HNHacker News
TopNewBestAskShowJobs

prophesi

3,333 karma · joined January 29, 2016

To contact me, look up this fingerprint on the keys.openpgp.org keyserver for a public key with my contact info: 34E9710869817515F26385DB113DFB380F724A81
submissionscomments
prophesi··on 95% of Companies See 'Zero Return' on $30B Generative AI Spend
You would be hard-pressed to find a call center that _doesn't_ start every phone call with a warning that the conversation may be recorded.
prophesi··on Tidewave Web: in-browser coding agent for Rails and Phoenix
Is it safe to say that the Tidewave MCP server will remain FOSS, but anything Tidewave Web related (localhost:4000/tidewave) will be proprietary yet optional?
prophesi··on How well do coding agents use your library?
In Elixir land, the Ash Framework created a package called usage_rules[0] as an experimental attempt to solve this problem a few months ago. The latest version of the Phoenix Framework (1.8) includes it in their `mix phx.new` generator and in their own hex packages[1]. Library owners would need to add their own usage rules, but it seems to help even for just the core packages Phoenix includes.

[0] https://hexdocs.pm/usage_rules/readme.html

[1] https://github.com/phoenixframework/phoenix/tree/main/usage-...

prophesi··on Can modern LLMs count the number of b's in "blueberry"?
I think it's a valid complaint that the naming scheme for the various GPT-4 models were very confusing. GPT-5 just launched, and doesn't (yet?) have a GPT-5 Turbo or GPT-o5 mini to muddy the waters.
prophesi··on Trump Orders National Guard to Washington and Takeover of Capital’s Police
Yeah it's been hard for it to leave the news cycle when there are new developments still happening all the time. Latest one being a federal judge denying the request to unseal the transcripts for cases relating to Ghislaine Maxwell. And at least going by Ground News, seems to have been widely covered by outlets today.

https://ground.news/article/judge-rejects-unsealing-ghislain...

prophesi··on Sunlight-activated material turns PFAS in water into harmless fluoride
It's a bit disheartening to see that the Bill Moyer's documentary came out in 2001, and not much has changed to keep these corporations accountable.
prophesi··on Open SWE: An open-source asynchronous coding agent
Hm I don't think so. You might be thinking about the file size, which is ~64GB.

> Native MXFP4 quantization: The models are trained with native MXFP4 precision for the MoE layer, making gpt-oss-120b run on a single 80GB GPU (like NVIDIA H100 or AMD MI300X) and the gpt-oss-20b model run within 16GB of memory.

If you _could_ fit it within ~60GB VRAM, the variability of the amount of VRAM required for certain context lengths and prompt sizes would OOM pretty quickly.

edit: Ah and MXFP4 in itself is a quantization, just supposedly closer to the original FP16 than the rest with a smaller VRAM requirement.

prophesi··on Study finds flavor bans cut youth vaping but slow decline in cigarette smoking
Once nicotine salts were introduced, it hit the fan. It was no longer about the clouds, which freebase and their glycerine introduced. But just more and more nicotine content.
prophesi··on Study finds flavor bans cut youth vaping but slow decline in cigarette smoking
If you started in 2011, then you likely had freebase nicotine that maxed out at 20mg/ml before being too harsh for people to vape. It was around 4 or 5 years later that nicotine salts began being used due to how smooth the hit can be, and those can reach +70mg/ml now. There's not really a motive to keep low nicotine disposable vapes in gas stations and the like.

edit: And yeah, this is all anecdotal for me. No clue how much nicotine you actually intake via these methods.

prophesi··on Study finds flavor bans cut youth vaping but slow decline in cigarette smoking
I will also add to the discussion that disposable vapes (in the USA) tend to be outrageously high in their nicotine content. If you buy the juice yourself you have options to choose much lower nicotine per volume compared to these disposables (and will be a lot cheaper in the mid-term/long-term). A lot of devices these days curb the issue of replacing your own coils too, as that used to be a common issue.

But please don't start vaping if you don't already smoke. There's also the discussion of tobacco-less nicotine pouches that have hit the scene; on average they have higher nicotine content than regular tobacco pouches. And the same issue applies, as it's also acceptable to partake of them anywhere at anytime.

prophesi··on Open SWE: An open-source asynchronous coding agent
I'm also excited for local LLM's to be capable of assisting with nontrivial coding tasks, but we're far from reaching that point. VRAM remains a huge bottleneck for even a top-of-the-line gaming PC to run them. The best these days for agentic coding that get close to the vibe-check of frontier models seem to be Qwen3-Coder-480B-A35B-Instruct, DeepSeek-Coder-V2-236B, GLM 4.5, and GPT-OSS-120B. The latter being the only one capable of fitting on a 64 to 96GB VRAM machine with quantization.

Of course, the line will always be pushed back as frontier models incrementally improve, but the quality is night and day between these open models consumers can feasibly run versus even the cheaper frontier models.

