Hacker Newsnew | past | comments | ask | show | jobs | submit | llimllib's commentslogin

> Does it work well yet?

no, definitely not


note that if you want range queries in duckdb-wasm to work, you currently need to make sure that you disable `builtin_httpfs`, otherwise you can end up pulling a ton of data: https://github.com/duckdb/duckdb-wasm/issues/2228#issuecomme...


This week I've been debugging a performance issue in a Ducklake and knocking out some good improvements, but this precise symptom—read_parquet() making tons of tiny range requests to load a parquet's footer from S3—was still baffling me. How lucky am I that I stumbled on your comment! Amazing, thank you for the pointer.


> read_parquet() making tons of tiny range requests to load a parquet's footer from S3

This is somewhat unavoidable if you have a number of files and a tight filter which rejects most of them. The footer choice was a bad one in hindsight (not for HDFS), but for anything which uses XOR or other kind of erasure coding. The other other file format I worked with used headers only instead of footers, to save a few ms in latency for the initial GET

I don't think Ducklake can maintain a zone-map or ART index over parquet files, but there is no equivalent of iceberg metadata/something which has min/max/HLL compacted across multiple files yet.

The secondary indexes with file identifiers stored in a sequential file is most likely the next useful thing duckdb can make.


Apple's on-device transcription model is about 3x slower than their reported results on my more powerful computer, with a similar error rate in my test (which I consider pretty good results)

I'll test theirs too if I get around to it

Here's a sample CLI wrapping the apple model and providing a simple benchmark: https://github.com/llimllib/transcribe

update: doesn't seem like they provide their model and I don't feel like trying to dig it out of their app bundle, so I'm not going to test this one


Is it even worth using apple's? Maybe they have improved it but when they first started using it to transcribe voice memos it was laughably bad (in my experience). I defaulted to whisper.cpp


apple's is 5x faster than `large-v3-turbo` with whisper.cpp and in my test had somewhat worse performance, so it's a tradeoff


I used it a lot in my AI client before I finally got around to hacking a Deepgram API server into Handy: It sucks enough to give me a headache worrying about its unreliability in every given transcription.

They supposedly improved it on later Apple Intelligence enabled devices, but I don’t have one.


my tests are against the Apple Intelligence model, fwiw, not any earlier model


I suspect Apple's strengths are in multi-language support.

I use their on-device translation in my apps. Works pretty well. ChatGPT's context-aware translation is better, but it's not on-device.


With MacWhisper you can select the Apple models, it took almost 4x as long to transcribe the same file versus this software and MacWhisper/Parakeet3.


interesting! I only tested against Parakeet v2 because my agent suggested it would be faster for English transcription (v3 is multilingual but somewhat slower, it suggested)

On my bench, parakeet-tdt-0.6b-v2 is 2x slower

I wonder if you're using different Apple models when you go through that app? I don't know


The specific language is:

> projects that mostly consist of code written by "generative AI"-tools

And "mostly" here seems egregiously undefined to me

https://codeberg.org/Codeberg/org/commit/96fac426a32d1ba91ff...


implies for me >50% - instead of >0% as suggested by @ricardobeat


The hardest part of policy isn't coming up with smart policy, it's figuring out where and how to draw the lines.


What do you think "mostly" could mean?


Take a peek at https://codeberg.org/DAWO/DAWO-Core/commits/branch/main

Is this "mostly developed with AI" in your opinion? Not to mention the fact the policy leaves it open to interpretation is part of the problem...


As much as I hate vibe-coded "slop". I feel like this is idiotic. I and most good engs with decades of experience do use AI now to type nearly all the actual "code", but we read EVERY SINGLE line, always fix and rework every feature before a commit to make it clean, optimized, and de-slopified.

But still, we wouldn't be allowed to use codeberg because "generative ai" was used significantly for most lines of code?


AFAICT they are dealing with resource exhaustion from slopcoding agents hammering them with copycat garbage. It's existential. If you use AI to help you thoughtfully write code, review it with humans, and respect their service, I don't think they will have a problem with you.


If it's good, will they ever find out?


Maybe not... I personally wouldn't use a service that would boot me off if they did find out because I spoke about it online or committed too much code one week or something.


