It's not goblins, it's honest!
The other ingredients would be doing other things: making the pill/drug easer to swallow/consume, extending shelf-life, etc. You need enough of the drug for it to be effective, but not too much to overdose or exhibit…
Yes that clip [4] is making the rounds again in light of Mythos, but his stance (and that of others) hasn't changed ([3] is from a new interview). [1] https://memeburn.com/amodei-says-open-source-ai-is-becoming-... [2…
LLMs grading the answers is relying on the LLM knowing the answer and not just hallucinating it. You also have issues if/when the model refuses to answer, or if it gets stuck in a loop (e.g. if running locally with a…
I'm not sure if they've fixed this, but older models have a tendency to ignore negation as `no`, `not`, etc. all occur frequently in the training data so are weighted less strongly than the verbs and nouns. The advice…
This approach is effectively seeding the context with how you want the LLM to behave/operate ("senior reviewer", i.e. the style of the responses you want) and the context/domain in which the LLM is operating in…
You're not reading someone else's RP, though. Not with the applications that support this. There are things like the character(s) you interact with, the initial setting, and some background information predefined, but…
Tell that to the LLM RP crowd.
A 27B model can fit easily on a 32GB VRAM card (e.g. 5090) or a 32GB computer in RAM at FP8/Q8 (unsloth have 28.6GB Q8 files). For 24GB VRAM cards (e.g. 4090) you can use Q6_K (22.5GB) or Q5_K_M (19.5GB) quants,…
A game company (https://www.zachtronics.com/) that have made a series of games where you either build machines or write instructions to solve a task. An example is Opus Magnum where you convert input elements to output…
If you don't want the thinking, you can pass `enable_thinking: false` to the `chat_template_kwargs`. If using promptfoo, this can be done via: providers: - # llama-server id: openai:chat:qwen config: apiBaseUrl:…
Dublin Core is effectively similar/related to schema.org's CreativeWork. If you have a creative work (audiobook, short story, news article, etc.) then Dublin Core is applicable, in addition to the corresponding…
Sandi Toksvig, the current host of the BBC program QI (Quite Interesting), previously hosted by Stephen Fry. She's also been on a number of other BBC TV and radio shows.
It could be based on things like word frequency. I'd expect obfuscate/obfuscation to be less common outside of programming and RPGs (Vampire the Masquerade).
Norman French due to the Norman invasion of 1066 resulting in Old English evolving into Middle English. You can see that in the words for animals vs meats (cow and boef/beef, sheep and mutton, etc.) where the Germanic…
I use promptfoo for evaluation. I'm experimenting with tests for my workflow/use cases. I have a custom assert for loop/repeat detection that works well: def count_repeats(text: str, length: int) -> int: n = len(text)…
llama.cpp is great. However, Llama 4 was a misstep for them: it was too big, so was out of reach of the LocalLlama crowd and hard to train/customize into different variants like has happened with the smaller models on…
It's all relative. For local use I'd classify it by hardware (VRAM size) using FP8 or Q6 quantization: 1. tiny <2-3B -- easily runnable on lower-spec hardware 2. small 4-8B -- runnable on 8GB GPUs 3. medium 9-12B --…
Yeah. I run LLM models locally and for me 22B-32B is the largest I'm willing to invest in trying out. Even though Mistral 4 has 6B active parameters per token (allowing 3-3.5 per token parameters to be loaded on a…
post != article From https://news.ycombinator.com/item?id=48281764: > > ARIA can help when devs want to use the wrong elements for some reason or for custom controls. > But it can't. See this article.
You said "see this article" re: how aria-label is not applicable to div elements, hence the second link which is the WAI-ARIA guide on labelling elements. You also said that ARIA can't help with custom controls in that…
NVDA is a free screen reader for Windows (written by blind devs) that works with Firefox and Chrome. You don't need to pay for a specialist browser as all web browsers (Firefox, Chrome, Edge, Safari, etc.) will…
See https://www.w3.org/WAI/ARIA/apg/patterns/ for a guide on how to create accessible markup for custom controls and the associated examples. See specifically…
IIRC, libraries like numpy and pytorch can already do that as they store the matrices as 1D arrays with information on things like the stride length (advancing to the next row). That allows you to implement operations…
That goes back to Ken Thompson's NFA regex interpreter from 1968 [1], [2], [3]. Note: that whole regex series by Russ Cox [4] is great. [1] https://dl.acm.org/doi/10.1145/363347.363387 -- Programming Techniques: Regular…
It's not goblins, it's honest!
