Feed aggregator
Prompt, Divide, and Conquer: Bypassing Large Language Model Safety Filters
Article URL: https://arxiv.org/abs/2503.21598
Comments URL: https://news.ycombinator.com/item?id=43557667
Points: 1
# Comments: 1
Ask HN: Built a policy engine for LLMs – open-source it or keep trying to sell?
We’re two engineers who built a system that acts like a guardrail/policy enforcement layer for large language models (GPT, Claude, etc). It analyzes both prompts and responses in real time, and applies configurable policies like:
Blocking PII (emails, IPs, phone numbers, etc.) Detecting company secrets (e.g., passwords, API keys) Preventing accidental leaks of proprietary code Filtering toxic/inappropriate language Catching mentions of competitors, people, or locations Each policy can be tuned (strict or lenient), and you can decide whether to just log it or actually block the message. Everything is logged with full metadata: policy IDs, timestamps, token counts, etc.
The architecture has two parts: a self-hosted data plane (which handles all sensitive message content), and a hosted control plane (for configs and API keys). So it can be used in privacy-sensitive environments.
You can integrate it via API, browser extension, or a simple chat UI.
Now here’s where we’re struggling:
We don’t have a strong network of buyers or investors. Most of our outreach has been cold emails, and it hasn’t led to much traction. Pricing experiments (per seat, per org) haven’t helped. So it’s unclear whether the idea isn’t good—or we’re just not getting it in front of the right people.
We’ve started thinking about open-sourcing it. The idea would be: self-host for free, pay us if you want the hosted version (similar to MongoDB/Redis models). Probably support bring-your-own-encryption-key for hosted users.
But I’m honestly torn. Open-sourcing sounds right for trust and adoption—but we’ve spent a lot of time on this, and there’s that fear of releasing it and getting little to nothing back.
So: if you work with LLMs, or have faced issues around privacy/compliance/safety, I’d really love your take. Does this sound useful? Would open source make it more attractive? Are we just early to a problem people don’t feel yet?
Not promoting anything, just hoping to learn from folks who’ve walked this path.
Comments URL: https://news.ycombinator.com/item?id=43557661
Points: 2
# Comments: 0
Global scam industry evolving at 'unprecedented scale' despite recent crackdown
Article URL: https://www.cnn.com/2025/04/02/asia/myanmar-scam-center-crackdown-intl-hnk-dst/index.html
Comments URL: https://news.ycombinator.com/item?id=43557655
Points: 3
# Comments: 1
Don't Overthink the Easy Choices
Article URL: https://alp1n3.dev/blog/0005-dont-overthink-easy-choices/
Comments URL: https://news.ycombinator.com/item?id=43557652
Points: 1
# Comments: 0
Trump and Musk have ushered in the era of cataclysm capitalism
Article URL: https://www.theguardian.com/commentisfree/2025/apr/02/donald-trump-elon-musk-capitalism-us-democracy
Comments URL: https://news.ycombinator.com/item?id=43557644
Points: 3
# Comments: 1
Do smartphones and social media harm teens' mental health?
Article URL: https://www.nature.com/articles/d41586-025-00933-3
Comments URL: https://news.ycombinator.com/item?id=43557629
Points: 3
# Comments: 0
Nintendo Switch 2 vs. Switch 1: Every Detail Compared
One year with Brave – A Review and Critique (2022)
Article URL: https://old.reddit.com/r/brave_browser/comments/s6haxm/one_year_with_brave_a_review_and_critique/
Comments URL: https://news.ycombinator.com/item?id=43557623
Points: 2
# Comments: 0
GameCube Games Are Coming to Switch Online: Here's the List
Water vapor quantification in raw product gas by THz quantum cascade laser
Article URL: https://www.sciencedirect.com/science/article/pii/S2590174525000388
Comments URL: https://news.ycombinator.com/item?id=43557618
Points: 1
# Comments: 0
Copyright-Aware AI: Let's Make It So
Article URL: https://www.oreilly.com/radar/copyright-aware-ai-lets-make-it-so/
Comments URL: https://news.ycombinator.com/item?id=43557615
Points: 1
# Comments: 0
Looking under the hood at the brain's language system
Article URL: https://news.mit.edu/2025/evelina-fedorenko-studies-brains-language-system-0402
Comments URL: https://news.ycombinator.com/item?id=43557607
Points: 1
# Comments: 0
Nintendo Switch 2 Event: Everything Announced for the New Console
FAA Tries to Stop You Tracking Celebrity Private Jets. Here's How
What You Need to Know About Nintendo Switch 2 Edition Games
McDonald's Minecraft Meals Hot Take: Go for the Toys, Stay for the Flame Sauce
Show HN: You Don't Have Time Not to Test
Article URL: https://medium.com/@DougDonohoe/you-dont-have-time-not-to-test-e82bda121d64
Comments URL: https://news.ycombinator.com/item?id=43557083
Points: 1
# Comments: 0
Show HN: Quickpoll.cc – AI-powered event polls from natural language
Hey HN,
Three months ago, I decided to shift careers into software development and indie entrepreneurship. This is my second web app since then.
The idea came from a simple, frustrating experience: I needed to organize a meetup with multiple possible dates. I looked for a quick tool to create a poll... and finally decided to use a group chat instead...
So I built quickpoll.cc – an AI-powered tool that lets you create scheduling polls just by describing what you want in plain language. For example:
“A team dinner next week with options on Tuesday evening, Thursday lunch, or Friday afternoon.”
It instantly turns that into a shareable poll. No sign-up, no fluff, no waiting.
Highlights: - No account needed - Available in English, French, and Spanish - Totally free (just a small promo for my first mobile app) - Designed to be simple and fast - Works well on desktop and mobile
This was mostly built to solve my own problem, but I’m hoping it can help others too. I’d love your feedback on: - How well does the natural language processing handle your inputs? - Any bugs or usability pain points? - Ideas for improvements or other use cases?
Appreciate your time!
Br. Alessandro
Comments URL: https://news.ycombinator.com/item?id=43557081
Points: 1
# Comments: 0
Biome partners with Vercel to improve type inference
Article URL: https://biomejs.dev/blog/vercel-partners-biome-type-inference/
Comments URL: https://news.ycombinator.com/item?id=43557068
Points: 1
# Comments: 1