Hacker News Reader

drittichon MiMo-v2.6-Pro: Intelligence, Performance and Price Analysis

Interesting - have you done MiMo-v2.6-flash?

drittichon GPT-6 Astra

Often the smartest thing is to do nothing.

drittichon Degraded performance for multiple models

API Error: 529 Overlorded

drittichon Apple Silicon and macOS VMs: Faster LLM Inference with llama.cpp

Things like the number of GPU layers, whether to use flash attention or not, enabling MTP, playing with context size.

drittichon Apple Silicon and macOS VMs: 11–16× Faster LLM Inference with Llama.cpp

There are certainly challenges. When setting up a new model, I get AI to walk me through the commands using llama-benchmark that determine the best parameters for my particular configuration and needs. Once you've got that it's pretty easy to port those parameters to llama-server. It takes me about an hour to run through this process. It would be great if there was a registry of hardware, models, configuration parameters, and resulting tokens per second. Maybe one day we'll get there.

drittichon I’ve built a virtual museum with nearly every operating system you can think of

And I thought I was killing it just saving some install disk images!

drittichon Granite 4.1: IBM's 8B Model Matching 32B MoE

Nanobanana for scale.

drittichon The Classic American Diner

Or do they? https://www.youtube.com/watch?v=EMNqJQaf08E

drittichon The Beauty of Bonsai Styles

...is the camera you have with you.

drittichon Simple self-distillation improves code generation

I used to invent TLAs on the spot for fun, and when someone asked what it was, would respond, "It's a PUA", eventually revealing that meant "previously unknown acronym". It was even more annoying that it sounds.

drittichon Scientific audio equipment analysis with analyzer shows no difference in quality

I think the point is to reproduce the sound of those hundreds of feet of standard run-of-the-mill cabling as faithfully as possible ;)

drittichon Don't post generated/AI-edited comments. HN is for conversation between humans

Voice is everything. Don't relinquish the best part of yourself.

drittichon GPT-5.4

I think it's time for an https://hotornot.com for AI models.

drittichon Setting up phones is a nightmare

Tell me why they only support Google Pixel phones, v6 through 10.

drittichon A macOS app that blurs your screen when you slouch

Yes, shower thinking with warm water on my neck is absolute peak. In those conditions I'm unafraid of tackling the most challenging of thinking.

drittichon A decentralized peer-to-peer messaging application that operates over Bluetooth

And of course you can now run local LLMs on your phone as well.

drittichon AI generated music barred from Bandcamp

Also a musician and I don't think it's that amusing. IMO this isn't an "AI can't be art" discussion. It's about the fact that AI can be used to extract value from other artists' work without consent, and then out-compete them on volume by flooding the marketplace.

drittichon NCSA Mosaic 2.7, one of the first graphical web browsers

Your description matches my recollection exactly.

drittichon So you wanna build a local RAG?

Do you have a standard prompt you use for this? I have definitely seen agentic tools doing this for me, e.g., when searching the local file system, but I'm not sure if it native behaviour for tool-using LLMs or if it is coerced via prompts.

drittichon Agent design is still hard

This sounds interesting. What about the agent behavior itself? How it decides how to come at a problem, what to show the user along the way, and how it decides when to stop? Are these things you have attempted to grapple with in your framework?

drittichon EXIF orientation info in PNGs isn't used for image-orientation: from-image

Yes, came to the same conclusion - it's a pain, but solves the problem permanently.

drittichon Why the push for Agentic when models can barely follow a simple instruction?

We all just feed the LLMs now.

drittichon Why the push for Agentic when models can barely follow a simple instruction?

Looks like it comes from Scientific American: https://spaf.cerias.purdue.edu/~spaf/Yucks/V5/msg00004.html

drittichon Microsoft BASIC for 6502 Microprocessor – Version 1.1

And recording software applications to cassette off the radio!

drittichon Microsoft BASIC for 6502 Microprocessor – Version 1.1

I cut my teeth on that language, and still keep a Commodore PET around for old times sake.

drittichon Adaptive LLM routing under budget constraints

"When you use Unpaid Services, including, for example, Google AI Studio and the unpaid quota on Gemini API, Google uses the content you submit to the Services and any generated responses to provide, improve, and develop Google products and services and machine learning technologies, including Google's enterprise features, products, and services, consistent with our Privacy Policy.

To help with quality and improve our products, human reviewers may read, annotate, and process your API input and output. Google takes steps to protect your privacy as part of this process. This includes disconnecting this data from your Google Account, API key, and Cloud project before reviewers see or annotate it. Do not submit sensitive, confidential, or personal information to the Unpaid Services."

Reference: https://ai.google.dev/gemini-api/terms

drittichon AOL to discontinue dial-up internet

For younger users of the internet, it's hard to overstate how omnipresent the AOL brand was. The marketing team was on overdrive, all the time. And their marketing CDs probably caused a noticeable increase in CD-ROM adoption.

drittichon The Titanic’s Best Lifeboat

tl;dr

HN values efficiency

drittichon The unreasonable effectiveness of an LLM agent loop with tool use

Perhaps that's a false dichotomy?

drittichon LLMs Get Lost in Multi-Turn Conversation

Also exists in LM Studio.