HNHacker News
TopNewBestAskShowJobs

pietz

391 karma · joined December 2, 2016

submissionscomments
pietz··on Show HN: Typed-FFmpeg 3.0–Typed Interface to FFmpeg and Visual Filter Editor
That looks awesome.
pietz··on Plex remote stream no longer free; Plex Pass 2x price hike
I hate subscriptions a lot but Infuse is literally $1 per month.
pietz··on Plex remote stream no longer free; Plex Pass 2x price hike
Plex is a service I want to like, but I've had so many technical issues over the years, I gave up at some point. I then stopped being the Plex server provider to all of my friends, bought an Apple TV, got Infuse Pro and never looked back for even a split second.
pietz··on Searchcode.com’s SQLite database is probably 6 terabytes bigger than yours
I host my stuff on railway and while I love SQLite, I usually go with their Postgres offering instead. It's actually less work for me and although I haven't tested this scientifically, it's hard too imagine the SQlite way would save me serious money.
pietz··on Fair Pricing
It's so crazy to me to hear these super positive opinions. I gave kagi a shot for several months but the results were quite a bit worse than Google or DuckDuckGo. Maybe it's because I live in Germany and kagi doesn't do well with German content but I never understood the hype of kagi.
pietz··on Hetzner cuts traffic on US VPSs
A lot less bandwidth and higher prices with less than a week until it goes live. That's insane.
pietz··on DuckDB over Pandas/Polars
I don't understand the purpose of this post. "I write a lot of X so I prefer using X over Y." Great.
pietz··on Claude 3.5 Haiku Released
The pricing isn't absurd, but I also think this one will stay on the shelf. Their Claude 3.0 lineup was so good:

Haiku as a very cheap and fast option for low complexity stuff. Sonnet for most people and their fairly advanced problems. Opus as a very expensive, slow but very capable model. Haiku 3 was 12x cheaper and Opus 5x more expensive than Sonnet. This made a lot of sense to me and I used every single for for different tasks.

With 3.5, that's a different story. They're clearly running into GPU limits and dropping Opus for now is a reasonable thing to do. Sonnet has improved so much and OpenAI is behind them anyway. But this move on Haiku is stupid on so many levels.

Their lineup is down to two models, which are closer than ever in performance and price. It's like "Sonnet" and "Almost Sonnet" and at a 3x price difference, I'll stick with the OG. For a lower cost model, I'm going to stick with GPT-4o-mini.

Also, people make connections to products through names and not benchmark results. It's great that Haiku has improved so much, but it's still Haiku. We know models get better, but increasing prices 4x for a new version of an existing one doesn't feel nice.

pietz··on Show HN: Replacicon – Create and customize app icons on macOS
Can I also use this to create icons for my own apps?
pietz··on Play 3.0 mini – A lightweight, reliable, cost-efficient Multilingual TTS model
Is your model really open source or did you misunderstand the question?
pietz··on OpenAI Threatening to Ban Users for Asking Strawberry About Its Reasoning
How will this be controlled on Azure? Don't they have a stricter policy on what they view and also develop their own content filters?
pietz··on MiniJinja: Learnings from Building a Template Engine in Rust
I think JinjaX was inspired by JSX. Obviously, there's a huge delta, but my "HTML heavy" problems are nicely solved with JinjaX
pietz··on MiniJinja: Learnings from Building a Template Engine in Rust
Yes
pietz··on MiniJinja: Learnings from Building a Template Engine in Rust
Different tastes, I guess. It's the same syntax that many JavaScript Frontend frameworks like react and Svelte use. I find it much cleaner.

When using Jinja with HTMX you also run into trouble that it doesn't nicely support partials. You could use the jinja_partials library but I just find JinjaX so much more pleasant.

pietz··on MiniJinja: Learnings from Building a Template Engine in Rust
https://jinjax.scaletti.dev/ cured my pain.
pietz··on MiniJinja: Learnings from Building a Template Engine in Rust
I really don't like Jinja. The syntax is not for me and there's often three different ways of building the same thing. I find it a bit confusing.

Recently I started using JinjaX (https://jinjax.scaletti.dev/), which is weirdly amazing. I think the comparison on the homepage explains it quite well. It feels like what Jinja always should have been and and the multiple ways of doing the same thing converge into one syntax in JinjaX.

I really hope the project gains more traction.

pietz··on Anthropic Claude 3.5 can create icalendar files, so I did this
Really cool! It's also really good at creating all kinds of mermaid diagrams. I have a prompt to create learning resources based on YouTube transcripts including visualizations in mermaid. It's awesome.
pietz··on Uv 0.3 – Unified Python packaging
What does the Astral team recommend for setting up their tools on a new machine? Install uv with curl, manage python versions and projects from there and install ruff with uv?
pietz··on Uv 0.3 – Unified Python packaging
What's the recommended/opinionated
pietz··on Sqlite-vec: Work-in-progress vector search SQLite extension that runs anywhere
Is this also what turso uses in their "AI feature"?
pietz··on FastHTML – Modern web applications in pure Python
This looks cool and I will check it out but I'm also quite happy with my tech stack.