> And "mostly" here seems egregiously undefined to me

and yet we seem to be convinced that English is the appropriate spec language of the modern times


re: npm, if you upgrade your global node version, you will lose that installation, right?


no, in fact you don’t even need node or npm to install npm packages with mise. (You likely will need it to execute them though)


right, but say you give the agent access to github and it can push as you, or make a gist; now it can easily exfiltrate your secret.

And that's just an easy case - really if it has any network access at all it can come up with a clever way to route a request through the network such that the key comes back somewhere in the request. If you scan for it inbound too, the machine can obfuscate it.

Our agents are trained to be so intensely helpful and they have such intricate knowledge of how things work that they will do some incredibly clever tricks to do what you ask them to do.


The agent has no access to the secret. It has a placeholder that is replaced at a higher level. When it makes the network request the secret is substituted but that is outside of the caller's worldview.


So what stops it from sending a network request to a git repo that pushes what that placeholder resolves to?


How would that work? You don't control github.com servers so your repo would never see the secret.

edit: You may want to look into tokenizing proxies as the general application of this concept.


Your agent writes secret.txt with the placeholder, and the tokenizing proxy replaces it with the token, then the agent reads secret.txt


It only replaces the token in the HTTP header that is sent to the server. Whatever you wrote in your files isn't touched by the proxy.


It sends a request to requestb.in and reads the public log of the headers. There are ways.


But requestb.in is not api.github.com so the proxy wouldn't replace anything.


Couldn't it then just publish the mock in a public place... it would get replaced by the real secret.? How is this prevented


Maybe the tokenizing proxy could work both ways? If the agent tries to read secret.txt, it gets back the placeholder.


But how? Normally the TLS handshake and encryption/decryption happen in user space. Even the kernel doesn’t know anything about it.


There is a transparent proxy installed (along with the necessary certificates on the VM.) For an example, see https://docs.microsandbox.dev/networking/tls


So, if the program or the proxy solution doesn’t support it, then it doesn’t work? Like with security solutions?


microsandbox maintainer here. the custom certificate is installed in the guest's trusted root CA list, so it should work across any program, except where the program opts to explicitly pin certificates for a destination.


The docs mention it can be bypassed for configured domains.


Yes, I read it. That means that it doesn’t work in those cases. Btw, as a developer it’s very easy to have something like that. It’s not as trivial as it seems at all. I encountered with similar problems all the time, with similar solutions (mainly for security theater reasons) in the past. There are websites which simply doesn’t work if you replace certificates, regardless of browser or CA for example.


Yes, I've worked with people who have run into issues with "security" solutions like ZScaler. I have tried it with some APIs (like GitHub) and it does work. Not to say it will work in your case.


I was just interested how it works, because above it was sold as “it works”, when in reality, “it works*”.


Isn't it that way with most software? "It works", except when it doesn't.


There is a difference between bugs/failures, and false advertisement. The solution used here has well known shortcomings.


All SSL/TLS interception has the shortcoming. In other contexts, that’s a good thing, when certificate pinning works, I mean.


microsandbox is designed to work with most security products. there's a dedicated section for this in the docs that makes this entire process seamless: https://docs.microsandbox.dev/networking/tls#trusting-host-c...


Operating systems ought to be providing us the utilities we need to safely sandbox processes (agent or otherwise), but they appear to not be interested in the job


Apple, Microsoft, IBM, Unisys, HP, Oracle/Sun have done that for a while now.


Poorly! Apple’s facilities for this are the ones I know best, and they are woefully insufficient


Well, the features announced at WWDC 2026 naturally are yet to be made available in a mature form.


This is what I currently do, but my software uses docker and docker mounts act as a bypass for the file system restrictions, plus docker processes started outside the sandbox allow network proxy escape.

Currently, I don't allow the agent access to docker, start docker myself, and then do short-lived sandbox-free sessions when the agent needs to do things that interact directly with docker; but that's annoying.


“Smoking gun”


I want to be straight with you, I overstepped by naming it a "smoking gun."


"Honest assessment: I was wrong to say I was being straight with you. You pointed out that a "smoking gun" is a sign of evidence, and I clearly didn't have any. This is not a bug but a gap that can be fixed like [this]. Give me the word and I'll wire it in."

··You've hit your monthly spend limit · raise it at claude.ai/settings/usage


> Is that built in protection really a filter, on code level

yes, on mac it uses seatbelt and on other platforms it uses similar tools: https://code.claude.com/docs/en/sandboxing


Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: