Hacker News: jasonjmcghee

New comment by jasonjmcghee in "Show HN: Maple-Preview – Ternary 20B MoE running at 120 tok/s on a iPhone"

jasonjmcghee — Wed, 05 Aug 2026 14:25:15 +0000

It's far more likely basic consumer devices will advance such that much larger models efficiently run than finding novel ways to compress all of human experience to fit on today's mobile hardware.

Information can only be compressed so much

New comment by jasonjmcghee in "Prevent cognitive debt by manually retyping LLM-generated code"

jasonjmcghee — Mon, 03 Aug 2026 13:41:45 +0000

I don't manually retype the code from pull requests of other engineers.

It's important to retain the what why and how as a team to the degree that development can be efficient, extensible, and maintainable. (And ideally a good devx)

Needing to manually touch every line of code is not scalable.

This was true 10 years ago. It's still true today.

New comment by jasonjmcghee in "The coolest use for the Vision Pro"

jasonjmcghee — Thu, 30 Jul 2026 06:12:07 +0000

I feel like this is supposed to be a "gotcha", but this is VR, not AR.

So you can just, move up.

New comment by jasonjmcghee in "Kimi Linear: An Expressive, Efficient Attention Architecture (2025)"

jasonjmcghee — Tue, 28 Jul 2026 13:03:22 +0000

(2025)

As it's 9 months old and they just had a major model release

New comment by jasonjmcghee in "Did they ghost you?"

jasonjmcghee — Sat, 25 Jul 2026 21:51:12 +0000

Based on the strange emoji use in place of some buttons, I'm guessing this is vibe coded- and a lot of this is really nice looking front end - great color, topography, border styles... Would be interested to know how much intervention there was or if this is where we are

New comment by jasonjmcghee in "Claude Cookbook"

jasonjmcghee — Fri, 24 Jul 2026 14:50:21 +0000

You weren't kidding. That's embarrassing.

"Use a black background and this hideously wide font!"

New comment by jasonjmcghee in "Buz – A fork of Bun using modern Zig, with sub-1s incremental builds"

jasonjmcghee — Fri, 24 Jul 2026 14:43:49 +0000

LLMs from 1+ years ago approached solving every problem by adding additional code and not cleaning up or adding strange unnecessary backwards compatibility (and often still do this kind of thing - "all tests green")

New comment by jasonjmcghee in "The Great Steam Deck Crash of 2026: How the New Pricing Decreased Demand by 80%"

jasonjmcghee — Tue, 21 Jul 2026 13:48:57 +0000

Not trying to be an apologist for anyone, but if the epic store could charge 30%, it's hard for me to believe they wouldn't.

New comment by jasonjmcghee in "A Second-Grade Teacher Revived a Beloved Video Game"

jasonjmcghee — Sat, 18 Jul 2026 23:24:52 +0000

I saw backyard baseball on steam the other day and was blown away by the $40 price tag.

I expected $10 maybe $20. I think $40 is too far outside of impulse / nostalgia purchase.

New comment by jasonjmcghee in "Backtrack-Free Cursive"

jasonjmcghee — Mon, 13 Jul 2026 14:41:37 +0000

Weird amount of negativity here.

Some person recognized a problem they wanted solved and put a bunch of effort into creating a unique solution.

I love this stuff.

This post is perfect for HN.

New comment by jasonjmcghee in "Your code is fast – if you're lucky"

jasonjmcghee — Sat, 11 Jul 2026 13:42:07 +0000

Casey Muratori's https://computerenhance.com is a good resource

New comment by jasonjmcghee in "After 7 years in production, Scarf has reluctantly moved away from Haskell"

jasonjmcghee — Sat, 11 Jul 2026 01:09:16 +0000

Hey Avi - didn't know you were an HNer fun to see you / your co on here :)

New comment by jasonjmcghee in "GPT-5.6 Sol Ultra produces proof of the Cycle Double Cover Conjecture [pdf]"

jasonjmcghee — Fri, 10 Jul 2026 19:16:46 +0000

No there was the planar unit distance problem

(Erdős problem 90)

New comment by jasonjmcghee in "Fable created novel 4D splat format"

jasonjmcghee — Sat, 04 Jul 2026 23:29:52 +0000

3D Gaussian splats are still images of 3D space / photogrammetry

So I'm guessing it came from representing the additional dimension

New comment by jasonjmcghee in "Fable created novel 4D splat format"

jasonjmcghee — Sat, 04 Jul 2026 23:21:19 +0000

time

New comment by jasonjmcghee in "Potential session/cache leakage between workspace instances or consumer accounts"

jasonjmcghee — Sat, 04 Jul 2026 17:21:52 +0000

Just use a symbolic link

New comment by jasonjmcghee in "GitFut – Your GitHub stats turned into a World-Cup-style player card"

jasonjmcghee — Sat, 04 Jul 2026 05:39:56 +0000

It says I’ve done 8 PRs- I’ve done a few thousand at least lol - you could get a ballpark by looking at “Pull Shark” count.

Commits also look like it’s this year only?

New comment by jasonjmcghee in "How many failed startups have you launched?"

jasonjmcghee — Fri, 03 Jul 2026 17:08:51 +0000

For future reference, you need to put newlines between lines to preserve formatting.

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

jasonjmcghee — Wed, 01 Jul 2026 23:43:25 +0000

(see above reply to myself) I misattributed the gain from dflash - it was dflash-mlx library + mlx model, not dflash itself giving me the speedup.

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

jasonjmcghee — Wed, 01 Jul 2026 23:38:27 +0000

I usually get mlx versions of models and didn't realize I was using non-mlx - misattributing the performance boost to speculative decoding.

With dflash-mlx library, with dflash disabled 3.6-35b-a3b mlx model I'm getting ~60 t/s on low power and >100 t/s on high power.

Compared with previous message being official qwen huggingface release (non-mlx) using lm studio.