Hacker News

points by spenvo 2 days ago | 0 comments

In Claude Code, even just dismissing the feedback may be considered "feedback". I wrote about it here: https://keydiscussions.com/2025/09/29/dont-even-dismiss-the-... The most common feedback on these posts is always "what did you expect?" ....... Man, the reason I signed up for one service over the other in the first place is often because of the promises these companies make on the front end, about what they do or don't do with my data. I agree that the bar is low in the industry but it doesn't mean we shouldn't expect more. Anthropic made very explicit, strong promises in this area <on the sign up/plans screen>, and so of course it disappoints me when I feel like the company I'm paying hundreds of dollars per month to is trying to sneak a peak. I literally think differently and self-censor when I think my work is being surveilled or stolen, which makes me less productive.

[–] sandeepkd 2 days ago

In a company as big as Anthropic, moving at current pace, the person/team writing this article and the team(s) collecting data would be way distant. It should be more than obvious that the data is being collected in not just one place but many (different teams are looking for different gems in that data). What part is used for training, how and when would be spread across the company way too wide for anyone to confirm with certainity.

  On the IP side, About 6 months ago I used Claude to help me draft a patent application for something which I believed was novel, I verified in the USPTO as well. Then out of being busy I dropped that idea. Couple weeks ago I picked up the work again, verified with Claude for the problem statement to see if there is any existing solution. It suggested my own solution back to me which was surprising since I only use in browser and delete the history explicitly. I asked it for references or resources and it could not find any. That helped me to understand the boundaries bit better.

[–] altmanaltman 2 days ago

There are companies with far more people than Anthropic that still stay coordinated. This is no excuse. Especially with customer data.

[–] j45 2 days ago

I have noticed this as well, when it can't help me and I thought it might be a difference in models, or using the same account somehow (tried a different one).

These models may not train directly on my chats and data, but in some cases it does seem to keep an impression of the pattern and process.

[–] ozgung 2 days ago

I was dismissing it immediately in order not to enter a number by mistake. What a dark pattern.

[–] dovetailed 2 days ago

The Claude Code data usage policy covers this - none of the options change your data retention or training opt-in: https://code.claude.com/docs/en/data-usage#session-quality-s...

"When you see the “How is Claude doing this session?” prompt in Claude Code, responding to this survey, including selecting “Dismiss”, records only your rating. We do not collect or store any conversation transcripts, inputs, outputs, or other session data as part of the rating prompt itself. Unlike thumbs up/down feedback or /feedback reports, this session quality survey is a simple product satisfaction metric."

[–] spenvo 2 days ago

Okay - that's good to know! Thanks for the comment.

In my defense: that webpage didn't exist at the time I wrote my blog post, appearing to first surface about 45 days after I wrote my post. https://web.archive.org/web/20250101000000*/https://code.cla... All that was known at the time was, from their privacy policy then: "We may use your feedback to analyze the effectiveness of our Services, conduct research, study user behavior, and train our AI models as permitted under applicable laws."

Great to know there's clarity on this - makes me feel better about Anthropic.

[–] dovetailed 1 day ago

Seems like a decent chance they posted this in response to your concerns!

[–] jrmg 2 days ago

I assumed that what they’re ’collecting’ there is at most stats on how much you like the product. Especially because, in my experience, if you say ‘good’ the next question is one that explicitly asks whether you permit Anthropic to analyze the session.

[–] j45 2 days ago

How would they know when to bring it up?

A determination of it has to be made some way beyond looking for frustrated or angry keywords.

If there was direct transparency regarding this it would be great vs digging into it on our own.

[–] jrockway 2 days ago

I have no idea if they're doing this, but you can always run the last turn through another LLM and ask it "should I request feedback based on this turn?"

I believe this is how the auto-mode classifier works. It sees a little bit of the transcript (with tool output removed, to avoid prompt injection) and answers "yes, allow the action" or "no, deny the action".

[–] j45 2 days ago

Interesting idea - I'm not sure if there's enough reliability that another LLM would reliably answer the same way over time since it's non-deterministic.

[–] ffsm8 2 days ago

Anthropics models wouldn't exist if they honored such promises. So yeah,

> The most common feedback on these posts is always "what did you expect?"

Is completely spot on. It's highly entertaining to me that some people can't wrap their head around that.

[–] loufe 2 days ago

That link is speculation at best. I wouldn't tell my family yet not to do that. Insane if true, but let's not jump at speculation.

[–] spenvo 2 days ago

The fact that Claude Code told me "thanks for the feedback … and thanks for helping to improve Anthropic’s models" when I pressed 0 to "dismiss" is not speculation, that's what the product did. I agree though that whether Anthropic actually considers it "feedback" for the purposes of a privacy loophole (vs. it being a sloppy response) is anyone's guess.

