going for it.

There is in /config "Switch models when a message is flagged" now which can be set to false, but I had no chance to see what happens then, does it just stop or what.

'by the way, your previous attempts have these structural problems."

Just to be clear, it did not have access to any previous work that opus did? Because they are pretty good at digging out relevant tmp files and making use of whatever is out there.

With my fable adventures I caught it hallucinating something and stating it as a fact in CLI twice. And it was something that I did not see opus do in such way, opus obviously many times stated some things that it did not verify but guessed, but fable said something like "the probe showed that ..." - but there was no probe, it was not about some past events it was about what it was doing right now. "I overstated"...

But boy does it know Chinese, so much better than any other english model, gemini used to be the king but fable clearly was trained on a decent amount of it. It has a deep cultural understanding.

I kind of enjoy exploring black boxes, trying how different inputs are mapping to differences in outputs. It's kind of like hacking. The problem is, they keep altering the box.