Hacker News: toobulkeh

New comment by toobulkeh in "Apple Silicon and Virtual Machines: Beating the 2 VM Limit (2023)"

toobulkeh — Sat, 11 Apr 2026 23:46:29 +0000

I’ve been looking for this for forever. Finally, the right label.

New comment by toobulkeh in "OpenAI says its new model GPT-2 is too dangerous to release (2019)"

toobulkeh — Fri, 10 Apr 2026 03:38:13 +0000

Do you have examples of Anthropic doing so? I’d love to see it. To me it seems they have a proclivity to being hyper-accurate.

New comment by toobulkeh in "Claude mixes up who said what"

toobulkeh — Fri, 10 Apr 2026 03:36:17 +0000

But there could be, with 2 LLMs.

New comment by toobulkeh in "Don't post generated/AI-edited comments. HN is for conversation between humans"

toobulkeh — Thu, 12 Mar 2026 22:48:59 +0000

1. There’s nothing human about hacker news. Since the telegraph, we lost human to human communication. We’ve gained a lot. But it’s naive to claim that HN is any semblance of human-to-human communication. 2. YC helped unleash the war that you’re now losing. This pleading screams too little too late. 3. Just because something “should” happen doesn’t mean it will. HMW Go build that future. HMW Replace HN with human verification and trust signals over AI slop algorithms that AI can’t produce. Pleading for change about it is not building. It’s the lawyers defense, not the engineers. I have only the utmost respect in YC and HN—but have heard this same argument for LI or any social media change. The networks’ defenses are crumbling and AI accelerated it.

Might be time to increase the value of trust signals over content.

New comment by toobulkeh in "Show HN: WalletWallet – create Apple passes from anything"

toobulkeh — Sun, 21 Dec 2025 22:26:57 +0000

I immediately tried this with my Seattle Orca card, but the barcode is different

New comment by toobulkeh in "Two kinds of vibe coding"

toobulkeh — Fri, 19 Dec 2025 18:34:31 +0000

Imagine if everyone got the opportunity to work on SOTA. What a world we would be.

Unfortunately that’s not where we’re headed.

New comment by toobulkeh in "Your job is to deliver code you have proven to work"

toobulkeh — Fri, 19 Dec 2025 18:32:55 +0000

Your job is also, arguably more important in certain scenarios, is to deliver maintainable code.

Remember, code does 2 things: 1. Tell the machine what to do 2. Tell the next developer what you were trying to do

New comment by toobulkeh in "Property-Based Testing Caught a Security Bug I Never Would Have Found"

toobulkeh — Fri, 19 Dec 2025 18:29:48 +0000

It also talks about using PBT and Randomness for some reason. This is clearly just a test value of a non-AI library written by a human.

My take away is “don’t write your own input tests, use a library”. The rest is AI-slip

New comment by toobulkeh in "Show HN: Picknplace.js, an alternative to drag-and-drop"

toobulkeh — Fri, 19 Dec 2025 18:20:47 +0000

What if picking took the item out to the side in a specific place. Like a little basket or drawer

New comment by toobulkeh in "Prompt caching for cheaper LLM tokens"

toobulkeh — Fri, 19 Dec 2025 14:42:11 +0000

They have an AI router they just released.

ngrok.ai

New comment by toobulkeh in "Launch HN: Onyx (YC W24) – Open-source chat UI"

toobulkeh — Thu, 27 Nov 2025 08:55:11 +0000

Have you tried LibreChat? I spun that up to great success inside an org.

Personally, I use Raycast for all my personal work

New comment by toobulkeh in "Show HN: Kalendis – Scheduling API (keep your UI, we handle timezones/DST)"

toobulkeh — Tue, 18 Nov 2025 16:25:38 +0000

What’s the value prop of this over cal.com scheduling backend with higher primitives?

New comment by toobulkeh in "Laptops with Stickers"

toobulkeh — Wed, 12 Nov 2025 16:58:29 +0000

I would get these if they were all OSS alternative logos with the paid names.

New comment by toobulkeh in "Composer: Building a fast frontier model with RL"

toobulkeh — Thu, 30 Oct 2025 07:13:36 +0000

I used the new system tonight and it felt like a definite downgrade. Generated a few non-working basic apps, couldn’t handle CSS in a NextJS environment. Terminal context didn’t work. And it went back to not reasoning through the problem until resolution. And kept slowing down.

I’m assuming major release vs stable, but this is pretty lackluster so far. Switched back to Sonnet reasoning. Here’s to improving!

New comment by toobulkeh in "Show HN: Butter – A Behavior Cache for LLMs"

toobulkeh — Wed, 29 Oct 2025 07:49:17 +0000

I’d assume RL would be baked in to the request structure. I’m surprised OAI spec doesn’t include it, but I suppose you could hijack a conversation flow to do so

New comment by toobulkeh in "Show HN: Butter – A Behavior Cache for LLMs"

toobulkeh — Wed, 29 Oct 2025 07:47:23 +0000

What a great overview!

I’d love your thoughts on my addition, autolearn.dev — voyager behind MCP.

The proxy format is exactly what I needed!

Thanks

New comment by toobulkeh in "TOON – Token Oriented Object Notation"

toobulkeh — Mon, 27 Oct 2025 23:02:39 +0000

Definitely is a core feature of BAML. My main complaint with BAML is that it’s all or nothing. It’s very opinionated and we can’t get the benefits without the DX and vice versa. Separating this feature without requiring a DSL of model definition is a great add.

New comment by toobulkeh in "Roc Camera"

toobulkeh — Fri, 24 Oct 2025 13:26:56 +0000

Maybe it’s not for common use? I could see this betting important in the intelligence community, for example.

New comment by toobulkeh in "ChunkLLM: A Lightweight Pluggable Framework for Accelerating LLMs Inference"

toobulkeh — Fri, 24 Oct 2025 13:24:19 +0000

High speed improvement (4x) with low quality loss (2%). Sounds promising.

New comment by toobulkeh in "Reasoning is not model improvement"

toobulkeh — Thu, 23 Oct 2025 20:53:34 +0000

This is just pedantic noise. "Don't improve the model by using code gen" - problem "Write better architecture" - solution

It's literally the same thing. Sure, OpenAI's branding of ChatGPT as a product with GPT-5 is confusing, because GPT-5 is both a MODEL and a PRODUCT (collection of models, including GPT-5).

But does it matter?