ComputerGuru

I just want to say 500ms isn’t fast.

Yes, I liked using them for NAS builds. Great for ZFS machines.

Intel is weird with their stuff like that, I wrote about how they introduced and then dropped an SHA instruction that could be used to optimize hashing (useful for things like block checksum hashes on storage servers) before: https://neosmart.net/blog/will-amds-ryzen-finally-bring-sha-...

Yes, here I'm referring to actual ECC, not the on-die ECC. All DDR5 chips have on-die ECC as a prerequisite just to reach error levels comparable to standard non-ECC DDR4, but from a hardware analysis perspective, even with the on-die ECC, DDR5 still needs ECC more than DDR4 (which already needed it).

Of course they disallow it. In saying they’re not thick enough to not realize how it looks to publicize their anti-distillation efforts. You must not have seen the tirades and screeds Anthropic posts in comparison, then.

It’s intentional product segmentation. Just like Intel not shipping ECC for client SKUs when DDR5 design makes it really necessary.

Read this last night. I was completely thrown off when I read this line:

> We expected the engine to return an error for any prompt it could not process, and our long-context tests returned successful responses, so we did not believe this to be an issue

Surely “we expected and error but got a successful response” merits investigating rather than moving on?

My interest was piqued and I tried to read on… Holy mother of all that is slop, I couldn’t even get to the text because of all the interactive slop I had to scroll through.

This is incorrect. You cannot identify the presence of text watermarks if you do not have access to the models so that you can spot where expected tokens diverge from actual tokens based off the weights. That does not preclude Pangram from being able to reliably identify LLM-generated text from models whose outputs they’ve trained their own detection model upon.

I feel like OpenAI hasn’t been as (publicly) hypocritical and petty about their distillation spats as Anthropic has been?

People in HN live in a bubble. Plenty of real world use cannot use subscription plans (not even talking about enterprise agentic use) and is absolutely at the mercy of the cost per task metric.

We agonize over cost per task to the extreme for services/functionality built atop of APIs and can absolutely guarantee you that it is the real world metric that counts most right after “is the model good enough for the task at hand,” and furthermore that, yes, in fact, it is quite often that we see certain models/providers easily using 3x or more tokens on average for the same tasks (or even the exact same task) than a supposedly similarly matched model from another provider. Once you become intimately familiar with the thinking traces from specific models, you can preempt this to some extent with manipulations to the system prompt (essentially heading off entire thinking branches) but the nuking of reasoning traces in the name of preventing against distillation attacks has made this harder.

Pretty sure my Samsung plasma TV from 2011 is safe. It was all downhill from there, just competing over which display looks better with its saturation and brightness boosted all the way up with the floor models all next to each other.

Shameless plug: stop storing secrets in environment variables. Better solutions exist.

Our own open-source, cross-platform, multi-language, cross-framework SecureStore is an embedded secrets store (sort of like SQLite) that lets you version your secrets alongside your code in git (and deploy updates via CI, accordingly), only securely.

Available for rust, golang, php, python, js/ts (node, bun, or in-browser), C#/.NET, PowerShell, bash, and more. Open Source (MIT), not vibe coded, and lovingly maintained for a decade.

https://github.com/neosmart/securestore-rs

Sorry for the twitter, link; I couldn't find this reported elsewhere.

Apparently only supported under darwin and linux for arm64 and x64? At least via npm.

So you’re saying.. they are that bad, actually?

Wow, Voxtral 3 sounds so much more artificial and “AI” as compared to the old Voxtral 1 in the comparison clips on that page! I absolutely hated listening to the new one.

Also, comparisons with Whisper always really tick me off. It’s not a dictation model so much as it is a transcription model. It’s really the only one you can use to capture incidental audio rather than one or more persons deliberately speaking into a microphone, e.g. the only one you can use to transcribe audio from surveillance footage or background audio recorders; it has a higher WER but it actually tries to capture everything (leading quite often to hallucination, unfortunately) whereas the other models they compare it against tune out background noise and dialog altogether, in my fairly extensive testing.

> When I last looked for this a few years ago, I only found a really primitive one that did a linear transformation of all timestamps in the video, and none to correct each line individually.

Not an answer to your current query, but my first published rust crate shortly after the language hit 1.0 was a tool to both calculate and correct linear drift, long before the era of LLMs or quality audio recognition. It might be fun to tackle this again with the new tech we now have at hand!

