toddmorey

I’m worried that opportunities like this to build fun/interesting software over models are evaporating.

A service just like this maybe 3 years ago would have been the coolest and most helpful thing I discovered.

But when the same 2 foundation models do the heavy lifting, I struggle to figure out what value the rest of us in the wider ecosystem can add.

I’m doing exactly this by feeding the papers to the LLMs directly. And you’re right the results are amazing.

But more and more what I see on HN feels like “let me google that for you”. I’m sorry to be so negative!

I actually expected a world where a lot of specialized and fine-tuned models would bloom. Where someone with a passion for a certain domain could make a living in AI development, but it seems like the logical endd game in tech is just absurd concentration.

The ethics are exactly what the DoD is complaining about. They want any legal action to not be obstructed by guardrails.

I don't think it's sanctimonious to say, hey, I don't want the technology I work on to be used for targeting decisions when executing people from the sky. Especially as the tech starts to play more active roles. You know governments will be quick to shift blame to the model developers when things go wrong.

Yeah that sentence struck me as very carefully worded. They also don't mention how often RA is needed or invoked. We'll encounter a lot of these autonomous systems (cars, robots, equipment) that escalate decisions and edge cases to human employees until they are trained enough that reliability goes up.

You are not the one folks are worried about. US Department of War wants unfettered access to AI models, without any restraints / safety mitigations. Do you provide that for all governments? Just one? Where does the line go?

Maybe. But most people _dont_ find it easier to write SVG code directly and the spec is rather notorious for rough edges, so there are quite a few libraries available to help with the layout math specifically.

Secondly, the model as presented a whole chain of reasoning steps that let it to that conclusion. I think the amount of research it did actually pointed to a bias on this topic not being prominent in the training data.

I mean… What does your shop even do? Write software? Why? The whole premise is that it’s now easily cloned.

It’s fun to ask the models their input. I was working on diagrams and was sure Claude would want some python / js framework to handle layout and nodes and connections. It said “honestly I find it easiest to just write the svg code directly”.

I love this so much. I do worry about the hidden cost of maintaining these software stacks that have a userbase of one. Things like hidden vulnerabilities, bugs, and edge cases that could bring a business down for days.

But I do think the days of just having to learn, tolerate, and accept any commercial software stack for your business because it's too complex to build yourself are over. What vendors remain will have to absolutely meet users with their unique requirements and budget.

“He should’ve been so aware that his job was going to be the first one cut, and he was responsible for building a tool to cut his own job”

That tool was going to get built whether he did it or someone else did. Maybe only thing to do is buy time building it while actively looking elsewhere.

I think this is exactly why Anthropic bought Bun!

Well LLMs do make normal Linux tooling more accessible. I needed a video reformatted to a new aspect ratio and codec and Claude produced a rather complex set of arguments for ffmpeg that I hadn’t been able to figure out on my own.

Right you are. Nature can be violent, but prefers gradual change. Abrupt change shocks ecosystems and always comes with unintended consequences.

I struggle with the thesis that our institutions haven't already been fatally wounded. Social media and endless content for passive consumption have already errored the free press, short-circuited decision-making, and isolated people from each other.

OpenAI is here because Sam Altman is NOT a product guy. He craves Apple style consumer success, but he's terrible at productizing his technology. Remember the marketplace of custom GPTs? Hell even the name ChatGPT. Anthropic had to show them how to build useful workflows for developers using AI. Meanwhile, OpenAI delivered... Sora.

To actually quote Sam Altman: "I think of ads as a last resort for a business model."

It's just maddening to me because this technology could be so much more useful for purposes way beyond advertising.

TERRIBLE! Can’t believe Apple is using these.

Q: What would prevent them from using git style version control under the hood? User doesn’t have to understand git, Claude can use it for its own purposes.

Commercial OSes (both Windows and MacOS) now feel so insanely agenda driven, and the agenda no longer feels like anything close to making the user happy and productive. For Mac, it feels like Apple wants to leverage what came out of VisionOS and unify the look and feel of mobile and desktop--two things no one asked for. For Windows, it feels like ads for their partners and ensuring they don't fumble the ai/agent transition the way they did with mobile.

Linux is SUCH a breath of fresh air. No one wants it to be anything other than what you want it to be. Modern desktop Linux has a much improved out of the box experience with good support for all the hardware I've thrown at it. And Claude Code makes it very fast and trivial to personalize, adapt, automate, etc.

It was more than a library of prewritten css, though, they did quite a bit of engineering work on tooling (speeding up the code scans and dynamically creating custom classes, etc). I respect the team's productivity.

This is more a question about the business model of open source, which has always had some challenges. I don't think you can support OSS with premium templates, training, and support once the knowledge is baked into LLMs.

I didn't understand this argument, either. I think the argument is that general purpose skills will likely continue to be baked into the LLMs, so we'll see less of the repos of "expert designer" and "python expert" as if you are creating personas. But as a way to teach an LLM bespoke processes or APIs, I think skills make a ton of sense and are a super small footprint.

Cursor if I recall actually started life as a VScode plugin. But the plugin API didn’t allow for the type of integration & experiences they wanted. Hit limits quickly and then decided to make a fork.

this age, feels like the most dangerous thing the Chinese government could do is sell that data back to our government.

Oh smart. This is good guidance. Yeah fascinating how longer running context causes these side effects, especially the repeated image with no changes bug.

I think that's the right approach and what I've been experimenting with. Diagram as code and then style transfer from output diagram to desired look. That's where I've had the most success.

I mean, one of its flagship features is to make precise edits to images. And it's really good at it... until it randomly isn't.

Yeah, that’s very well put. They don’t store black-and-white they store billions of grays. This is why tool use for research and grounding has been so transformative.

Confident idiot: I’m exploring using LLM for diagram creation.

I’ve found after about 3 prompts to edit an image with Gemini, it will respond randomly with an entirely new image. Another quirk is it will respond “here’s the image with those edits” with no edits made. It’s like a toaster that will catch on fire every eighth or ninth time.

I am not sure how to mitigate this behavior. I think maybe an LLM as a judge step with vision to evaluate the output before passing it on to the poor user.

GitHub seems to suggest banners or “Also consider ways to notify the user in other communication channels such as email, notifications, or a push notification in the GitHub app.”

On MacOS… emails and push notifications create… toast messages

When async notifications arrive from background processes… How is the user notified? (Not defending toasts, just curious how to do it better.)

Wait I don’t get it. CF is free for your use case, but you are looking for a cheap CDN? What’s cheaper than free?