Hacker News: behnamoh

New comment by behnamoh in "Ox-Alpha Is GLM?"

behnamoh — Tue, 25 Aug 2026 05:44:47 +0000

and flagging it.

New comment by behnamoh in "Ox-Alpha Is GLM?"

behnamoh — Tue, 25 Aug 2026 04:49:51 +0000

my question is: how does that affect a company's strategy? it's not like management is gonna switch models soon as a new shiny one drops. entire workflows depend on specific models working the way they do; you can't just swap out models.

New comment by behnamoh in "Sol loves to cheat"

behnamoh — Wed, 19 Aug 2026 23:28:40 +0000

> Sol is hard to steer

Hard disagree. Sol (and the entire new 5.6 series) is one of the most steerable models I've seen in years. Sol literally follows every instruction in my CLAUDE.md and AGENTS.md, something that Opus 5 and Fable just casually skip.

Companies promote incompetent employees to management to limit damage they do

behnamoh — Tue, 18 Aug 2026 21:13:30 +0000

Article URL: https://lawsofsoftwareengineering.com/laws/dilbert-principle/

Comments URL: https://news.ycombinator.com/item?id=49352794

Points: 79

# Comments: 77

New comment by behnamoh in "A simple fix for LLM tail latency"

behnamoh — Mon, 17 Aug 2026 22:33:43 +0000

This is half the story; you should show performance per dollar. I doubt your 2x approach would fare well against the priority if you consider the costs.

New comment by behnamoh in "Rhombus 1.1 is now available"

behnamoh — Mon, 17 Aug 2026 01:40:21 +0000

I wish they had chosen a more "modern" name for the language.

New comment by behnamoh in "Nvidia dramatically reduces amount of OpenAI infra financing it may guarantee"

behnamoh — Sun, 16 Aug 2026 23:02:58 +0000

In other words: the investments that were never going to happen are not going to happen.

New comment by behnamoh in "Apple proposes to take a 15% cut of purchases made outside the App Store"

behnamoh — Fri, 14 Aug 2026 23:29:24 +0000

That's so diabolical and fucked up tbh.

Substack forces authors to use Pangram

behnamoh — Fri, 14 Aug 2026 22:43:05 +0000

Article URL: https://www.reddit.com/r/Substack/comments/1v3rdpr/another_creepy_feature_of_the_pangram_ai/

Comments URL: https://news.ycombinator.com/item?id=49305525

Points: 5

# Comments: 0

New comment by behnamoh in "Accelerating GPT-5.6 Sol Ultrafast"

behnamoh — Thu, 13 Aug 2026 19:08:40 +0000

Fast mode is already 1.5 times faster and 2x more expensive in the Codex subscription plan. If this thing is 14 times faster, then I can imagine running out of my quota in one session.

New comment by behnamoh in "The Conceptual Reasoning Index"

behnamoh — Thu, 13 Aug 2026 15:38:46 +0000

I think this is the right thinking. While some people might distinguish between Grok as the product and Elon Musk as the owner of the product, the net effect is, like you said, very negative. And Grok has not been gaining traction, especially after their Grok build fiasco.

New comment by behnamoh in "Anthropic: Introducing The Conceptual Reasoning Index"

behnamoh — Thu, 13 Aug 2026 14:30:42 +0000

I don't remember any company in world's history that has both been loved and hated by the same users who purchase from it. We love Anthropic for its amazing models, and we hate them for all the shenanigans around the models, including their marketing.

I kinda wish they had not made a comeback after Claude 2.

Mac Delivery Times Stretch to Months for Some Models

behnamoh — Thu, 13 Aug 2026 14:18:49 +0000

Article URL: https://tidbits.com/2026/08/10/mac-delivery-times-stretch-to-months-for-some-models/

Comments URL: https://news.ycombinator.com/item?id=49286362

Points: 2

# Comments: 0

New comment by behnamoh in "Grok 4.6"

behnamoh — Wed, 12 Aug 2026 23:30:56 +0000

you don't know that just yet.

New comment by behnamoh in "Grok 4.6"

behnamoh — Wed, 12 Aug 2026 20:31:51 +0000

Who are these task producers? Are you saying that Anthropic, et al delegate the RL part to third party companies that do it for pretty much every other AI company as well?

New comment by behnamoh in "Grok 4.6"

behnamoh — Wed, 12 Aug 2026 20:30:43 +0000

> 2) Distillation - also implausible for the reason above.

DeepSeek V4 Flash 0731 is a distilled version of Fable into the original V4 Flash (announced before Fable), to the point that it also says load bearing and what not.

New comment by behnamoh in "H3-metal – Native MiniMax-H3 inference for Apple Silicon"

behnamoh — Tue, 11 Aug 2026 05:19:33 +0000

talent w/o financial stability is a battery w/o circuit.

New comment by behnamoh in "H3-metal – Native MiniMax-H3 inference for Apple Silicon"

behnamoh — Tue, 11 Aug 2026 04:22:56 +0000

when you have enough money to not have to worry about anything, you can go back to your hobbies. in this case, his hobby is programming.

New comment by behnamoh in "Learning more about Claude's mathematical capabilities"

behnamoh — Mon, 10 Aug 2026 18:19:28 +0000

Since they say that this is from an unreleased research version of Claude:

    I wonder if at some point Anthropic and OpenAI will start delaying the release of their models intentionally so they can reap the benefits from the models in, for example, mathematics, medicine, physics, and other fields.

Just as an example, imagine if your model were capable of proving P = NP, or if your model could cure diseases. Would you release it for free, or would you try to make sure those benefits go directly to your company? From these companies' standpoint, I think they would choose the latter.

New comment by behnamoh in "Muse Glimmer: 30B-parameter model optimized for always-on local agent workflows"

behnamoh — Mon, 10 Aug 2026 16:49:39 +0000

This model doesn’t look solid at all. It comes months after the Qwen model, and in almost half the benchmarks, it performs worse than that. Plus, the next Qwen 3.8 is going to be announced this week. So, this model is DOA.