HNHacker News
TopNewBestAskShowJobs

jasonjmcghee

4,910 karma · joined March 25, 2015

Website: https://jason.today

GitHub: https://github.com/jasonjmcghee

submissionscomments
jasonjmcghee··on Nanopass Framework: Clean Compiler Creation Language
Hmm, there was this recent article - not exactly right but overlaps and has interesting ideas

https://sequoiacap.com/article/from-hierarchy-to-intelligenc...

jasonjmcghee··on Qwen3.6-Max-Preview: Smarter, Sharper, Still Evolving
Wow I couldn't disagree more.

For me, Opus 4.5 and 4.6 feel so different compared to sonnet.

Maybe I'm lazy or something but sonnet is much worse in my experience at inferring intent correctly if I've left any ambiguity.

That effect is super compounding.

jasonjmcghee··on Qwen3.6-Max-Preview: Smarter, Sharper, Still Evolving
You may be interested in "radiance cascades"
jasonjmcghee··on Nanopass Framework: Clean Compiler Creation Language
This is a question folks are asking about in terms of organization building too.

Bottlenecks are changing and it's pretty interesting.

jasonjmcghee··on Measuring Claude 4.7's tokenizer costs
It's more like, if it gets it right 99% of the time, that sounds incredible.

Until it's making 100k decisions a day and many are dependent on previous results.

jasonjmcghee··on €54k spike in 13h from unrestricted Firebase browser key accessing Gemini APIs
So someone took a picture of the key at the live training session or something? What's the suspected cause?
jasonjmcghee··on A simpler, faster way to unlock 1Password
Genuinely trying to understand what exactly the offering is.

Optionally turn locking off?

What does it mean to unlock when the device unlocks? I think every single one of my apps might have built this feature too.

jasonjmcghee··on Lumina – a statically typed web-native language for JavaScript and WASM
I see syntax highlighting on the website fwiw
jasonjmcghee··on Anthropic downgraded cache TTL on March 6th
All the weird stuff happening with anthropic / Claude aside- just talking about this post:

Looking at the table with February and April- I don't get it. What am I missing?

The cost and number of calls look pretty aligned on all rows

jasonjmcghee··on LÖVE: 2D Game Framework for Lua
Yes, idk, and of course
jasonjmcghee··on Filing the Corners Off MacBooks
Depending on how I'm using the computer, I may definitely have deep marks after working laying down, but if I sit in a wood chair for a while it's the same thing- and my forearm is much tougher than behind my knee.

I suppose I would prefer it nice and rounded and soft on my wrist - but I don't feel like it's quite as extreme as this thread would have you believe lol

jasonjmcghee··on Claude Code Down
People keep saying this and idk what I'm doing wrong. Using q8_0 on all the latest and greatest local models and they just don't come close to sonnet.

I've tried different harnesses, building my own etc.

They are reasonably close to haiku? Maybe?

jasonjmcghee··on Nanocode: The best Claude Code that $200 can buy in pure JAX on TPUs
fwiw, other models seem to / are reported to struggle much more with using claude code compared with codex / opencode / pi etc.

that being said, there are other potential explanations

jasonjmcghee··on LÖVE: 2D Game Framework for Lua
Definitely SDL2.

Author is currently building version 12 which will be using SDL3. But it's been in development for quite some time with no clear end date afaik.

jasonjmcghee··on HarfBuzz Slug Support with WebGL
Note to author: the WebGL has great zoom settings, but WebGPU has different ones that feel very low-sensitivity (at least on trackpad on macos)
jasonjmcghee··on April 2026 TLDR Setup for Ollama and Gemma 4 26B on a Mac mini
I googled it- supposed fixed template

https://github.com/ggml-org/llama.cpp/issues/21347#issuecomm...

jasonjmcghee··on April 2026 TLDR Setup for Ollama and Gemma 4 26B on a Mac mini
Haven't had time to try yet, but heard from others that they needed to update both the main and runtime versions for things to work.
jasonjmcghee··on From 300KB to 69KB per Token: How LLM Architectures Solve the KV Cache Problem
> OpenAI applies it automatically and charges 50% less for cache hits

This is incorrect. It's 90% cheaper.

https://developers.openai.com/api/docs/pricing

jasonjmcghee··on Copilot edited an ad into my PR
What am I supposed to opt out of? The only setting in "Privacy" is "Suggestions matching public code" which is blocked and seems wholly unrelated to this.
jasonjmcghee··on Everything old is new again: memory optimization
Idk about "memory is cheap" but storage certainly is.

Whether you're talking about buying or renting in the cloud - RAM is 3 orders of magnitude more expensive

jasonjmcghee··on Whistler: Live eBPF Programming from the Common Lisp REPL
This is very cool.

I'm in danger of being nerd sniped.

jasonjmcghee··on ARC-AGI-3
v2 was a static fill in the blank task instead of v3 which is interactive.

There's world state that you can change. Not just place pixel.

Here's v2:

https://arcprize.org/tasks/ce602527

jasonjmcghee··on Apple Announces New Mac Sales Record Following MacBook Neo Launch
I certainly haven’t done anything outside of general happy path but i swapped out docker for it and it “just worked” - and i got more battery life
jasonjmcghee··on Apple announces new Mac sales record following MacBook Neo launch
0 improvements is unfair

https://github.com/apple/container

jasonjmcghee··on Push events into a running session with channels
I think they gave two sample implementations to demonstrate it.

I'm guessing their expecting the community to run with it

jasonjmcghee··on Drawvg Filter for FFmpeg
Didn't know this was a thing- thanks for posting.

Is there something similar that supports shaders? Like metal / wgsl / glsl or something?

Sounds like a fun project...

jasonjmcghee··on Push events into a running session with channels
This isn't quite as good unfortunately- you can't accept / deny permissions prompts.

Maybe there should be a Claude code that facilitates others that is connected. Like sub agents but can "choose what to do" on permissions check.

Or some other means to listen for permissions check

jasonjmcghee··on Push events into a running session with channels
I don't see this said anywhere - maybe I missed it. Why is it only one conversation at a time?

Couldn't you have multiple sessions using different plugins or whatever?

jasonjmcghee··on Leanstral: Open-source agent for trustworthy coding and formal proof engineering
That's great news- I wonder if that will help drive cloud costs down too
jasonjmcghee··on Leanstral: Open-source agent for trustworthy coding and formal proof engineering
Is it though?

Most people I know that use agents for building software and tried to switch to local development, every single time they switch back to Claude/codex.

It's just not worth it. The models are that much better and continue to get released / improve.

And it's much cheaper unless you're doing like 24/7 stuff.

Even on the $200/m plan, that's cheaper than buying a $3k dgx or $5k m4 max with enough ram.

Not to mention you can no longer use your laptop as a laptop as the power draw drains it - you'd need to host separately and connect

← PreviousPage 5 of 34Next →