The other ingredients would be doing other things: making the pill/drug easer to swallow/consume, extending shelf-life, etc. You need enough of the drug for it to be effective, but not too much to overdose or exhibit…
Yes that clip [4] is making the rounds again in light of Mythos, but his stance (and that of others) hasn't changed ([3] is from a new interview). [1] https://memeburn.com/amodei-says-open-source-ai-is-becoming-... [2…
LLMs grading the answers is relying on the LLM knowing the answer and not just hallucinating it. You also have issues if/when the model refuses to answer, or if it gets stuck in a loop (e.g. if running locally with a…
I'm not sure if they've fixed this, but older models have a tendency to ignore negation as `no`, `not`, etc. all occur frequently in the training data so are weighted less strongly than the verbs and nouns. The advice…
This approach is effectively seeding the context with how you want the LLM to behave/operate ("senior reviewer", i.e. the style of the responses you want) and the context/domain in which the LLM is operating in…
You're not reading someone else's RP, though. Not with the applications that support this. There are things like the character(s) you interact with, the initial setting, and some background information predefined, but…
Tell that to the LLM RP crowd.
A 27B model can fit easily on a 32GB VRAM card (e.g. 5090) or a 32GB computer in RAM at FP8/Q8 (unsloth have 28.6GB Q8 files). For 24GB VRAM cards (e.g. 4090) you can use Q6_K (22.5GB) or Q5_K_M (19.5GB) quants,…
A game company (https://www.zachtronics.com/) that have made a series of games where you either build machines or write instructions to solve a task. An example is Opus Magnum where you convert input elements to output…
If you don't want the thinking, you can pass `enable_thinking: false` to the `chat_template_kwargs`. If using promptfoo, this can be done via: providers: - # llama-server id: openai:chat:qwen config: apiBaseUrl:…
Dublin Core is effectively similar/related to schema.org's CreativeWork. If you have a creative work (audiobook, short story, news article, etc.) then Dublin Core is applicable, in addition to the corresponding…
Sandi Toksvig, the current host of the BBC program QI (Quite Interesting), previously hosted by Stephen Fry. She's also been on a number of other BBC TV and radio shows.
It could be based on things like word frequency. I'd expect obfuscate/obfuscation to be less common outside of programming and RPGs (Vampire the Masquerade).
Norman French due to the Norman invasion of 1066 resulting in Old English evolving into Middle English. You can see that in the words for animals vs meats (cow and boef/beef, sheep and mutton, etc.) where the Germanic…
I use promptfoo for evaluation. I'm experimenting with tests for my workflow/use cases. I have a custom assert for loop/repeat detection that works well: def count_repeats(text: str, length: int) -> int: n = len(text)…
llama.cpp is great. However, Llama 4 was a misstep for them: it was too big, so was out of reach of the LocalLlama crowd and hard to train/customize into different variants like has happened with the smaller models on…
It's all relative. For local use I'd classify it by hardware (VRAM size) using FP8 or Q6 quantization: 1. tiny <2-3B -- easily runnable on lower-spec hardware 2. small 4-8B -- runnable on 8GB GPUs 3. medium 9-12B --…
Yeah. I run LLM models locally and for me 22B-32B is the largest I'm willing to invest in trying out. Even though Mistral 4 has 6B active parameters per token (allowing 3-3.5 per token parameters to be loaded on a…
post != article From https://news.ycombinator.com/item?id=48281764: > > ARIA can help when devs want to use the wrong elements for some reason or for custom controls. > But it can't. See this article.
You said "see this article" re: how aria-label is not applicable to div elements, hence the second link which is the WAI-ARIA guide on labelling elements. You also said that ARIA can't help with custom controls in that…
NVDA is a free screen reader for Windows (written by blind devs) that works with Firefox and Chrome. You don't need to pay for a specialist browser as all web browsers (Firefox, Chrome, Edge, Safari, etc.) will…
See https://www.w3.org/WAI/ARIA/apg/patterns/ for a guide on how to create accessible markup for custom controls and the associated examples. See specifically…
IIRC, libraries like numpy and pytorch can already do that as they store the matrices as 1D arrays with information on things like the stride length (advancing to the next row). That allows you to implement operations…
That goes back to Ken Thompson's NFA regex interpreter from 1968 [1], [2], [3]. Note: that whole regex series by Russ Cox [4] is great. [1] https://dl.acm.org/doi/10.1145/363347.363387 -- Programming Techniques: Regular…