HNHacker News
TopNewBestAskShowJobs

hnuser123456

2,025 karma · joined January 30, 2019

a
submissionscomments
hnuser123456··on Gemma 3 270M: Compact model for hyper-efficient AI
For keys, you probably want to use at least q5 or q6, for values q4 is fine
hnuser123456··on April Fools 2014: The *Real* Test Driven Development (2014)
We're beyond that, now we can vibecode both the tests and the implementation.
hnuser123456··on A Comprehensive Survey of Self-Evolving AI Agents [pdf]
Thanks for the writeup. I wonder if it would be plausible to run this kind of self-optimization for a wider variety of problem sets, to generate "context pathways" for various tasks that are all optimized, and maybe even learn patterns from multiple prompt optimizations to generalize.
hnuser123456··on High-severity WinRAR 0-day exploited for weeks by 2 groups
What I don't get is why people kept installing WinRAR when 7zip can do all the same things and doesn't beg for money.
hnuser123456··on GPT-OSS-120B runs on just 8GB VRAM & 64GB+ system RAM
Research also often finds that fine-tuning in more guardrails also decreases performance, which is done to all non-base instruction tuned models, which most people are using.
hnuser123456··on GPT-OSS-120B runs on just 8GB VRAM & 64GB+ system RAM
You need to find an abliterated finetune, where someone sends prompts that would hit the guardrails, traces the activated neurons, finds the pathway that leads to refusal, and deletes it.
hnuser123456··on GPU-rich labs have won: What's left for the rest of us is distillation
>memory-centric compute

This already exists: https://www.cerebras.ai/chip

They claim 44 GB of SRAM at 21 PB/s.

hnuser123456··on Tor: How a military project became a lifeline for privacy
Back when I tried, it was a modified Firefox build.
hnuser123456··on Running GPT-OSS-120B at 500 tokens per second on Nvidia GPUs
Yeah, Crysis came out in 2007 and could run physics on the GPU.
hnuser123456··on GPT-5: Key characteristics, pricing and system card
I agree, we have now proven that GPUs can ingest information and be trained to generate content for various tasks. But to put it to work, make it useful, requires far more thought about a specific problem and how to apply the tech. If you could just ask GPT to create a startup that'll be guaranteed to be worth $1B on a $1k investment within one year, someone else would've already done it. Elbow grease still required for the foreseeable future.

In the meantime, figuring out how to train them to make less of their most common mistakes is a worthwhile effort.

hnuser123456··on GPT-5
I see now on the website, the screenshot cut off the header for the first benchmark, looked like it was just comparing 1-pass and 2-pass.
hnuser123456··on GPT-5
Haha, even with that, it says 4o does worse with 2 passes than with 1.

Edit: Nevermind, just now the first one is SWE-bench and 2nd is aider.

hnuser123456··on Brennan Center for Justice Report: The Campaign to Undermine the Next Election
I renewed my license 4 years ago and got an enhanced license with no special request.
hnuser123456··on Litestar is worth a look
It's a python web framework, for those curious to know more before clicking through.
hnuser123456··on Qwen3-4B-Thinking-2507
A 24GB GPU can run a ~30b parameter model at 4bit quantization at about 8k-12k context length before every GB of VRAM is occupied.
hnuser123456··on "This question has been retired"
I'm sure it's still on their drives somewhere, they'd just rather you ask copilot
hnuser123456··on GPT 5 coming tomorrow confirmed
gpt-oss-5b
hnuser123456··on AI is propping up the US economy
Nobody gets fired for choosing Microsoft
hnuser123456··on Open models by OpenAI
Native FP4 quantization means it requires half as many bytes as parameters, and will have next to zero quality loss (on the order of 0.1%) compared to using twice the VRAM and exponentially more expensive hardware. FP3 and below gets messier.
hnuser123456··on Scientific fraud has become an 'industry,' analysis finds
Those whose parents stressed nothing but academics hit a dead end if the parents can't keep paying the kid to get high grades.
hnuser123456··on Open models by OpenAI
Maybe someone got tired of waiting paid them to release something actually open
hnuser123456··on Open models by OpenAI
Text only, when local multimodal became table stakes last year.
hnuser123456··on Genie 3: A new frontier for world models
It's feasible you could have a personal neural net that fine-tunes itself overnight to make less inference mistakes in the future.
hnuser123456··on GitHub pull requests were down
Also the foo_bar method from v1 worked better so pull that back in
hnuser123456··on Things that helped me get out of the AI 10x engineer imposter syndrome
I think there are three factors to this: 1. What to code (longer, more specific prompts are better but take longer to write), and 2. How to code it (specify languages, libraries, APIs, etc.) And if you're trying to write code that uses a newer version of a library that works differently from what's most commonly documented, it's a long uphill battle of constantly reminding the LLM of the new changes.

If you're not specific enough, it will definitely spit out a half-baked pseudocode file where it expects you to fill in the rest. If you don't specify certain libraries, it'll use whatever is featured in the most blogspam. And if you're in an ecosystem that isn't publicly well-documented, it's near useless.

hnuser123456··on NASA's Curiosity picks up new skills
Yeah, sloppy writing. They're maximizing how quickly they can complete tasks by multi-tasking and enter sleep mode sooner, reducing recharge time and reducing the amount of energy wasted on systems that are in active standby. They rediscovered race to idle.
hnuser123456··on A GitHub repo with no code, just a memory and a license written for her
The trick is to shut your brain off and say what you wanted to say without thinking about it while you do it. That's how I did it the first time anyways (she said no but wasn't mad)
hnuser123456··on The anti-abundance critique on housing is wrong
And you obviously already got your big expensive house and just want number to go up...
hnuser123456··on Following Up on the Python JIT
Yep, I had a script that was doing some dict mapping and re-indexing, wrote the high level code to be as optimal as possible, and switching from cpython to pypy brought the run time from 5 minutes to 15 seconds.
hnuser123456··on GEPA: Reflective prompt evolution can outperform reinforcement learning
It has been commonly observed that the current crop of LLMs can be too agreeable/sycophantic (or on some topics, too disagreeable) due to the commonly chosen RLHF priorities.

Simply asking the LLM in two separate contexts the same question but from opposing perspectives, then in a third context asking it to analyze both responses and choose the most neutral and objective take, you wipe out any "(dis)agreeableness" bias and dig closer to a deeper, more nuanced synthesis of a given topic. This paper is just taking this idea to the next level.

This isn't really possible with RLHF alone unless you train the LLM to often give two opposing perspectives, which would get tiring.

← PreviousPage 9 of 34Next →