wren6991

6 ms·

61.
by wren6991 3mo ago
"Now I have the full picture. Let me just undo that last change." $ git checkout . "Oh no, that undid everything."
62.
by wren6991 3mo ago
Another banger from the mind that brought us Beads, the solution to having your repositories littered with Markdown files, written in a svelte 250,000 lines of Go. What's that? The Beads repo itself is littered with Markdown files? And
63.
by wren6991 3mo ago
I guess they are saying sizeof(...) should return the annotated static size rather than sizeof(int*) in this case
64.
by wren6991 3mo ago
Opus 5 clearly frogmaxxed. gemini-3.6-flash runs 2 and 3 responded best to the royal portrait context.
65.
by wren6991 3mo ago
> (everything here also applies to c2y's newly introduced _Countof) Stop the press, they're adding what ? So I don't have to keep adding #define count_of(arr) (sizeof(arr) / sizeof(*arr)) to every codebase and living
66.
by wren6991 3mo ago
Trying OpenRouter gave me a strong impression that LLMs are already a commodity and individual providers are not that important. Providers don't seem to be too happy about it.
67.
by wren6991 3mo ago
As a frog I'm happy to report the water is comfortably warm today, the same as it was yesterday.
68.
by wren6991 3mo ago
Huh, so every interactive event has a network request in the loop, with perceptible latency? And you still end up obligating client-side JS for the plumbing? I'd been meaning to look into HTMX but based on the content of this post it s
69.
by wren6991 3mo ago
I don't think so. Probability of what? If you put something through a softmax the output is (trivially) a valid PMF. Does that matter? You're not sampling from it.
70.
by wren6991 3mo ago
The worst part of this workflow is when you accidentally run `git commit -am` at the end and destroy all of your careful staging.
71.
by wren6991 3mo ago
This seems functionally similar to OpenAI having a step in pricing once you exceed a certain context length (also at 272k aka 2^18 aka 256k). Having a lot of active context increases the per-token cost (flops issued and bytes read per token
72.
by wren6991 3mo ago
I'd guess because it costs them nothing and it gives you a smoother transition back towards paying for their products.
73.
by wren6991 3mo ago
Ooh. This looks like an interesting paper and there were a couple of things in the intro that I found counter-intuitive. It'll take me a while to digest the whole thing. > Coverage and perplexity analyses show that the observed reas
74.
by wren6991 3mo ago
That's true. The fact that an LLM is a pure function of (all previous tokens) -> (next token), with internal state like KV cache only existing for optimisation purposes, is pretty mind-blowing. I guess it was more the "predicto
75.
by wren6991 3mo ago
The "next-token predictor" framing is also a bit shaky. It's an accurate description of pre-training, where next-token prediction is a useful learning objective to force the model to learn higher-level representations. It&#x2
76.
by wren6991 3mo ago
Sounds a lot like running the Qwen3.5/3.6-series models at home: you need checkpoints for the recurrent state (GDN in the case of Qwen). You avoid the miss for the common case of 100% prefix match (e.g. during interleaved tool calls an
77.
by wren6991 3mo ago
> Anthropic has never advocated for a ban on open-weights models. What are the legal ramifications of this statement if it turns out Anthropic have lobbied for this? Does it just get swept under the rug? I can't say this is bullsh
78.
by wren6991 3mo ago
Older models did show the full unredacted thinking trace, but I don't think Opus has ever shown full CoT. Here is an archived version of Anthropic's API docs saying that Sonnet 3.7 (only) has unredacted CoT on API: https:/&#
79.
by wren6991 3mo ago
The main source I'm aware of is this Reuters article: https://www.reuters.com/world/beijing-is-looking-curbing-ove... However you should take it with a pinch of salt because IMO their direct quotations do not supp
80.
by wren6991 3mo ago
I think the type of persistence rewarded by benchmarks may be misaligned with instruction following
81.
by wren6991 3mo ago
The minimum fix would be to read it once per user turn, rather than once per tool call turn. As it is, if you tell an agent to update the file and it decides to do multiple edits on separate turns, you take a full cache miss per edit . If
82.
by wren6991 3mo ago
> some things mention are outdated - particularly the tool call pruning feature. this has been disabled by default for a while specifically because of complaints like this Thanks, I added a small update section near the top to call this
83.
by wren6991 3mo ago
> and you need to hash chunks of tokens in order to look up in these, meaning you need to be able to slice up your tokens by token count. The hash just has to uniquely identify the contents. I still don't see what stops you from wal
84.
by wren6991 3mo ago
> It's all above board though, as you, the user, agreed to the terms & conditions for installing said app. I have an LG TV, about 6 years old now. For a while it was connected to the internet. Every time I turned it on, after a
85.
by wren6991 3mo ago
> I'm not sure about the proprietary inference engines, but in the open source ones tokenization is done before looking up if a text sequence is present in the KV-cache Is this necessary? Tokenisation is deterministic, so for a hit&
86.
by wren6991 3mo ago
I like the explicit and imperative style of this prompt. Makes me feel like LLMs are just weird Turing machines and CoT is their tape.
87.
by wren6991 3mo ago
Having watched Qwen kill its own llama-server instance to free up a port, I think this is a bold presumption and you should test it at your earliest convenience.
88.
by wren6991 3mo ago
Title is editorialised. Here is one editorialised in the opposite direction, for balance: "OpenAI model breached HF, meanwhile OpenAI model safeguards refused to help HF's defense."
89.
by wren6991 3mo ago
Author here. Just wanted to say: thank you for posting this. You've given me something to think about, and I regret writing this post the way that I did.
90.
by wren6991 3mo ago
It felt unreasonable when I was using it at midnight and had to wait 10 minutes to refill the KV cache on my local GPU :-) A simple solution here would be to evaluate the date once per session, or once each time the `opencode` binary is lau

More ›