HNHacker News
TopNewBestAskShowJobs

qeternity

8,841 karma · joined May 4, 2016

submissionscomments
qeternity··on Qwen 3.8 27B is excellent, but it defaults to overthinking things
The real breakthrough is going to be thinking in latent space.
qeternity··on Auto-research with codex: How I achieved a 232x Faster Kernel
You say this like that isn’t how the vast majority of problems are solved…
qeternity··on Qwen 3.8 27B
Has nothing to do with Chinese.

Frontier labs have already been doing this for a while, verified in smuggled traces from OAT/Ant.

Simply a way to reduce tokens.

qeternity··on Qwen 3.8 27B
35A3 might be more comparable to 10 dense.

27 dense is far more capable than 35A3.

qeternity··on Gemini 3.7 Flash
> more than half its price

Less than half its price.

More than 50% discount.

qeternity··on Qwen3.8-Max: A New Bar for Coding and Cowork
LLMs are, in theory, deterministic. Sampling is not intrinsic to LLMs.

Greedy decoding a single batch in most libraries will give you mostly deterministic outputs. Higher batch sizes can increase variance.

But all of this is down to CUDA and/or kernel implementation issues.

qeternity··on Moonshot’s Kimi uses 20k Nvidia chip cluster from Alibaba
I don't follow this. Clearly all of the frontier labs are doing these things.

When OAI released gpt-oss it was released as an mxfp4 checkpoint.

OAI, Ant, et al are also obviously employing QAT.

qeternity··on Kimi-K3 Releases on HuggingFace 7/27
I cannot be sure what the likes of Cursor have done, but I think it's incredibly unlikely that they have trained a QLoRA for Composer.

It's almost certainly full parameter post training of the original model weights.

qeternity··on “We have information that Moonshot distilled Fable for the development of K3”
I think it’s fairly obvious the Chinese labs are doing mass distillation.

I also think the Fable accusation is wrong and it was most likely Opus 4.8 which itself is likely a distillation of Fable.

qeternity··on AWS: Inaccurate Estimated Billing Data – $1.7 billion
But these aren't the right services where the test should be, right?

There's another service that says "ok we take the 100 bytes from A, and we take the $17 SKU from B, and this should equal $X".

It's the third service that multiplies these things that failed. Where are the tests for that?

qeternity··on Kimi K3: Open Frontier Intelligence
It's a defensive tactic to reduce the effectiveness of distillation.

Say of that what you will, but it's not because they want to wrest control from users.

It's because they don't want Chinese companies to do exactly what Moonshot (Kimi creators) and others have done.

qeternity··on SpaceX bond worth 10% less than issue price – heading for junk bond status
Public in this context just means publicly listed on a stock exchange.

It does not mean it is state-owned.

qeternity··on SpaceX bond worth 10% less than issue price – heading for junk bond status
Empirically this isn't true. However you feel about "true" capitalism vs socialism, countries underpinned by capitalism have prospered, even the socialist flavors (China, Scandinavia)...while the countries that have attempted pure socialism have all failed.
qeternity··on S&P downgrades Oracle to BBB – only one notch above junk level
> is supposed to be the central bank's job.

What? No it's not, and never has been.

Without even getting into the practical vs. theoretical of Fed dual mandate (funding deficits), even the most uncharitable take on modern CBs wouldn't suggest this.

qeternity··on SpaceX wants to launch 100k more Starlink satellites for 100x the bandwidth
We are quite literally talking about launching 100k satellites into space.

This is necessarily a question that requires thought and speculation about the future.

You can't just flip a switch and do it overnight if it turns out there is demand.

And if there isn't demand, you are going to lose a lot of money.

qeternity··on SpaceX wants to launch 100k more Starlink satellites for 100x the bandwidth
It's sort of amazing that people still say these sorts of things today.

I don't think it's at all inconceivable where people in the future are streaming high resolution multi modal personal sensor arrays to AI that is running in a data center...and ditto we are streaming ever more content back.

That's just the in vogue answer of 2026. There are undoubtedly endless innovations that I can't envision that will require ever more resources.

My point is just that every single "X ought to be good enough" has been proven false. And the only times where we see consumption really plateau are due to other reasons than desire (like cost).

qeternity··on Hetzner Price Adjustment
There is a huge amount of vendor financing available for this exact reason.

Hetzner presumably have been making use of this for years already.

It's far more likely they simply began raising prices and realized the market would tolerate it.

qeternity··on KVarN: Native vLLM backend for KV-cache quantization by Huawei
I think it very much is worth it!

But the point was that quality didn't magically increase.

qeternity··on KVarN: Native vLLM backend for KV-cache quantization by Huawei
It's not better quality: 59.3% vs 59.4% fp16 on AIME 25
qeternity··on Anthropic raises $65B in Series H funding at $965B post-money valuation
> Venture capitalists & private investors are sucking all of the possible growth and future upside from these companies and then dumping them on retail investors when there's nothing left.

A lot of the money that is deployed by VCs comes from pension funds and asset managers that ultimately manage money for the average Joe.

qeternity··on Introspective Diffusion Language Models
I haven't read TFA yet but a common technique is speculative decoding where a fast draft model will generate X tokens, which are then verified by the larger target model. The target model may accept some Y <= X tokens but the speedup comes from the fact that this can be done in parallel as a prefill operation due to the nature of transformers.

So let's say a draft model generates 5 tokens, all 5 of these can be verified in parallel with a single forward pass of the target model. The target model may only accept the first 4 tokens (or whatever) but as long as the 5 forward passes of the draft model + 1 prefill of the target model is faster than 4 forward passes of the target, you will have a speedup while maintaining the exact output distribution as the target.

qeternity··on Anthropic downgraded cache TTL on March 6th
> They paid a billion dollars for a vibe coded mess just for the opportunity to associate themselves with the hype.

Lol no they didn't. It wasn't even an acquihire. They just hired Peter.

Maybe they are paying him incredibly well, but not a billion dollars well.

qeternity··on Meta removes ads for social media addiction litigation
> It's not any company, its Meta and the channels they administrate come with a set of responsibilities and principles

Sorry, which laws stipulate these special responsibilities and principles?

qeternity··on Claude mixes up who said what
> or if the model might actually have emitted the formatting tokens that indicate a user message.

These tokens are almost universally used as stop tokens which causes generation to stop and return control to the user.

If you didn't do this, the model would happily continue generating user + assistant pairs w/o any human input.

qeternity··on Claude mixes up who said what
This does not solve the problem at all, it's just another bandaid that hopefully reduces the likelihood.
qeternity··on Mamba-3
Yes, it is written for a specific audience.

That is not a reason for snark.

As other commenters have noted, it’s well written.

qeternity··on Better JIT for Postgres
Yes, lots of things can create indeterminism. But nothing is inherent.
qeternity··on Better JIT for Postgres
Yes, lots of things can create indeterminism. But nothing is inherent.
qeternity··on Better JIT for Postgres
> LLMs are inherently non-deterministic.

This isn't true, and certainly not inherently so.

Changes to input leading to changes in output does not violate determinism.

qeternity··on MCP server that reduces Claude Code context consumption by 98%
> With prompt caching, verbose context that gets reused is basically free.

But it's not. It might be discounted cost-wise, however it will still degrade attention and make generation slower/more computationally expensive even if you have a long prefix you can reuse during prefill.

← PreviousPage 3 of 34Next →