scosman

15 ms·

271.
by scosman 1y ago
Great work
272.
by scosman 1y ago
You were specific. Don't start throwing around insults for reading what you wrote and pointing out it seemed like a weird statement. There's more than 1 device that run's GrapheneOS so the comment didn't make sense unles
273.
by scosman 1y ago
A UI framework? What’s the other $608m for?
274.
by scosman 1y ago
Zero iOS devices are produced in the US. Production is all China/India.
275.
by scosman 1y ago
That means there are infinitely more devices produced in the US that can run graphene than can run iOS. Seems like a weird metric.
276.
by scosman 1y ago
+1. Smartphones aren't a monopoly. GrapheneOS is a thing. More companies can build hardware for it if there's demand. Not every piece of hardware needs to be general purpose computer. I've been delighted to get my parents on
277.
by scosman 1y ago
How does the process management comparison work for GPU vs full systems?
278.
by scosman 1y ago
No one dismissed asking a question. I'm pushing back against a logical hole in the argument. Even with better healthcare, everyone's heath eventually fails. Using something that happens to 100% of people as an indication of anythi
279.
by scosman 1y ago
I don't follow the author's logic. They seem to assume anyone choosing MAID has been failed by the healthcare system. While that's certainly possible for some of cases, every single person eventually reaches a point where the
280.
by scosman 1y ago
None of this comment is true. Compilation works fine without notarization. It isn't called by default for the vast majority of complications. It is only called if you submit to an App Store, or manually trigger notarization. The notari
281.
by scosman 1y ago
Notarization isn’t called for the vast majority of builds on the vast majority of build systems. Your local dev builds don’t call it or require it. It’s only needed for release builds, where you want it notorized (required on iOS, highly re
282.
Show HN: Breaking IDE silos with MCP for prompts and dev-tools
1 points by scosman 1y ago|0 comments
283.
by scosman 1y ago
I made a MCP server that tries to address some of these (undocumented, security, discoverability, platform specific). You write a yaml describing your tools (lint/format/test/build), and it exposes them to agents MCP. Kinda l
284.
by scosman 1y ago
Agreed. This is the only thing stopping it from being the undisputed best terminal for me. I’ve referred 2 people to it in the last few months, both like it but didn’t adopt it because of this.
285.
by scosman 1y ago
I saw you may rename your project. IANAL, but don't do that. They might have invalidated a European registered trademark, but you have first use and continuous worldwide use which means a lot. Not all trademarks need to be registered.
286.
by scosman 1y ago
This is the comment people should read. GG is amazing. Ollama forked to get it working for day 1 compatibility. They need to get their system back in line with mainline because of that choice. That's kinda how open source works. The up
287.
by scosman 1y ago
But the price difference is only $38…
288.
by scosman 1y ago
He's comparing to a studio when he should compare to the mini for this performance. They are almost the same price at a 64gb RAM + 500gb storage config (CAD). - Framework, Max+ 64GB: $2,861.16 - Apple Mini M4 Pro, 64GB: $2,899.00 Apple
289.
by scosman 1y ago
The qwen coder CLI gives you 1000 free requests per day to the qwen coder model (405b). Probably the best free option right now.
290.
by scosman 1y ago
I liked the idea a lot! But yeah, saying no-servers when there are servers will get ya roasted here. You got me thinking about this and it was fun to whip up.
291.
by scosman 1y ago
Okay, for fun I took a stab at this problem removing the server (using public keys instead), and minimizing the dependencies: https://news.ycombinator.com/item?id=44846495
292.
Show HN: SecretShare – Easy, secure one time secret sharing CLI
3 points by scosman 1y ago|2 comments
293.
by scosman 1y ago
There's a bunch of benchmarks on the intro page including AIME 2025 without tools, SWE-bench Verified, Aider Polyglot, MMMU, and HealthBench Hard (not familiar with this one): https://openai.com/index/introducing-g
294.
by scosman 1y ago
I feel like I’d rather send “uuid:cipertext” so the cipertext never touches a server, but logically the security seems the same.
295.
Eval-maxing an AI FFmpeg command generator(getkiln.ai)
6 points by scosman 1y ago|0 comments
296.
by scosman 1y ago
even simplier: 1) It becomes essential for workflows while it cost $1 2) OpenAI can increase price to any amount once they are dependent on it, as the cost for changing workflows will be huge Giving it to them for free skews the cost/b
297.
by scosman 1y ago
> GGML library is llama.cpp. They are one and the same. Nope…
298.
by scosman 1y ago
I build an app against the Ollama API. If this will let me test all Ollama models, I'm so in.
299.
Show HN: Give coding agents MCP access to lint/test/format with 1 YAML file
2 points by scosman 1y ago|0 comments
300.
by scosman 1y ago
Anyone get this working in Cursor? I can connect openrouter just fine, but Cerebras just errors out instantly. Same url/key works via curl, so some sort of Cerebras/Cursor compatibility issue.

More ›