nl

26 ms·

511.
by nl 6mo ago
Second model to get 25/25 on my benchmark (after Opus 4.7): https://sql-benchmark.nicklothian.com/?highlight=openai_gpt-... Cheaper and slower than Opus.
512.
by nl 6mo ago
The Pro model is giving 429 Overload errors
513.
by nl 6mo ago
The claim that APIs are subsidized is very common. eg: Token prices are significantly subsidized and anyone that does any serious work with AI can tell you this. https://news.ycombinator.com/item?id=47684887 (the claims
514.
by nl 6mo ago
Amazingly, the current depreciation overestimates the retained value of GPUs. In 2023, the depreciation schedule for H100s was 2 years, but they are still oversubscribed and generating signficant income. Coreweve has upped their depreciatio
515.
by nl 6mo ago
> upload and restore it when the user starts their next interaction The data is the conversation (along with the thinking tokens). There is no download - you already have it. The issue is that it gets expunged from the (very expensive,
516.
by nl 6mo ago
> ~/.config/uv/uv.toml > exclude-newer = "7 days" Note the if you get failed to parse year in date "7 days": failed to parse "7 da" as year (a four digit integer): invalid digit, exp
517.
by nl 6mo ago
It doesn't look dissimilar to the Apple Watch Ultra plus a Casio GShock/Garmin Fenix vibe
518.
by nl 6mo ago
> the pro and flash variants are 5x to 10x smaller than opus and gpt-5 class models. The rumor is that Gemini Pro is the largest model being served today (or at least was prior to Mythos) Source: some podcast where they were discussing
519.
by nl 6mo ago
Strix Halo version available for pre-order!
520.
by nl 6mo ago
> The entire purpose of LLMs is to be non-static: they have no deterministic output and can't be validated the same way a non-LLM function can be. Adding another LLM layer is just adding another layer of swiss cheese and praying the
521.
by nl 6mo ago
> We’re supposed to be fixing LLM security by adding a non-LLM layer to it, If people said "we build a ML-based classifier into our proxy to block dangerous requests" would it be better? Why does the fact the classifier is a LL
522.
by nl 6mo ago
$30B ARR says otherwise.
523.
by nl 6mo ago
It's a better model in my usage. I have benchmarks.
524.
by nl 6mo ago
> So how would you explain the increase in token usage, considering the fact that conventionally tokenizers are trained to minimize the token usage within a given vocabulary budget? Just modeling whitespace as its own token would seem to
525.
by nl 6mo ago
This is almost certainly wrong. Case sensitive language models have been a thing since way before neural language models. I was using them with boosted tree models at least ten years ago, and even my Java NLP tool did this twenty years
526.
by nl 6mo ago
I don't believe route limitation for dangerous goods is a thing. I looked on https://www.iata.org/en/publications/newsletters/iata-knowle... Interestingly I asked both Claude and ChatGPT "does the I
527.
by nl 6mo ago
https://academic.oup.com/jiel/article/26/2/233/7071723 One case study involves economic coercion. (Disclosure: I worked with the author, although not on this paper)
528.
by nl 6mo ago
love your work on this!
529.
by nl 6mo ago
It's irrelevant because no matter if the system memory is unified or not the point of the article is if WASM adds extra memory copiy. That's the same no matter the physical memory system architecture.
530.
by nl 6mo ago
I'm pretty sure this is just "yes (parts of), memory control in WASM works"[1]. The whole Apple Silicon thing is (in this case) just added details that don't actually matter. [1] https://github.com/WebAss
531.
by nl 6mo ago
This is a bad take. It's not really wrong in the sense that yes higher performance does cost more. But it ignores completely the fact that the same intelligence is dropping by an order of magnitude (at least) every 12 months. GPT o1 la
532.
by nl 6mo ago
Not really. For coding I care mostly about reasoning ability which is uncorrelated with cut off
533.
by nl 6mo ago
I've invested significant time into getting open models to work, and investigating what works well. The TL;DR is that unless you are doing it as a hobby or working in an environment where none of the data privacy options supported by A
534.
by nl 6mo ago
Yeah, multi-step SQL generation and debugging. Some models surprised me and Grok Fast was one of them. It is consistently good at this task though!
535.
by nl 6mo ago
First model to get 100% on my agentic benchmark: https://sql-benchmark.nicklothian.com/?highlight=anthropic_c...
536.
by nl 6mo ago
> By all accounts, the best LLM cyber scanning approaches are really primitive - it's just a bash script that goes through every single file in the codebase What accounts are these? I've seen some people use this but I cannot i
537.
by nl 6mo ago
Good point! But they argue that: > PT_DENY_ATTACH (ptrace constant 31): Invoked at process startup before any sensitive data is loaded. Instructs the macOS kernel to permanently deny all ptracerequests against this process, including fro
538.
by nl 6mo ago
They use the Apple TEE which they claim also protects GPU memory (I wasn't aware of this). NVidia data center GPUs have a similar path, but not their consumer ones. Not sure about the NVidia Spark. It's possible AMD Strix Halo can
539.
by nl 6mo ago
They use the TEE to check that the model and code is untampered with. That's a good, valid approach and should work (I've done similar things on AWS with their TEE) The key question here is how they avoid the outside computer bein
540.
by nl 6mo ago
> Companies usually want to IPO with a decent profit margin showing on the books, revenue doesn't usually come into it. Untrue. Read what the OP said about Amazon. And Figma - the most recent high-profile IPO I could think of - isn&

More ›