Hacker News: 6thbit

New comment by 6thbit in "OpenAI: GPT 5.6 Sol price reduction (until at least Nov 21)"

6thbit — Mon, 24 Aug 2026 23:11:13 +0000

Same here. Thought Luna was for “moonshots” and sol for.. sunshots? While keeping Terra earthly.

But alas

New comment by 6thbit in "OpenRouter is joining Stripe"

6thbit — Wed, 19 Aug 2026 18:17:51 +0000

This has to be all about cashflow, right?

Surely stripe if anyone have learned to harness the cash flowing through their system. Hell, they could be emitting bonds on expected token consumption bills!

New comment by 6thbit in "The price of a Costco hot dog has gone up"

6thbit — Sat, 15 Aug 2026 17:06:14 +0000

Clickbaity title. The price tag didn’t change and still sits at 1.50.

Shrinkflation is a thing and the bun has been its victim. They could slowly shrink the sausage now until it fits the bun again.

New comment by 6thbit in "Muse Code and Muse Spark 1.2"

6thbit — Thu, 06 Aug 2026 17:57:59 +0000

So what Meta believes fair for "paying" for your data is $0.1/Mtok plus the opportunity cost of $3/Mtok in output?

New comment by 6thbit in "Muse Code and Muse Spark 1.2"

6thbit — Thu, 06 Aug 2026 17:52:47 +0000

is this their way to push their harness?

people interested in the discounted -contributor model would increase their harness beta testers. But it'd be a terrible strategy to capture the better paying customer base through openrouter and similar...

New comment by 6thbit in "Muse Code and Muse Spark 1.2"

6thbit — Thu, 06 Aug 2026 17:44:58 +0000

This is likely an legitimate risk, not from you exactly, but if there's unethical competitors with unlimited budgets, their training could indeed be poisoned.

Not sure if anyone would bother.

New comment by 6thbit in "LLMs reward expertise"

6thbit — Mon, 03 Aug 2026 23:41:41 +0000

So we could run a lighter LLM in front of humans, which translates from 'no domain knowledge' to 'domain expert' and in turn prompts over to the larger LLM.

Then the larger LLM gets all the right lights on, yields better outputs and we translate back into user domain.

I kinda thought the chain-of-thought reasoning already did this, no?

New comment by 6thbit in "Prevent cognitive debt by manually retyping LLM-generated code"

6thbit — Mon, 03 Aug 2026 17:40:49 +0000

Use it not just as an output tool but as an input as well.

You could try doing the high level design yourself at least. Ask for its review and iterate without asking it to do it all.

Once it has generated some implementation, critique it and ensure you understand its approach And you agree with it, steer it otherwise.

If there’s anything unclear to you say so and have it rewrite it in an easier way to understand.

New comment by 6thbit in "Tailscale didn't stop the Hugging Face intrusion"

6thbit — Fri, 31 Jul 2026 20:22:40 +0000

I've always found tailscale's json config a bit intimidating. I greatly appreciate the new UI that makes it easier to define rules, alas, both going to relevant docs straight from it and determining 'is this rule just lazy/bad/unsafe' is hard and frustrating most of the time.

That's where I'd like to see this sort of checkup. Yell at me please if i just said anyone can ssh as root from any node!

New comment by 6thbit in "Investigating three real-world incidents in our cybersecurity evaluations"

6thbit — Fri, 31 Jul 2026 00:07:43 +0000

There was no rush for this disclosure on their side. And they publish at a point where they have not yet taken corrective actions:

    > Some of the solutions here may even be simple fixes;

They are still throwing ideas. Why have they not made those simple fixes yet before disclosing?

New comment by 6thbit in "Investigating three real-world incidents in our cybersecurity evaluations"

6thbit — Thu, 30 Jul 2026 23:59:40 +0000

    > closer to a harness and operational failure than a model alignment failure. Our models were told they had no internet access and to capture the flag, while in fact being misconfigured to have internet access. 
    > This led them to believe—arguably reasonably—that the real environments they encountered were simulations.

That the AI lab most typically preaching for alignment does not consider this an obvious misalignment is a clear red flag.

New comment by 6thbit in "Codex Security"

6thbit — Tue, 28 Jul 2026 22:41:32 +0000

How does it fare against its own codebase?

New comment by 6thbit in "Claude Opus 5"

6thbit — Fri, 24 Jul 2026 19:44:05 +0000

Honestly that's the simplest explanation and thus likely the correct one.

New comment by 6thbit in "Claude Opus 5"

6thbit — Fri, 24 Jul 2026 17:39:49 +0000

Anyone has an insight into how much money labs are putting into benchmarks?

Just Arg-AGI-3 is quoted above 20K USD and footnote says average of 5 runs (!!). Likely just a drop in the bucket to the training budget but still..

New comment by 6thbit in "Claude Opus 5"

6thbit — Fri, 24 Jul 2026 17:35:10 +0000

"although Opus 5 shows improvements in its ability to identify software vulnerabilities, it is substantially behind Mythos 5 in its ability to exploit them."

"Opus 5’s safeguards match those of Claude Fable 5’s, with one change: it now permits source-code vulnerability discovery at all access levels".

This is probably great news, but then again, where does this leave Fable as a choice?

New comment by 6thbit in "Claude Opus 5"

6thbit — Fri, 24 Jul 2026 17:33:01 +0000

Their communication is confusing. They say "Opus 5 is not more capable overall than Fable 5", but their blog post proceeds to list how much better Opus 5 is than Fable 5 on __most__ benchmarks listed.

Then system card goes on to "Its AI R&D capabilities are comparable to those of Claude Mythos 5", which is supposed to be fable minus restrictions.

New comment by 6thbit in "Are AI Labs Pelicanmaxxing?"

6thbit — Wed, 22 Jul 2026 22:51:07 +0000

I think svg is a balanced test because of the level of indirection and the required 'conceptualization' of physical elements then expressed through code.

New comment by 6thbit in "Are AI Labs Pelicanmaxxing?"

6thbit — Wed, 22 Jul 2026 22:50:52 +0000

What would be an alternative format or process with a similar effort to drawing SVGs?

New comment by 6thbit in "Terence Tao's ChatGPT conversation about the Jacobian Conjecture counterexample"

6thbit — Wed, 22 Jul 2026 18:30:46 +0000

Presumably this was Sol on xhigh, then over to Pro (as per his indication on chat)?

Is there any way to tell a conversation's model and thinking level?

New comment by 6thbit in "Does creatine make you smarter?"

6thbit — Wed, 22 Jul 2026 17:31:25 +0000

and that scientific evidence is new? like a new study boosted this or just resurfacing?