smahs

There was a time when local apps written in binary compiled languages were notably faster to run, but notoriously difficult to write and scale across platforms. Both of these have improved but the JS/CSS run times used for local web apps (V8/Blink and JSCore/Webkit) are now fast enough for well written code to not notice any difference for many use cases. I think soon most SaaS apps without much of a deep tech moat (not easy to replicate) will be labelled as rip-off or anti-patterns.

The above comment, to which you responded, wrote about CEOs who responded to mass hysteria, not those who automated anything.

Not wrong, perhaps a little weird. HTML is a first class citizen in the commonmark (spec)[https://spec.commonmark.org/0.31.2/#html-blocks].

In an agentic loop, the model can keep calling multiple tools for each specialized artifact (like how claude webapp renders HTML/SVG artifacts within a single turn). Models are already trained for this (tested this approach with qwen 3.5 27B and it was able to follow claude's lead from the previous turns).

There is nothing wrong with big money backing, often is necessary for long term bets, but rug pulling is a serious threat. VC funded open source has become a pattern/playbook.

27B dense model is probably the best in the 3.5 lot, not absolutely but for perf:size. Its also pretty good at prose, which is a rarity for a Qwen.

The author doesn't explain (or is ignorant about) why this happens. These are special tokens that the model is trained on, and are part of its vocab. For example, here are the <think> and </think> tokens defined in the [Qwen3 tokenizer config](https://huggingface.co/Qwen/Qwen3-235B-A22B-Thinking-2507/bl...).

The model runtime recognizes these as special tokens. It can be configured using a chat template to replace these token with something else. This is how one provider is modifying the xml namespace, while llama.cpp and vllm would move the content between <think> and </think> tags to a separate field in the response JSON called `reasoning_content`.

I only did it once some 15 years back (in a happy memory) using LFS. It took about a week to get to a functional system with basic necessities. A code finetuned model can write a functional chat UI with all common features and a decent UX in under a minute.

Still O(n).

Thanks for noticing this, and yes I have also noticed what you're pointing out, but workable for many use cases. I use this workflow for making images for marketing or web (so images are more artistic than photo realistic generations to begin with). Think of stuff you can find on undraw, but generated by image models from prompts. Then run them through VTracer. The reproductions are not perfect, but are often good enough (can be slow depending on how sharp you want the curves, and often very large file sizes as you mentioned). Then make any changes in inkscape and convert back to raster for publishing.

> logo generation app

For logo generation, I would actually prefer code gen. I thought of this problem when reading about the diffusion language models recent (if there is lots of training data available in form of text-vector-raster triplets).

Gemma 3 4B (QAT quant): Yes, Paul Newman was indeed known to have struggled with alcohol throughout his life. While he maintained a public image of a charming, clean-cut star, he privately battled alcoholism for many years. He sought treatment in the late 1980s and early 1990s and was reportedly very open about his struggles and the importance of seeking help.

^^ I have vague memories of Clippy, but I remember it as obnoxious, often consuming the precious screen real estate on the low res monitors of the day, without offering anything valuable. But tooltips on the web with CSS and libraries like floating-ui can be much more compact, agile and barely noticeable.

I have tried showing help text in an internal app using tooltips when the user would hover over the target element (or show a small icon on touch devices), and the feedback was good as the tooltips were never in the way but easily available for help (accessibility for keyboard users needed some thinking, but for the limited audience for that app, it was not a problem). And while you're at it, may be make it more engaging than a simple text only tooltip (which can be done without any intelligence), and let the host to customize and offer complex workflows.

No it won't (most likely). VTracer (which the authors compare with) is fast, runs in browser via wasm, consumes way less resources and can even convert natural images very decently. But the model seems cool for the usecase of prompt to logo or icon (over my current workflow of getting a jpg from flux and passing it through VTracer). I hope someone over at llama.cpp notices this (at least for the text-to-svg usecase, if not multimodal).

On the contrary, I have found reasoning models (DS R1 mostly) to be very good at complex positioning and transition problems. They can't "visualize" anything, so can't design well unless you explain them well (which is the problem in design, most people have vague ideas but can't explain it well in CSS terms).

And geometry, which is high school level for the most part IMO but combined with cascading in the third (z) dimension. Introduce 'relative' with its own coordinate system and then do transforms in it (to be fair, its only complex in some cases like where the transformed parent is different from the relative parent). And then get into the time domain for transitions. Its math after all, but not the same that most programming courses teach.

Leaving the distortions from inflated and unrealistic expectations (case in point: people expecting evolution of AGI somehow have not yet well defined what AGI is), I also think that in the mid-long run the current state of LLMs will bloom an entire economy for migration of legacy apps to have conversational APIs. The same investors will then have a new gold rush to chase, as it always happen.

But that's the story of every entrepreneur. Whether you're the smartest builder or sales person, nobody goes big in a single attempt, you keep improving. An attempt doesn't necessarily mean changing businesses, but improvising until you find what works. What seemed implicit to me is that the OP is looking for responses from early career entrepreneurs.

I first found Seal which uses arai2 under the hood. After getting used to Seal, I had to start using yt-dlp with aria2 on linux as well. The only nitpick is that you can't get multiple parallel downloads with Seal.

This sentiment is pretty common I guess. Outside of a niche, the massive scale for which a vast majority of the data tech was designed doesn't exist and KISS wins outright. Though I guess that's evolution, we want to test the limits in pursuit of grandeur before mastering the utility (ex. pyramids).

I think I get your perspective (but also you should really understand that your arguments are just going unidirectional instead of building a healthy dialog), but it made me wonder what really went wrong (or changed, depending on your perspective) from Hegel's utopian egalitarianism to how communism was implemented in practice a century later. I guess people just tend to "adapt" legal and governance theories built with perfectly good intentions to their advantage and then things go awry.

Since when is left == religious fundamentalism?

This is surely a gross oversimplification of what's happening. If you see layoffs.fyi, there are a lot of rows with more than 5% fired or where the numbers are in several hundreds to thousands.

While some companies are known to regularly use the tactics you mention (and a lot others also do the same from time to time), they don't use mass firing as a means.

+1. I have vague memories of my time at a Japanese automation vendor out of the uni. I had quit out of frustration that the software was super buggy, there was no one to help except just a couple of people in Japan who knew the software but would not reply to emails. I also remember feeling neglected as folks working on the hardware or on customer projects were paid higher than me.

A recent experience at a neobanking startup from SE Asia reaffirms the point. Despite the product built around an API-only model, the firm was operations heavy when it came to decision making and investing in people, as it was believed to be the core company strength (for a variety of reasons including the institutional bureaucracy, corruption in these markets, etc.)

TL;DR people work where the money flows. Companies get what they pay for. And the investors pay for what they think is the strength or is likely to sell at inflated valuations.

+1 I recent saw a couple of cases of small companies who chose to depend on Stripe for managing their payment data to save cost on initial development. As a result, their operations staff struggled between Stripe's admin panel and their internal tools even to answer simple user questions.

Eventually, when they decided to streamline their payment and billing operations on their own tools, due to historical data on Stripe servers, they built a custom presentation layer on top of Stripe APIs. They ended up spending pretty much the same time and effort as they would have done by managing the business flows on their own systems and storing all the data on their DBs and using Stripe just for processing payments. Worse yet, they still depend on Stripe for stuff like reporting.