Hacker News: macwhisperer

Show HN: PolyRoll – A 3D Marble Game for Mobile/Desktop

macwhisperer — Tue, 06 Oct 2026 10:41:02 +0000

Inspired by games like "super monkey ball" and good 'ol "marble blast".. PolyRoll is basically their evil clone..

Uses a combo of "three.js" and "cannon-es"

--------- Backstory: ---------

3 months ago I stopped at my buddy's house.. he showed me his Claude-code setup.. I had never tried agentic coding so I figured, why not try to make a game?

After some brainstorming and developing a plan with gemini-pro, we set sonnet off on his merry way and went for a swim... About an hour or two later we had a working game (though it was't great, buggy)... After some more prompts to kill the bugs, tweak the play-feel and camera physics- we did some mobile optimizations, added the animations, and designed a d-pad by sending sonnet a photo mockup of the custom buttons we wanted...

For a game made in the span of about 6-8 hours total, it's not bad! Quite difficult/ frustrating at times though.. Some levels get glitchy, and we left it kinda messy/low-poly on purpose!

I don't think my friend has much interest in developing the game further, and I don't have Claude-code, so I figured why not post the game here now so others may enjoy...

---- NOTE: ----

The levels are entirely procedurally generated, no two levels are ever the same...

If you get stuck on a level just reload, you may get easier levels at the expense of starting the game over at level 1 (sorry no saves).. technically infinite, but the levels get very difficult!

Good luck have fun!

Comments URL: https://news.ycombinator.com/item?id=49976619

Points: 2

# Comments: 0

New comment by macwhisperer in "ChatGPT Is Throwing 404"

macwhisperer — Thu, 03 Sep 2026 15:46:14 +0000

lmao *qwen3.8-27b runs in the background*

New comment by macwhisperer in "Claude outage – Resolved"

macwhisperer — Thu, 03 Sep 2026 15:40:52 +0000

honestly boggles my mind that actual companies put load-bearing (lol) work behind 3rd party api's like claude or gpt.. also its really funny because some of these companies fired actual humans to put said 3rd party api to work in place of humans... like what happens when ur boy claude stops showing up for work huh!?

New comment by macwhisperer in "Show HN: A techno machine in one HTML file, with verifiable renders"

macwhisperer — Mon, 24 Aug 2026 20:46:45 +0000

super dope!! the future of musical instruments may be web-based.. exciting times

New comment by macwhisperer in "Qwen3.8 27B scores 52 on Artificial Analysis"

macwhisperer — Mon, 17 Aug 2026 21:21:20 +0000

4-bit quant sits at 15.72GB

New comment by macwhisperer in "Qwen 3.8 27B"

macwhisperer — Sat, 15 Aug 2026 07:40:09 +0000

yeah this model is chefs kiss..

running an untouched, vanilla 4-bit version (Q4_0) I baked myself today (benched it against Q4_K_M (16gb) and IQ3_M (12gb), Q4_0 (15gb) is king)...

this model--

1: over 60% faster than qwen 3.6 version of the same dense 27b model, same engine setup (don't ask how, im not sure either)

2: has better reasoning quality, less "loopy" with its thinking patterns.. most certainly the smartest model on my roster currently

3: has the longest task horizon ive ever experienced (locally or otherwise)...I sent it a bunch of compressed ideas for an app, it sent me back the largest python app ive ever seen in one single ai response pass (80kb text file)

thanks qwen!! hoping to see the full model range get released...

New comment by macwhisperer in "llama.cpp"

macwhisperer — Wed, 12 Aug 2026 18:24:43 +0000

pro tip: if u are building ur own harness with python (recommended) , use "llama-cpp-python"...