Srt-resync: https://github.com/mqudsi/srtresync

TFA has a picture as well; just scroll down.

Sorry about your father. He needs a dictation model, not a general purpose speech-to-text model. They ignore umms and ahhs, change things like “an elephant, no a monkey, went up the tree” to “a monkey went up the tree,” support saying punctuation aloud sometimes, etc.

Gemini team just released Gemini 3.5 Transcribe that’s supposed to be good at this; it’s available via api: https://blog.google/innovation-and-ai/models-and-research/ge...

I agree about the chargers!

Im sure you know this, but do note that fully dumping the charge can a) damage the battery and dramatically shorten its lifespan, b) give you different results depending on your discharge rate and the batteries you are testing, as they all have different current-dependent discharge curves.

No, that’s just white labeling. Like you said though, they can switch manufacturers and pull the rug from underneath you at any time.

My example is like if you opened the Amazon Basics box and found Eneloop branded white/black/blue batteries directly.

Amazon Basics is incredible in this regard, they’ve optimized SKU identification down to a pipeline. They’ll essentially randomly pick items off their internal list of highest netting sales and test them to see how dependent they are on brand name recognition and price-quality signalling. To do this as efficiently as possible, they simply purchase a few hundred units of a high quality product in the space, stick it in an Amazon Basics box and list it on their site under their Amazon Basixs brand at a price they feel they can achieve via white labeling, and wait to see how it sells. The use of high quality items (with quality above what can actually be had at the listed price point for the duration of the experiment) means they are really only testing the user base’s willingness to forgo a brand name for the category in exchange for a discount. If it sells well, they then work on sourcing it in bulk as a white labeled item “for real”, while if it sells poorly they simply delist and move on.

I (used to) buy pre-spliced/terminated fiber optic cables with some frequency from Amazon and came to be familiar with the brands and their quality. One time while shopping for some fiber optics, I saw Amazon Basics-labeled OM-3/OM-4 MMF cable at a very tempting price, so I purchased some to see if it was any good.

To my utter shock and surprise, when I received the trademark plain cardboard boxes with the Amazon Basics label on them and proceeded to open them, I found that I was sent boxes of cables still factory wrapped with labels that clearly read “Corning Optical” – which if you know anything about optical fiber, was pretty much the premium brand in the game. I should have stocked up because the next time I went to order I found out their experiment had ended and they no longer sold “Amazon Basics” finer cables.

Yeah, I’m similarly not sure how to square away claims of “trained 100% exclusively on synthetic data” with the fairly obvious fact that it’s derived from an open model, almost all of which are trained on a mix of scraped and data and distilled traces.

Your last bit about it being different is a known issue with models that have been trained on purely synthetic data, no?

Wow, gpt-5.6-Luna was already a fairly unbeatable bargain and now gpt-6-luna is both cheaper and better. And they did a phenomenal job getting gpt-6-sol to max out right where Astra begins; funny how they just so happened to avoid cannibalizing their best model while still being quite cost-competitive near the frontier.

At least it sounds good on paper, the the graphed results do give me pause as it seems the lower cost might come from a slightly nerfed base model combined with more thinking, going by the more erratic scoring curves and the lower no thinking baseline score. I’ll have to try it out but I really hope they haven’t nerfed Luna/Sol to make this price point possible!

Funny. Look up errata AMD-SB-7055: RDSEED Failure on AMD “Zen 5” Processors.

Zen 5 rdrand16/32 return zero with CF=1 on entropy exhaustion and their recommended approach directly leads to the issue you observed: treat all-zero result of rdseed as if cf=0 (failure) and re-roll the dice, effectively recreating the zen 1/zen 2 issue all over again!

They say this might be addressed by a future microcode update… meaning there’s a chance they’ll just patch it to do just that in software. Maybe that’s how they got into this mess in the first place?

Also, am I a complete idiot or is asserting the relative distribution of a mere 64k possible results a rather easy black box validation test that I would’ve assumed they’d be doing? When I used to write cycle-accurate emulators in the past, that would have been an obvious test to include. This isn’t some arcane instruction no one uses or a really complicated case with deep dependency and/or timing issues; it’s like getting rdtsc wrong.

I have been told that ZDR is available upon manual request.

That’s if you have disparate prompts and don’t want to actively select a model. If you’re developing a pipeline, it’s a terrible idea. You want to choose a model, validate it, then stick to it.

Your sensors need calibration, then! This one is as obvious as they come.