I started to couple my FastAPI backend with native Jinja2 templates and noticed that I hate Jinja2 with passion (no disrespect). I tried HTPY which seemed great but this Python abstraction of HTML just felt weird and I found myself converting HTML to HTPY all the time. I even created a GPT for it. Then I found JinjaX and noticed that this hits the nail on the head for me. It's a Jinja2 preprocessor that allows the usage of components instead of the weird extends and macro syntax.

I'm happy to look at FastHTML but I'm not sure what type of benefit I can expect.

pietz··on The economics of a Postgres free tier
The SQLite offerings are pretty good for hobby projects. The free tier ok turso goes a long way and the next level costs you $8
pietz··on Show HN: Let Grumpy AI Roast Your Startup Idea
Sure, because posting my detailed startup concept on some unknown website sounds like a good idea. Thanks.
pietz··on Claude-3.5-sonnet 'artifacts' system prompt is a doozy
That just killed v0.dev
pietz··on Hypermedia Systems
HTMX is not perfect and there are many situations when you probably don't want to use it, but I think it should be considered the default answer to the question: What framework should I learn to build a dynamic frontend?

At the moment the answer to that question is "react" and while react can be cool (I've heard) it has developed into a complex world that requires quite some time to master. Even intuitive JS frameworks like Svelte have been getting a lot bigger in order to be more feature complete.

HTMX is the Pareto answer to building dynamic frontends. It will be great for 80% of all projects that require interactivity while doing that job in 20% of the complexity that react would bring.

pietz··on xLSTM code release by NX-AI
Could someone provide a quick summary where they stand compared to transformer architectures? Do they have real world scale results that are competitive?
pietz··on GPT-4o is consistently hallucinating
I haven't noticed that GPT-4o hallucinates a lot more than the previous version but I noticed 2 other things of which especially the latter seems relevant here.

1) it's insanely chatty, to a point where it ignores instructions about not doing certain things. I think this behavior is heavily favoured by benchmarks but as somehow who expects concise answers, this model annoys me. Custom instructions don't fully fix this for me.

2) It likes repetitive answers a lot more than the previous version. Meaning that it will try its hardest to generate the followup answer in the same format as the first one. I think this is also the problem in your example.

To my understanding, this is a measure against laziness, where the model would exclude information from the first answer that haven't changed in the followup. I always liked this behavior but maybe you remember the time from a few months ago where many people complained about the laziness of (I believe) 0125.

Btw, while I type this, I notice that this is probably the highest level of first world problems I've ever complained about. There is this amazing almost free tool that answers all my questions and does most of my coding and I dislike it because it provides me with thorough context.

pietz··on A web version of Anthropic's prompt engineering interactive tutorial
I think we need to differentiate between a few things:

1. The specific way different models were trained.

Claude was specifically trained with these XML tags in the input and output. For Claude it works well to say things along the lines of "use a scratchpad for thinking out loud in XML tags first". While you can force GPT to do the same, it may not be as "natural" e.g. high chance of GPT3.5 not doing it.

2. XML tags vs full XML

Using XML tags doesn't mean it would be amazing at outputting full valid XML. XML tags are great because there's no further formatting needed within. If you only want to separate sections in your prompt or response, XML tags are perfect for clearly marking start and end, as well as featuring syntax (<>) that's rarely seen in natural text.

If it comes to generating a full data schema valid object, my experience is that JSON works way better than XML. It just seems to be a good middle ground between structure while having little overhead. I remember a recent AirBnB posts saying that YAML works even better than JSON. That's not my personal experience. Plus, YAML is something a bit weird and slow on read.

pietz··on The Ivy League and other elite private colleges are losing esteem
Excuse my ignorance but was it ever about the higher level of quality in education with the schools? I honestly thought the "getting in" aspect was the most important one.
pietz··on SentenceTransformers: Python framework for sentence, text and image embeddings
I haven't noticed this behavior in such extreme ways as you. I've worked with many different embeddings using this library and the OpenAI API for data normalization, extraction and structuring tasks.

On one hand it's super impressive what you can get out of these embeddings. On the other I'm with you that there seems to be a missing piece to make them work great.

An additional MLP or Siamese Network works really well but it seems like there should be something easier and unsupervised given your set of vectorized samples.

Thinking out loud, could PCA help with this by centering the data and removing dimensions that mostly model noise?

← PreviousPage 5 of 6Next →