I started with "llama-server" and custom stuff around it, which is great for single model setups.. but for multi-model harness with quick switching, llama-cpp-python is peak

New comment by macwhisperer in "Nvidia Nemotron 3.5 Lightning and NeMo Switchyard"

macwhisperer — Wed, 12 Aug 2026 04:44:37 +0000

big week for open models... seems like companies are noticing the 26-35b sweet spot... though I think a 12b-a1b-MoE model would be helpful for the 16gb folks

New comment by macwhisperer in "Mark Zuckerberg attacks 'closed' AI rivals as Meta returns to open models"

macwhisperer — Tue, 11 Aug 2026 10:27:56 +0000

I feel like all the ppl complaining itt don't even run open weight models.. wahh billionaire bad is true, but you are missing the forrest for the trees.. they are literally bending the knee and ur mad about it? it literally means local models won.

New comment by macwhisperer in "I Benchmarked Local LLMs on the Laptop I Have"

macwhisperer — Tue, 11 Aug 2026 08:48:25 +0000

yeah you realize you can turn reasoning off on the newer the models right? also try my version of Gemma-12b https://huggingface.co/macwhisperer/Gemma4-12B-SuperDense or try my qwen 3.5-9b

Show HN: Is the Moon Full?

macwhisperer — Tue, 04 Aug 2026 20:52:49 +0000

its a simple moon tracker lol...

Comments URL: https://news.ycombinator.com/item?id=49174981

Points: 2

# Comments: 0

New comment by macwhisperer in "GPT-5.6 used a prompt to close a 30-year gap in convex optimization"

macwhisperer — Sat, 18 Jul 2026 23:20:13 +0000

Basically, he proved that *information is power.* If you don't know which way to go (the subgradient), you're gonna be calculating forever!

New comment by macwhisperer in "Qwen 3.6 27B is the sweet spot for local development"

macwhisperer — Mon, 29 Jun 2026 23:21:20 +0000

also for those with only 16gb-- try this model https://huggingface.co/macwhisperer/Gemma4-12B-SuperDense its exceptional!

New comment by macwhisperer in "Qwen 3.6 27B is the sweet spot for local development"

macwhisperer — Mon, 29 Jun 2026 23:18:33 +0000

hi guys... I run specialized quants on my 24gb air.. (I specialize in 3-bit quants that punch above their weight).. try out my version of 3.6-27b I think you be impressed https://huggingface.co/macwhisperer/Qwen3.6-27B-SuperDense

New comment by macwhisperer in "Ask HN: Best local LLM under 2B paramater and consuming RAM less than 3gb"

macwhisperer — Sun, 28 Jun 2026 18:27:03 +0000

qwen3 1.7b- q4_k_m is your best best for that size

New comment by macwhisperer in "The unbearable cheapness of open weight models"

macwhisperer — Fri, 26 Jun 2026 20:17:03 +0000

literally ask cloud ai like the free gemini or chatgpt.. they could make you an expert on the subject overnight..

New comment by macwhisperer in "Ask HN: Has anyone replaced Claude/GPT with a local model for daily coding?"

macwhisperer — Tue, 16 Jun 2026 17:41:19 +0000

I code with like a slew of 20+ custom baked models of all sizes, in various fully custom multi-model harnesses that use different bindings...

the harnesses themselves are just as important as the models...different harnesses give different responses with the same prompt, same model...

if you have the 20/mnth claude sub or codex, you really should be using that to build a good local harness for yourself... claude won't be 20$ forever

build the stack first! when you get that new comp with massive ram, youre already set, just run a larger model!

big cloud models are incredibly good at building and teaching about local ai!

have fun in the rabbit hole!

if you are memory constrained like me, check out my custom models https://huggingface.co/macwhisperer

New comment by macwhisperer in "SpaceX to buy Cursor for $60B"

macwhisperer — Tue, 16 Jun 2026 16:57:56 +0000

ai is like the first technology with a conversational service manual inside it..

you should be foaming at the mouth to use claude or codex to make a custom harness, just for your own personal use with local models...

New comment by macwhisperer in "Making a vintage LLM from scratch"

macwhisperer — Fri, 12 Jun 2026 15:03:43 +0000

super inspiring! thanks for sharing!

New comment by macwhisperer in "Ask HN: What are tools you have made for yourself since the advent of AI?"

macwhisperer — Mon, 08 Jun 2026 23:14:55 +0000

retro-inspired fully custom, swiss army knife style notepad --

https://convert.neocities.org