jug

7 ms·

91.
by jug 1y ago
The worst thing with C pointers was for me that the asterisk is inexplicably used both to declare a pointer and a COMPLETELY different operation of dereferencing a pointer. I still don't understand this decision. I think it should&#x27
92.
by jug 1y ago
Or OAI-SearchBot for "web search" feature to augment queries and GPTBot for training? I swear I've even read how aggressive GPTBot is. Surely they aren't just googling stuff? https://platform.openai.com/d
93.
by jug 1y ago
Right, that's the sticky point? The likelihood of a planet in the Goldilocks zone to be too hot in the early stage of stabilizing its chemistry that it requires seeding with "post-formation" chemistry? Is that likelihood clos
94.
by jug 1y ago
I feel so bad for the climate now.
95.
by jug 1y ago
At least on OpenRouter, you can often verify what quant a provider is using for a particular model.
96.
by jug 1y ago
Even Sam Altman himself thinks we’re in a bubble, and he ought to have a good sense of the wind direction here. I think the contradiction here can be reconciled by how these tests don’t tend to run on the typical hardware constraints they n
97.
by jug 1y ago
Probably wrecked their AdWord views.
98.
by jug 1y ago
I initially found this project kind of fun, almost like an art project. But realistically, it’s easier to just use the web. Nothing about the web implies heaviness. If you want a minimalist website with no photos, you are absolutely free to
99.
by jug 1y ago
And the story of why HLT was backed out :) Microsoft had no trouble adding it -- even in Windows 95, but... https://devblogs.microsoft.com/oldnewthing/20030828-00/?p=42...
100.
by jug 1y ago
Hmm. Let's just say if this is true, that this is actually better with such a much lower total parameter count, it's the greatest accomplishment in over a year of LLM development. With the backdrop of bechmaxxing in 2025, I'l
101.
by jug 1y ago
Llama 4 is neither from last year nor a dense model.
102.
by jug 1y ago
> The Qwen3-Next-80B-A3B-Instruct performs comparably to our flagship model Qwen3-235B-A22B-Instruct-2507 I'm skeptical about these claims. How can this be? Wouldn't there be massive loss of world knowledge? I'm particular
103.
by jug 1y ago
Mistral teased Mistral Large 3 within weeks back in May but it never materialized. If this doesn’t do it, I don’t know what… I’m eager to see this one because Mistral models actually perform pretty well against top tiers in their class. It’
104.
by jug 1y ago
I've also thought about this. It has to be a terrible AI to scale like this and provide these instantaneous answers. And probably heavy caching too.
105.
by jug 1y ago
You should be able to sue Google for libel for this and disclaimers on AI accuracy in their fine print should not matter. It's obvious that too many people don't care about these to make these rumors reach critical mass and become
106.
by jug 1y ago
Yeah. The oldest known gag is a Sumerian proverb from 1900 BC: "Something which has never occurred since time immemorial; a young woman did not fart in her husband's lap" You also have another joke from 1600 BC: "How do
107.
by jug 1y ago
The tech report compares against DeepSeek R1 671B, DeepSeek V3 671B, Qwen3 235B which have been regarded as SOTA class among ”open" models. I think this one holds its own surprisingly well in benchmarks for using the nowadays rather, l
108.
by jug 1y ago
Should be coming up now.
109.
by jug 1y ago
I'd say it's more like comparing Sonnet 3.5 to Sonnet 4. GPT-4 was a rather fundamental improvement. It jumped to professional applications compared to the only causal use you could use ChatGPT 3.5 for.
110.
by jug 1y ago
No paywall: https://archive.is/TAU2L
111.
by jug 1y ago
It's a language that describes how to render a document. For example, you could describe how to render your own domain specific markup language as HTML. Which is then happily viewed by the browser as a regular webpage, that can of cour
112.
by jug 1y ago
Yeah, it truly IS transformative for industries, no denying anymore at this point. What we have will remain even after a pop. But I think AI was special in how there were massive improvements the more compute you threw at it for years. But
113.
by jug 1y ago
Someone needs to tell me why Grok 4 with Vision, a very powerful model, is at the bottom? In fact, the test depends on vision - yet all models perform poorly with that capability?
114.
by jug 1y ago
I’m always disappointed in journalism like this when they don’t ask the obvious questions, like in this case where the added cost for VW lies to not offer it part of the vehicle.
115.
by jug 1y ago
This sure took some time and is not really a unique feature. Microsoft Copilot has ended chats going in certain directions since its inception over a year ago. This was Microsoft’s reaction to the media circus some time ago when it leaked i
116.
by jug 1y ago
I don’t think they should be interpreted like that (if this is still about Anthropic’s study in the article), but the innate moral state from the sum of their training material and fine tuning. It doesn’t require consciousness to have a mor
117.
Commodore Vic-II (and C=64) Color Analysis (Preview)(unusedino.de)
5 points by jug 1y ago|0 comments
118.
by jug 1y ago
I was just thinking that. It has many, many errors. 1. Not seen browsing ”ai.dev”. 2. The text ”Imagen 4 is now generally available!” is spoken, not a comic caption. 3. Invalid second panel. 4. Hallucinates ”Meet Imagen 4 fast!” 5. Hallucin
119.
by jug 1y ago
I showed this blog post to ChatGPT and even an AI disapproved, lol.
120.
by jug 1y ago
It’s not free FREE but if you deposit at least $10 on OpenRouter, you can use their free models without credit withdrawals. And those models are quite powerful, like DeepSeek R1. Sometimes, they are rate limited by the provider due to their

More ›