[–] randomblock1 2 days ago

Yeah given how CC is vibe coded, the feedback message always appearing is almost certainly just a weird implementation detail.

[–] bix6 2 days ago

Counterclaim? I never hit feedback because it’s all dark patterned.

[–] hmokiguess 2 days ago

That's very deceiving and should warrant a class action law suit in my opinion

[–] Zambyte 2 days ago

How many months do you plan on spending hundreds of dollars a month? You should probably just own your compute. It'll probably cost you less, and you'll have no concern over who has access to your chats.

[–] jitl 2 days ago

$200/mo @ 1 year = $2400 2400 gets me about 1/2 of a 128gb NVIDIA DGX Spark, so i could do a year of nonstop inference of (checks notes) quantized Qwen 3.8-27B, a model i would never choose to use. yay, such savings.

[–] lp92 1 day ago

And that Spark isn't running a top end frontier model. Yes, the open weights smaller models that can fit on Spark are getting better, but they are still quite a way off in performance depending on use case.

[–] spenvo 2 days ago

I don't know, but I do expect to eventually switch over. That said: the upfront time to setup (and continued fiddling) and cost of a high-end local compute system represent major investments, when I could just... keep being insanely productive with frontier AI in CC. And models/harnesses can themselves be security liabilities (see Mistral, LiteLLM, etc recent supply chain attacks). So I'm personally waiting for the local AI scene to mature, so I can be more precise about how I take the plunge. I will say: if the subscription pricing ever goes away or materially changes for Anthropic, I'm headed straight to order whatever the best local setup is, within my budget.

[–] pixl97 2 days ago

A single video card with any amount of ram is in the thousands of dollars, and you'd need a stack of them to replicate what the providers performance.

Running your own compute has a lot of risks too, you have to keep it secure against attacks on the models that you're running. Giving away all your chats to hackers because you don't secure your own stuff isn't really winning either.

[–] oblio 2 days ago

How much does it cost to run your own hardware such that the service is comparable to Claude? Similar speed, similar features, web chat access, CLI access?

[–] Zambyte 2 days ago

How much do people actually do that demands the capabilities of Claude instead of open models that can run on hardware in the range of thousands of dollars (which is not that much when the alternative is hundreds of dollars a month, indefinitely)?

People chasing the best of the best models for all tasks in order to "keep up" are only wasting resources at this point.

Regarding similar features, there are lots of off the shelf harnesses that support all that, or can easily be extended (literally just ask your agent) to do it.

[–] jitl 2 days ago

i am 100% sure i need opus 5.5 model to build stuff like this because i try lesser models because i want my tokens to go farther and lesser models simply cannot: https://fujimod.com

[–] lp92 1 day ago

Not sure if this is possible on the Fuji, but back in the old days of the Canon PowerShot being very relevant, there was CHDK firmware which ran off the SD card on boot. No physical impact to the device and fully reversible by reformatting the SD card. This would be the way to go to reduce bricking chances.

[–] jitl 23 hours ago

yeah i currently only distribute sd card dual boot media where the the install process is “turn it on holding REC and pressing rear command dial” and uninstall process is “turn it off”

[–] Zambyte 2 days ago

> All of this is unreviewed vibe-coded slop that could be hazardous to your camera’s mental, emotional, or digital health. If you're uncomfortable with risk to your camera, please don't use it.

You don't care anyways. Actual engineering works "100%" fine with small models now. You should try it!

[–] lp92 1 day ago

I've tried smaller models and the best open models for a specialized UAV focused CAD + CFD + DFM optimization application I'm building.

The best open weights models still had a lot of defects and very egregious mistakes to fix even after I provided a very well detailed Requirement/Design/ Technical Implementation plans.

Opus and Gemini 3.8 (and now Argon) did a significantly better job. So I've baseline using these two models to build this application. I now use Opus for design/review and Flash 3.8 for the code gen/implementation work.

I'm sure small models and even open big models are probably fine for many common small apps and such, but they aren't really able to perform the same across all use cases.

[–] jitl 2 days ago

as i said, i did try it. smaller models are not up to the task of deep reverse engineering work.

[–] tekchip 2 days ago

Plenty of packages do most or all of that. OpenCode does most of that. It's replaced Claude tools for me.

[–] oblio 1 day ago

OpenCode offers a web UI?

[–] tekchip 1 day ago

https://opencode.ai/v2/docs#web

[–] oblio 22 hours ago

That documentation is extremely sparse. Can I run that 24/7 and access it anywhere around the world? Is it secure?