Hacker News: randomblock1

New comment by randomblock1 in "LLM Ass Bench"

randomblock1 — Tue, 22 Sep 2026 22:03:45 +0000

Now put them on a bike.

New comment by randomblock1 in "Claude Opus 5.5"

randomblock1 — Tue, 22 Sep 2026 16:45:05 +0000

> On our benchmarks, Claude Opus 5.5 leads in agentic coding, computer use, and knowledge work. That said, at these levels of capability we’ve found that benchmark margins have become a less reliable guide to real-world differences. In our own use, the gap between Opus 5.5 and Claude Fable 5.1 is narrower than these scores suggest.

New comment by randomblock1 in "I don't like passkeys"

randomblock1 — Fri, 18 Sep 2026 18:46:37 +0000

It's amazon, I have 1password and it always asks me to create another security key

New comment by randomblock1 in "Bend – A language that blocks AI mistakes via proof, on CPU and GPU"

randomblock1 — Thu, 17 Sep 2026 21:24:33 +0000

Multiple times, even. Still no real reason why. https://github.com/bendlang/bend/activity?ref=main

One time they force pushed and erased everything except a 2-line README... on purpose.

Pre-obliteration version: https://github.com/bendlang/bend/tree/814453670d0e0d6777c131...

New comment by randomblock1 in "TSMC revealing details about next gen A14 node"

randomblock1 — Thu, 17 Sep 2026 19:18:27 +0000

It's not quite SRAM, it needs to be refreshed like DRAM. That's the main downside compared to SRAM but it's still very interesting.

New comment by randomblock1 in "Hister: A private search engine for the pages you visit and the files you keep"

randomblock1 — Thu, 17 Sep 2026 17:51:12 +0000

Nothing about this depends on the provider, you could spin up a local Qwen and give it a search tool like SearXNG or something. At this point local models are more than good enough for simple tasks like that. Using ChatGPT is just (usually) faster and simpler

New comment by randomblock1 in "Cognition launches new SWE-2 model, Rivaling Fable 5.1 and GPT-Astra"

randomblock1 — Mon, 14 Sep 2026 02:16:40 +0000

No it still requires Pro in the CLI. Only free model is SWE-1.6 slow

New comment by randomblock1 in "Why is the x86 undefined instruction called ud2? Why 2?"

randomblock1 — Mon, 14 Sep 2026 01:57:56 +0000

Schrodinger's instruction: simultaneously defined and undefined

New comment by randomblock1 in "Why is the x86 undefined instruction called ud2? Why 2?"

randomblock1 — Mon, 14 Sep 2026 01:55:52 +0000

0 undoubtedly comes first, but we all know 1 == true...

New comment by randomblock1 in "So you want to use OpenRouter?"

randomblock1 — Fri, 11 Sep 2026 23:41:53 +0000

You know how there's a router mode to use the cheapest provider? That only takes into account uncached rates, last I checked. Make another one that takes into account effective rates (the ones that include cache).

New comment by randomblock1 in "Cognition launches new SWE-2 model, Rivaling Fable 5.1 and GPT-Astra"

randomblock1 — Fri, 11 Sep 2026 09:35:38 +0000

Oh I tried in cloud. I'll give it another shot

New comment by randomblock1 in "More questions about whether researchers can trust OpenAI with unpublished math"

randomblock1 — Thu, 10 Sep 2026 17:35:37 +0000

So then they DIDN'T "learn the secret to cracking the problem". They simply knew that part of the problem was solved. Knowing a problem can be solved and knowing the solution are not the same thing.

New comment by randomblock1 in "Cognition launches new SWE-2 model, Rivaling Fable 5.1 and GPT-Astra"

randomblock1 — Thu, 10 Sep 2026 17:28:19 +0000

I just gave it a try and it doesn't appear to be free, it used up some of my on demand usage. It does say 75% off though. Seems like for Pro subscribers SWE-1.7 is free, maybe SWE-2 is free for them?

New comment by randomblock1 in "Searching for the best silicone USB cable"

randomblock1 — Wed, 09 Sep 2026 16:06:31 +0000

It's the website's "grid-bg" effect. It's glitching out and also normally barely visible at all.

New comment by randomblock1 in "Protecting Engineers' Skills in the AI Era"

randomblock1 — Thu, 03 Sep 2026 23:54:41 +0000

The writer is the CEO of... whatever this is: https://auraspark.com/

Completely slopped up website, zero human touch. I would even go so far as to call them a hyprocrite, for getting AI to do all that for them instead of, you guessed it, training a junior engineer to do it.

New comment by randomblock1 in "Show HN: FrontierHarness Eval – 9 harness, same model, cost per pass varies 17x"

randomblock1 — Wed, 02 Sep 2026 19:45:43 +0000

I think it's still a useful data point. For example, omp, which is pi with some default extensions, scores worse. I do agree that adding more configurations of Pi would help though.

New comment by randomblock1 in "Can I opt out of my input or output data being used for training?"

randomblock1 — Wed, 02 Sep 2026 19:40:14 +0000

The retain it for "the duration necessary to achieve the intended purposes", which could mean forever.

New comment by randomblock1 in "Apple caught off guard by AI demand for Mac Mini and Mac Studio"

randomblock1 — Mon, 31 Aug 2026 21:22:37 +0000

It's not that far off anymore. On my 7900 XTX 24GB, I can run Qwen3.8 27B with 131K context at Q4_K_M (55 tok/s with MTP). Excluding hardware cost, it's about $0.02 tok/M in and $0.40 tok/M out (cached in $0.0001). On OpenRouter, that would cost more than 10x what it actually costs me.

Of course, 131k context at 4-bit quant is a trade off, but even then, it's VERY capable. It doesn't feel that far behind something like GPT 5.6 Luna.

New comment by randomblock1 in "Previewing the Model Hardware Standard"

randomblock1 — Thu, 27 Aug 2026 21:15:49 +0000

To me, this seems like OPC UA / SiLA but instead of being software <-> machine semantics/control it's AI <-> machine semantics/control. Or in simpler terms, an AI-facing hardware abstraction / device-description standard.

This is how I understand it:

Agent <-> MCP/CLI/code <-> MHS <-> vendor API/SCPI/OPC UA/ROS/etc <-> CAN/Modbus/USB/etc <-> hardware

MHS can describe capabilites, metadata, safety limits, as well as provide read/write control and discovery. Things like "can measure temperature", "arm weighs X kg", "never exceed X RPM", that agents can easily understand. (as opposed to that being buried in a datasheet somewhere, or having to be included in the prompt)

Also see: https://en.wikipedia.org/wiki/OPC_Unified_Architecture, https://en.wikipedia.org/wiki/Standardization_in_Lab_Automat..., https://xkcd.com/927/

New comment by randomblock1 in "YouTube Format IDs"

randomblock1 — Wed, 26 Aug 2026 19:31:04 +0000

Anything with motion. Sports, games, vlogs, and so on experience immense improvements. And it's not like it takes 2x the bandwidth, because inter-frame compression can be smarter about it. It's like 30-50% more bandwidth.