Hacker News Reader

lagrange77on Launch HN: VibeFlow (YC S25) – Web app generator with visual, editable workflows

That's the most 2025 startup name and idea i've come across so far.

lagrange77on Show HN: An ncurses CUDA-based fluid simulation

Cool!

Jos Stem style?

lagrange77on Some thoughts on LLMs and software development

Right, all models are inherently wrong. It's up to the user know about its limits / uncertainty.

But i think this 'being wrong' is kind of confusing when talking about LLMs (in contrast to systems/scientific modelling). In what they model (language), the current LLMs are really good and acurate, except for example the occasional chinese character in the middle of a sentence.

But what we mean by LLMs 'being wrong' most of the time is being factually wrong in answering a question, that is expressed as language. That's a layer on top of what the model is designed to model.

EDITS:

So saying 'the model is wrong' when it's factually wrong above the language level isn't fair.

I guess this is essentially the same thought as 'all they do is hallucinate'.

lagrange77on Some thoughts on LLMs and software development

I'll steal that.

lagrange77on Malleable Software

This is comparing two orthogonal properties.

SaaS is a business model while malleable vs. rigid is a property of the software itself.

lagrange77on A guide to Gen AI / LLM vibecoding for expert programmers

Absolutely agree. The watches do tell time better. But the factory worker does not become better at the craft of watchmaking or EE.

lagrange77on A guide to Gen AI / LLM vibecoding for expert programmers

Yeah i agree with most of what you say.

> looking like really good product managers.

Exactly and that's a different field with a different skillset than developer/programmer.

And that's the purpose of technology in the first place tbh, to make the hard/tedious work easier.

lagrange77on A guide to Gen AI / LLM vibecoding for expert programmers

You have forgotten the most important part: Lay off 90% of those devs.

lagrange77on A guide to Gen AI / LLM vibecoding for expert programmers

But automating isn't a programming paradigm.

> They are young and inexperienced today, but won't stay that way for long.

I doubt that. For me this is the real dilemma with a generation of LLM-native developers. Does a worker in a fully automated watch factory become better at the craft of watchmaking with time?

lagrange77on Show HN: OpenAI/reflect – Physical AI Assistant that illuminates your life

Is it my browser, or does the video in the readme not have sound?

lagrange77on Dokploy is the sweet spot between PaaS and EC2

Does any of you use one of these (Dokploy, CapRover, Dokku, Coolify) like Netlify, as advertised by some?

For me, the core feature of Netlify is building and deploying static websites quickly, with minimal configuration and triggered by git commits.

Does any of these really resemble that experience (except for the CDN Netlify uses, of course)?

lagrange77on Dokploy is the sweet spot between PaaS and EC2

https://docs.dokploy.com/docs/core/comparison

Dokploy vs. CapRover, Dokku, Coolify

lagrange77on Jim Lovell, Apollo 13 commander, has died

> I became the math consultant for A Beautiful Mind in part because I was such an Apollo 13 buff.

Cool, awesome job, as far as i can tell as a fan of the movie!

So you did what was best for yourself... and the group.

lagrange77on Running GPT-OSS-120B at 500 tokens per second on Nvidia GPUs

Thanks for your answers!

While it is seemingly hard to calculate it, maybe one should just make a database website that tracks specific setups (model, exact variant / quantisation, runner, hardware) where users can report, which combination they got running (or not) along with metrics like tokens/s.

Visitors could then specify their runner and hardware and filter for a list of models that would run on that.

lagrange77on Running GPT-OSS-120B at 500 tokens per second on Nvidia GPUs

While you're here..

Do you guys know a website that clearly shows which OS LLM models run on / fit into a specific GPU(setup)?

The best heuristic i could find for the necessary VRAM is Number of Parameters × (Precision / 8) × 1.2 from here [0].

[0] https://medium.com/@lmpo/a-guide-to-estimating-vram-for-llms...

lagrange77on Attention is your scarcest resource (2020)

Oh i do.

lagrange77on Attention is your scarcest resource (2020)

That's why i despise advertising and marketing so much.

It forcefully tries to hack you and steal your scarcest resource.

lagrange77on Research suggests Big Bang may have taken place inside a black hole

Right, i had the same thought. What we can say is, that the quantities we call t and S are correlated.

lagrange77on Research suggests Big Bang may have taken place inside a black hole

If universes can exist in black holes and obviously black holes exist in universes, doesn't that smell like infinite recursion?

lagrange77on Research suggests Big Bang may have taken place inside a black hole

Cool anecdote!

lagrange77on Research suggests Big Bang may have taken place inside a black hole

Of course these are models. And to be able to recognise that entropy increases with time is an example for their usefulness.

lagrange77on "AI Will Replace All the Jobs " Is Just Tech Execs Doing Marketing

If humanity survives long enough, it will eventually. It's the end goal of technology.

lagrange77on Pope Francis has died

I can imagine that for people of faith there is a lot to be read into the time of death.

lagrange77on Show HN: We made an MCP Server so that Cursor can build anything from API Docs

Actually, programming languages are also for humans.

lagrange77on Moving away from US cloud services

That would be HN, i guess.

lagrange77on Fastplotlib: GPU-accelerated, fast, and interactive plotting library

Thank you!

lagrange77on Fastplotlib: GPU-accelerated, fast, and interactive plotting library

Nice, i'd be interested to know which method for drawing lines (which is hard [0]) it uses.

[0] https://mattdesl.svbtle.com/drawing-lines-is-hard

lagrange77on Solving First Order Differential Equations with Julia

I second that.

Also recommend this course by Gilbert Strang @ MIT:

https://ocw.mit.edu/courses/res-18-009-learn-differential-eq...

https://www.youtube.com/playlist?list=PLUl4u3cNGP63oTpyxCMLK...

lagrange77on Show HN: While the world builds AI Agents, I'm just building calculators

For me it's also the joy of working with a deterministic system.

lagrange77on I think Yann Lecun was right about LLMs (but perhaps only by accident)

> But with Attention mechanism

I would think LeCun was aware of that. Also prior sequence to sequence models like RNNs have already incorporated information about the further past.