aleksiy123

Idk, I think it pretty clear to me that

“Find the issues in this code” vs “write this feature” are different problems.

The first one is easier because the feedback loop is llm to code/test and trivially scalable.

The second one is llm to human which is going to be slower because it can’t read your mind.

The tighter and more deterministic the loop the more llms excel.

Turning RSS, atom, blog feeds into audio feeds with kokoro tts. Free, web app only.

https://porchlit.app/

I built the original for myself because I wanted to listen to acoup and other history blogs while sleeping. And in the spirit of getting better at finishing projects decided make it public even tho it’s nothing really unique or special.

Still lots to polish especially in the text/copy and ui but the core features should be working.

It’s cause they don’t know what they are talking about.

Easier to shit on people and things with hot take then actually have a grounded and coherent argument or realise that different companies/teams have different constraints, problems and environments they operate as well as different solutions.

I think it’s relative size of literary work.

Mythos > Fable ? Opus > Sonnet > Haiku

Tho I think it’s definitely not obvious at first glance. And requires some thinking to order if you are even familiar with the terms in the first place (I had to look up opus)

Yeah I think the op is overestimating the effort it take to order them properly especially cross culture.

The ordering seems to be based on size?

I actually like the google one. Alphabetical categories. Essentially alpha beta without the existing connation.

I think the bigger problem it’s getting harder to order models linearly. As later versions of the model make different tradeoffs across dimensions.

Google does this research not sure if the others are doing it. Embodied reasoning

https://deepmind.google/models/gemini-robotics/

This comment is so snobby and obnoxious :/

There’s a lot of people different people from a lot of different ages and backgrounds.

Just because they haven’t seen all of the same sci-fi movies they aren’t qualified to work on ai?

Also, I love sci-fi but it’s also not always representative of reality.

It is bleak. But the cinematography is amazing, and I think worth it for that alone.

Maybe even the best I’ve ever seen.

Cool idea and I really like the design of the landing page.

The site overall looks nice too. It’s fun not to see slop for a ui for this kind of project.

I haven’t used jev yet. I’m curious if it can be used for entity labeling or extraction in text somehow?

I see box is now a thing but no span?

So render text to image then use box right?

Wow i was just thinking the same thing. Going to take a cbc course at my local shop.

Does anyone know what the SOTA is for cad models from image or multiple images?

I was hoping to try doing some reproductions of stuff i saw in museums/palaces in Europe.

I wanted to build this for phone numbers at some point but it was way outside of my skill level at the time.

I feel like unique address per sender, receiver pair could cut down on so much spam.

And you can see exactly who is leaking your info.

LLM/Agents have made the first 80-90% of a feature/project really fast.

But getting the last 10-20% actually requires your attention and many often many iterations.

Then you have 5+ things that are 80% done but kinda janky and broken in weird ways. Finishing them still takes work and discipline and can end up feeling overwhelming. Or you can swap to this new idea you just had in the shower, and get the dopamine rush all over again.

It sneaks up on you. Ironically discipline seems to be more important than ever. Easy to spend a bunch of tokens/quota and get nowhere but feel productive.

It’s a tradeoff for sure. But i think the problems just manifest a bit differently.

With multi repo it’s just a smaller constant friction on everything you do.

With mono repo everything is working amazing. Then you hit something big you need to solve to move on. So the friction is more spiky.

Edit: to add another thought. I feel like multi repo is more process oriented, while in mono repo you try to resolve that through tooling. Devs are just generally going to write about that more.

Also, Reduces bugs, increases agent velocity, doesn’t poison your code base.

All of that plus runtime performance and features.

All of these things still matter for agents. I think it’s just less visible to a dev now.

But velocity, bugs, token costs are still being impacted. You only see that if you are tracking it.

Scaling maintainable code base is probably more important than ever with the fire hose of code incoming.

Agreed. I tend to think of it like this.

Runtime config is a pure function of environment.

A versioned binary/artifact should be the same across environments.

Release stage is a property of environment. (local/dev -> preview/staging -> prod)

If behaviour or value of something changes across environments it belongs as config and not code.

I really believe having good rules like this can make a big difference in llm quality. So this is awesome thanks.

I’m personally using both oxlint for typescript and my own vibe coded tool for doing cross language rules ast rules and llm hook outputs.

And recently added vale as well.

I think in this case it’s from the element argon

What a flood of nostalgia.

I wonder if AI could help in reviving these.

With my zero knowledge of flash is it possible to somehow bridge it into modern browsers using web assembly or something?

To be fair this is worse.

It’s verbose with a layer of looking legit play first glance sloppy UI.

I don’t mind vibe code ui but at least some ui efffort would be nice. Not just 1 shot.

I also have been doing this every round of models releasing.

Thanks for reminding me to try this new round of models.

I wonder if there are some good tools/apis that can help here.

A lot of information is often login gated

Woosh?

As a sidenote from the slop language, which is awful.

Feel like latest models have really got hung up on “facts” and “evidence” in a really unnatural way.

And especially Claude seems to have a lot of trouble separating it’s internal thoughts, from prompts/instructions, to output.

So you get all of these weird comments and artifacts in the output.

“neighboring versions do not inherit certification. No tag, npm publication, Pages deployment, or release authorization is implied.”

Tyty I was thinking about something similar

You can also get LLM to optimize rules for a rules engine iteratively against some dataset.

It’s sort of like memoizing or distilling the knowledge. Works really well for certain type of problems.

The font, the colour, and the dot on the tags all scream Claude slop ui.

PlayStation store not allowing returns for games that have been downloaded is the most bs policy.

In a world where steam allows instant returns with under 2 hours play time.

I have a few games now on ps where I barely played or found out the controls were terrible and have no way to return.

You can do much more with feature flags like ab testing/experiments, integrations with analytics.

There can be whole UIs and tooling and infrastructure to manage around them and that’s what the sass offer

Managing complexity while scaling is a concept just keeps reappearing.

Funnily enough it’s extremely relevant to agents/subagents.