HN user

bertman

5,314 karma

What are we even doing.

Posts167
Comments327
View on HN
ludic.mataroa.blog 4d ago

AI Mania Is Eviscerating Global Decision-Making

bertman
75pts13
www.heise.de 1mo ago

Gödel and Turing outlined the limits of AI

bertman
2pts0
restofworld.org 6mo ago

Silicon, not oil: Why the U.S. needs the Gulf for AI

bertman
1pts0
mathstodon.xyz 7mo ago

Terence Tao: Cleverness versus Intelligence in AI Tools and Humans

bertman
4pts0
mathstodon.xyz 7mo ago

Tao: Intelligence is too high-dimensional for accurate low-dimensional narrative

bertman
3pts1
github.com 8mo ago

Yt-dlp: External JavaScript runtime now required for full YouTube support

bertman
1106pts627
www.economist.com 9mo ago

China's chipmakers are cleverly innovating around America's limits

bertman
2pts1
infrequently.org 9mo ago

The App Store was always authoritarian

bertman
150pts103
www.theregister.com 9mo ago

Former UK prime minster Sunak becomes human Clippy for Microsoft, Anthropic

bertman
6pts0
restofworld.org 9mo ago

Indonesia's film industry embraces AI to make Hollywood-style movies for cheap

bertman
28pts5
mathstodon.xyz 10mo ago

Terence Tao: The role of small organizations in society has shrunk significantly

bertman
1060pts565
lemire.me 10mo ago

Splitting a long string in lines efficiently

bertman
3pts0
ploum.net 10mo ago

I Fell in Love with Calendar.txt

bertman
7pts1
rachelbythebay.com 10mo ago

Making the most of a dumb fax switcher box in the old days

bertman
79pts21
www.electrospaces.net 10mo ago

Preventing telephone manipulation: the TSG standards

bertman
1pts0
restofworld.org 11mo ago

Dubai's robotaxi dreams are coming for human drivers

bertman
2pts0
thenewstack.io 11mo ago

Vercel Goes All in on Vibe Coding Web Apps

bertman
1pts1
mathstodon.xyz 12mo ago

Terence Tao: Applying Red Team / Blue Team Duality to AI Workflows

bertman
2pts0
mathstodon.xyz 12mo ago

Quantitative AI progress needs accurate and transparent evaluation

bertman
209pts104
www.notesfromthecircus.com 12mo ago

Bitcoin Coup:How Crypto Accelerationists Engineered America's Financial Collapse

bertman
5pts1
mathstodon.xyz 1y ago

A human metaphor for evaluating AI capability

bertman
155pts31
blog.jonudell.net 1y ago

Xmlui

bertman
2pts4
mathstodon.xyz 1y ago

Terence Tao: Game theory, politics and control of information

bertman
59pts13
purplesyringa.moe 1y ago

"AI discourse" is a joke

bertman
14pts25
www.notesfromthecircus.com 1y ago

The Magical Thinking That's Killing Our Humanity

bertman
7pts5
www.noemamag.com 1y ago

To Save Nature, Make It Sacred

bertman
4pts0
www.theverge.com 1y ago

AI Infiltrated Perfume

bertman
2pts0
lcamtuf.substack.com 1y ago

0.9999 ≊ 1

bertman
10pts2
mathstodon.xyz 1y ago

Terence Tao on the importance of funding basic mathematical research

bertman
5pts1
bernsteinbear.com 1y ago

A simple search engine from scratch

bertman
296pts59

Your description is not accurate and quite misleading.

QuickVPN runs entirely in your browser

Like I said: What do you mean with this? You did not build a "VPN" that runs in the browser. Your extension sets up a web proxy.

And worse:

or log user data.

You're implying you're running some kind of "no log" service which is false.

Interesting quotes:

“We try to disclose complexity only when necessary,” he said. “So if you don’t ask for a particular tech stack or anything, we will build you a web app. It’ll be using Next.js, it’ll be written in TypeScript, and it’ll be styled with Tailwind.” These are the “most modern technologies” that are available for web development, he added.

So this makes sense I guess:

The Next.js team is updating the framework so that it plays better with LLMs and agentic AI, Khandelwal said.

At first, I was confused because I didn't get why they would call it "hand-picked" when literally every single "article" on that site is attributed to the site itself.

And then I clicked a random page and, well, it's slop:

By amplifying the signal from minority class data and leveraging the diversity of models, these methods enhance prediction accuracy and fairness across all classes. When paired with complementary techniques such as resampling, adjusting class weights, or generating synthetic data, ensemble methods can yield even more robust results in handling imbalanced datasets.

Yeah, but if you're trying to promote your AI generated React bloat wrapper in a "Show HN", you should at least be honest and describe how it's actually working instead of writing mindless, non-coherent, intentionally misleading ad blurbs without any actual information imho.

Using CompressorJS means lossy image compression. You really shouldn't be calling this "Instant, Lossless File Compression" on your page.

Also, because you're using CompressorJS's default settings, you should know that PNG files > 5 MB are converted to JPG (https://github.com/fengyuanchen/compressorjs?tab=readme-ov-f...). Your site, however, keeps the .png file name ending for the converted image, again falsely suggesting lossless compression.

Synopsis from the project's "instructor guide:

This is not a computer science course, nor even an information science course—though naturally it could be used in such programs.

Our aim is not to teach students the mechanics of how large language models work, nor even the best ways of using them in various technical capacities.

We view this as a course in the humanities, because it is a course about what it means to be human in a world where LLMs are becoming ubiquitous, and it is a course about how to live and thrive in such a world.