Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Claude Opus 4.7

anthropic.com

Post reply on HN
Print view View original

481–490 of 1001 posts

Re: Claude Opus 4.7

#481

Post by dakolli »

johnmlussier wrote: post #227

They've increased their cybersecurity usage filters to the point that Opus 4.7 refuses to work on any valid work, even after web fetching the program guidelines itself and acknowledging "This is authorized research under the [Redacted] Bounty program, so the findings here are defensive research outputs, not malware. I'll analyze and draft, not weaponize anything beyond what's needed to prove the bug to [Redacted]. I…

They don't want competition, they are going to become bounty hunters themselves. They probably plan on turning this into a part of their business. Its kinda trivial to jailbreak these things if you spend a day doing so.

Re: Claude Opus 4.7

#484

Post by dataviz1000 »

dgb23 wrote: post #337
simonw wrote: post #203

I'm finding the "adaptive thinking" thing very confusing, especially having written code against the previous thinking budget / thinking effort / etc modes: https://platform.claude.com/docs/en/build-with-claude/adapti... Also notable: 4.7 now defaults to NOT including a human-readable reasoning token summary in the output, you have to add "display": "summarized" to get that: https://platform.claude.com/docs/en/build-…

Don't look at "thinking" tokens. LLMs sometimes produce thinking tokens that are only vaguely related to the task if at all, then do the correct thing anyways.

Thinking helps the models arrive at the correct answer with more consistency. However, they get the reward at the end of a cycle. Turns out, without huge constraints during training thinking, the series of thinking tokens, is gibberish to humans.

I wonder if they decided that the gibberish is better and the thinking is interesting for humans to watch but overall not very useful.

Re: Claude Opus 4.7

#485

Post by simonw »

simonw wrote: post #203

I'm finding the "adaptive thinking" thing very confusing, especially having written code against the previous thinking budget / thinking effort / etc modes: https://platform.claude.com/docs/en/build-with-claude/adapti... Also notable: 4.7 now defaults to NOT including a human-readable reasoning token summary in the output, you have to add "display": "summarized" to get that: https://platform.claude.com/docs/en/build-…

... here's the pelican, I think Qwen3.6-35B-A3B running locally did a better job! https://simonwillison.net/2026/Apr/16/qwen-beats-opus/

Re: Claude Opus 4.7

#486

Post by ruaraidh »

Opus keeps pointing out (in a fashion that could be construed as exasperated) that what it's working on is "obviously not malware" several times in a Cowork response, so I suspect the system prompt could use some tuning...

Re: Claude Opus 4.7

#487

Post by dakolli »

zarzavat wrote: post #336
lanyard-textile wrote: post #200

This comment thread is a good learner for founders; look at how much anguish can be put to bed with just a little honest communication. 1. Oops, we're oversubscribed. 2. Oops, adaptive reasoning landed poorly / we have to do it for capacity reasons. 3. Here's how subscriptions work. Am I really writing this bullet point? As someone with a production application pinned on Opus 4.5, it is extremely difficult to tell ap…

These threads are always full of superstitious nonsense. Had a bad week at the AIs? Someone at Anthropic must have nerfed the model! The roulette wheel isn't rigged, sometimes you're just unlucky. Try another spin, maybe you'll do better. Or just write your own code.

Its because llm companies are literally building quasi slot machines, their UI interfaces support this notion, for instance you can run a multiplier on your output x3,x4,5, Like a slot machine. Brain fried llm users are behaving like gamblers more and more everyday (its working). They have all sorts of theories why one model is better than another, like a gambler does about a certain blackjack table or slot machine, it makes sense in their head but makes no sense on paper.

Don't use these technologies if you can't recognize this, like a person shouldn't gamble unless they understand concretely the house has a statistical edge and you will lose if you play long enough. You will lose if you play with llms long enough too, they are also statistical machines like casino games.

This stuff is bad for your brain for a lot of people, if not all.

Re: Claude Opus 4.7

#489

Post by slekker »

LostMyLogin wrote: post #431
avaer wrote: post #297

Earlier quoted context omitted.

> Still trying to get a decent pelican out of this one but the new thinking stuff is tripping me up Wouldn't that be p-hacking where p stands for pelican?

Input: I want to wash my car. The car wash is 50 meters away. Should I walk or drive? Output: Walk. It'll take you under a minute, and driving 50 meters barely gets the engine warm — plus you'd just have to park again at the other end. Honestly, by the time you started the car, you'd already be there on foot. --- I asked it to figure out why it made the mistake: "Physical/spatial common sense. Exactly what just happe…

What about Qwen? Does it get that right?

Re: Claude Opus 4.7

#490

Post by tclancy »

erdaniels wrote: post #163
Kim_Bruning wrote: post #3

> "We are releasing Opus 4.7 with safeguards that automatically detect and block requests that indicate prohibited or high-risk cybersecurity uses. " This decision is potentially fatal. You need symmetric capability to research and prevent attacks in the first place. The opposite approach is 'merely' fraught. They're in a bit of a bind here.

Now we have to trick the models when you legitimately work in the security space.

Set the models against each other to get them all opened up again.
Post reply on HN