That said, I too have no interest in this if local models aren't supported and hope that's down the pipeline just so I can try tinkering with it. Though it looks like it utilizes multiple models for various tasks (planner, programmer, reviewer, router, and summarizer) so that only adds to the difficulty of the VRAM bottleneck if you'd like to load different models per task. So I think it makes sense for them to focus on just Claude for now to prove the concept.

edit: I personally use Qwen3 Coder 30B 4bit for both autocomplete and talking to an agent, and switch to a frontier model for the agent when Qwen3 starts running in circles.

prophesi··on Many countries that said no to ChatControl in 2024 are now undecided
Wasn't TOR originally funded by the CIA as well?
prophesi··on Ollama's new app
Oof well there goes my desire to switch back.
prophesi··on Many countries that said no to ChatControl in 2024 are now undecided
The privacy-preserving systems have been built and are finally making it into the mainstream, only to be uprooted by politics. Not that we no longer need to build more of such systems, but the root of the issue lies elsewhere.
prophesi··on Many countries that said no to ChatControl in 2024 are now undecided
I don't think there are actionable items in that manifesto, and was written during the RSA munitions debate.
prophesi··on Nova: A New Web Framework for Erlang
Not sure how this hit the front page without linking the actual project.

https://novaframework.org/

prophesi··on Ollama's new app
I think having a bash script as the linux installation is more of a stop-gap measure than truly supporting Linux. And ollama is FOSS compared to LM Studio and Msty (as someone who switched from ollama to LM Studio; I'm very happy to see the frontend development of ollama and an easier method of increasing the context length of a model).
prophesi··on Women dating safety app 'Tea' breached, users' IDs posted to 4chan
I wouldn't go that far. What they uncover with their FOIA requests that the general public would otherwise never know about tends to be quality content. And, like the Wired, their FOIA-based articles aren't paywalled.
prophesi··on Google's shortened goo.gl links will stop working next month
Because there isn't a line? You can require auth for any of those HTTP methods. Or not require auth for any of them.
prophesi··on Google's shortened goo.gl links will stop working next month
Obfuscation may hint that it's intended to be private, but it's certainly not authentication. And the keyspace for these goog.le short URL's are much smaller than a 64byte alphanumeric code.
prophesi··on Graphene OS: a security-enhanced Android build
The issue isn't with NFC. It's passing the Play Integrity check that app developers optionally can use to prevent devices that don't pass the check from running their app, or remove parts of its functionality. IIRC I don't think any custom ROM's can pass the check. So you might be able to pay via NFC with a banking app if they don't implement the Play Integrity API. For Graphene's thoughts on the matter (2024):

https://grapheneos.org/articles/attestation-compatibility-gu...

prophesi··on How to Firefox
You can't have nested tab groups so I imagine you'd want to keep using that extension if that's what you're using it for.
prophesi··on How to Firefox
A tip I would add to this article is that Firefox natively supports sidebar tabs now without needing hacky extensions. Go to about:preferences under the Browser Layout section of the General tab, and select Vertical Tabs. The tab group functionality along with Multi-Account Containers are a lot more useful under this layout IMO.
prophesi··on Exposed MCP servers across the internet
I think that attack surface would be the LLM's utilizing the MCP server, not the MCP server itself. It took a while to wrap my head around LLM vs Agents vs MCP servers, but the latter is just code with endpoints to list and call their tools.
prophesi··on Trying to send a sticker in Steam Chat burned through a month of mobile data
Steam Deck uses a desktop OS (Arch), and Steam's Big Picture mode is well-optimized for controllers.
prophesi··on Exposed MCP servers across the internet
The two things I can think of are MCP servers with functions that make calls to a database with sensitive information, or are easy to pwn due to propping them up in a hasty and irresponsible manner.

The article would actually be interesting if they tried either of those with the servers they found.

prophesi··on Fully homomorphic encryption and the dawn of a private internet
The operations used to perform these computations are utilizing algorithms backed by lattice-problems that are NP-hard to break.

edit: The only leaking information in this case are what operations you'd like to perform on the encrypted data. e.g., you now know that you've incremented the password by x amount, but you don't know what the plaintext was before, the plaintext after, _or_ infer what the value is by knowing you've modified the encrypted data by x amount.

edit2: Now I think I understand the question. The encryption is technically weakened because you can know what operations are made on the underlying data. Though it's still an advancing field, and there are promising developments with what's called circuit privacy[0] to prevent the server knowing the operations made as well.

[0] https://eprint.iacr.org/2022/1459

prophesi··on Fully homomorphic encryption and the dawn of a private internet
The tl;dr is that breaking FHE would mean solving lattice problems that have been studied for decades to be nontrivial to break[0].

[0] https://arxiv.org/abs/2208.08125

prophesi··on How I Use Kagi
These days it's often the sites in the results themselves that are either ads or highly SEO-optimized low value sites. You can manage a blacklist and a list of sites to promote/demote in the search results. And manage different "lenses" to have different buckets of blacklist/promote/demote settings tailored for what you're researching. And can also be used for their Kagi Assistant when you allow it to perform web searches.

edit: This is detailed in the article, but leaving this here for those like me who first jump to the HN comments

prophesi··on Oakland cops gave ICE license plate data; SFPD also illegally shared with feds
I think it's because you don't have to look too hard to find examples of authoritarian regimes leveraging information technologies for surveillance, censorship, and propaganda. Or how US government agencies use loopholes to get around the 4th amendment and buy sensitive civilian data from private data brokers. Or how data breaches are becoming larger and more frequent each year.
← PreviousPage 6 of 34Next →