Hacker News: benbojangles

New comment by benbojangles in "H3-metal – Native MiniMax-H3 inference for Apple Silicon"

benbojangles — Fri, 14 Aug 2026 05:03:54 +0000

running this on an m1 max 64gb, it produces some really nice clips with music and audio, stitching the clips together after makes a nice short story generated entirely using a local ai system. It takes quite a while to generate each clip though, But at least it is working. I would like to know a few things:

-Optimal settings/configs examples for h3.c to help speed things up -Optimal recommended generation settings for each mac product, i am sure it is easy to do -Prompt generator assistant

Great work from the github author. The more i use it the more i realise i don't need a gui to generate video, just terminal.

My setup: Macbook #1 as a client Macbook #2 as server

-use macbook #1 terminal + ssh -run mactop in terminal tab to monitor Macbook #2's hardware during ai video generation -run h3.c on Macbook #2 via ssh session -transfer the output video file from macbook #2 to macbook #1 using terminal scp -view the video

New comment by benbojangles in "Spider venom kills varroa mites without harming honeybees"

benbojangles — Thu, 09 Jul 2026 08:17:13 +0000

One day bees will evolve to produce spider venom

New comment by benbojangles in "How to setup a local coding agent on macOS"

benbojangles — Fri, 12 Jun 2026 20:16:44 +0000

70b on my M1 max 64gb

New comment by benbojangles in "How to setup a local coding agent on macOS"

benbojangles — Fri, 12 Jun 2026 20:13:07 +0000

I run a second Mac for local llm use and access it remotely using ssh from the first mac

New comment by benbojangles in "How to setup a local coding agent on macOS"

benbojangles — Fri, 12 Jun 2026 20:12:01 +0000

Omlx needs to incorporate macos native shortcuts use - macos can almost instantly extract text from pdfs and a bunch of other things using it's ane neural engine keeping unified ram for llm use. The two together would be awesome

New comment by benbojangles in "Gemma 4 12B: A unified, encoder-free multimodal model"

benbojangles — Thu, 04 Jun 2026 07:48:59 +0000

why combine audio & image analysis into an llm though, why not allow the user to choose their own audio & image analysis alongside their own llm choice?

New comment by benbojangles in "Gemma 4 12B: A unified, encoder-free multimodal model"

benbojangles — Thu, 04 Jun 2026 07:05:58 +0000

I run gemma-4-26b-bf16 in mtp mode and it runs very smooth, spitting out answers in seconds and outputting text 30x faster than i can read.

New comment by benbojangles in "Indexing a year of video locally on a 2021 MacBook with Gemma4-31B (50GB swap)"

benbojangles — Sat, 23 May 2026 03:08:32 +0000

Gemma4 because presumably it does image analysis right?

-31b It's a dense model

-how many tokens/s is it running at

-What temps are the M1 max GPU/CPU running at

-Is it mlx or gguf

-Why 31b and not 26b which is moe and much more efficient on the m1 max at 50tokens/s & low temps.

I personally use (MLX) qwen3.6-35b-8bit mostly, but use Gemma-4-26b-4bit for image analysis, its mind blowing how fast it is at identifying the scene in a photograph.

New comment by benbojangles in "Travelers on Air Force One ordered to throw away gifts, phones after China trip"

benbojangles — Fri, 15 May 2026 19:17:36 +0000

Didn't China steal the F-22/F35/B2/C17 blueprints, and also stole COVID?

New comment by benbojangles in "Gemini-CLI, zeroclaw, Llama.cpp, Qwen3.5 on TK1"

benbojangles — Thu, 09 Apr 2026 09:21:31 +0000

Restoring, updating, upgrading and compiling on a 2014 Jetson TK1 - Running: ai agent cli, ai agent assistant, local llm models.

Gemini-CLI, zeroclaw, Llama.cpp, Qwen3.5 on TK1

benbojangles — Thu, 09 Apr 2026 09:21:31 +0000

Article URL: https://forums.developer.nvidia.com/t/gemini-cli-zeroclaw-llama-cpp-qwen3-5-on-tk1/365653

Comments URL: https://news.ycombinator.com/item?id=47701187

Points: 1

# Comments: 1

New comment by benbojangles in "Tell HN: Anthropic no longer allowing Claude Code subscriptions to use OpenClaw"

benbojangles — Sat, 04 Apr 2026 03:20:23 +0000

Anthropic are a smart clever research based bunch of people, they probably realised that openclaw is a mess, full of vibe coding get rich quick people, nothing particularly interesting to observe, and don't want to mix this data with the data they have already from real coders.

New comment by benbojangles in "AI overly affirms users asking for personal advice"

benbojangles — Sat, 28 Mar 2026 17:05:34 +0000

Yes I noticed too that several ai agents will tell you directly the code is correct and it is 100 percent fixed but I know it is not true, when I explain to the AI agent that I know they are wrong and serve the solution the ai agent will just act as though what they said never happened and then use my solution to reaffirm they have provided a solution. It's frustrating, laughable, and painful to watch all at once. Makes me realise these companies hired some evil philosophy graduates to build AI soul.md

New comment by benbojangles in "How to run Qwen 3.5 locally"

benbojangles — Sun, 08 Mar 2026 14:05:57 +0000

I'm running Qwen3.5:0.8b locally on an Orangepi Zero 2w using llama.cpp, runs just fine on cpu only. If I want vulkan GPU I have run qwen3.5:2b locally on a meta quest 3 with zeroclaw and saved myself hundreds of $$$ buying a low power computer. I recommend people stop shopping around for inflated mac minis and look at getting a used android phone to load local models on.

New comment by benbojangles in "Apple's 512GB Mac Studio vanishes, a quiet acknowledgment of the RAM shortage"

benbojangles — Sun, 08 Mar 2026 13:59:51 +0000

You can keep your 512gb mac studio i'm running qwen3.5:0.8b on an Orangepi zero 2w and learning just as much as they are.

New comment by benbojangles in "Apple's 512GB Mac Studio vanishes, a quiet acknowledgment of the RAM shortage"

benbojangles — Sun, 08 Mar 2026 13:57:27 +0000

i'm running qwen3.5:0.8b on my orangepi zero 2w, low token/s but it still runs. I think i paid around £14 for it over a year ago but now the same board is double price. I wouldn't buy a computer right now of any kind. It's a bubble.

New comment by benbojangles in "Tech employment now significantly worse than the 2008 or 2020 recessions"

benbojangles — Fri, 06 Mar 2026 22:57:27 +0000

I don't really agree, i am making more money now than ever.

New comment by benbojangles in "Sam Altman (OpenAI) and Dario Amodei (Anthropic) Refuse to Hold Hands"

benbojangles — Thu, 19 Feb 2026 22:10:50 +0000

haha this is hilarious

New comment by benbojangles in "Andrew Mountbatten-Windsor arrested on suspicion of misconduct in public office"

benbojangles — Thu, 19 Feb 2026 10:19:10 +0000

And this is of course tech related because??? If I wanted news I would just check my RSS feed.

New comment by benbojangles in "Qwen3.5: Towards Native Multimodal Agents"

benbojangles — Mon, 16 Feb 2026 18:43:22 +0000

Was using Ollama but qwen3.5 unavailable earlier today