ambewas

Exactly. It's been around so long and we still don't know how to mimic it.

The way an llm learns is a very interesting way of doing it, but it sure isn't what the brain is doing.

But it's indisputable.. We can get enormous results with this technique. It's just probably not the way forward for faster learning to remediate the issue of context loss.

That's exactly the issue. Retraining is too expensive & needs too much iteration to work efficiently I think.

Sleep would probably be a part of the equation for consolidating , but there's still the question of how exactly does the brain process the information during sleep in a way that it permanently consolidates the information.

It's not how an llm can work right now, it needs too much iterations & a much bigger dataset than what we can work with. A single time experiencing something and we can remember it. That's orders of magnitude more efficient than an LLM right now can achieve.

You're right to be skeptical. Without a way to actually implement how the human brain processes experiences into a consolidated memory, we won't be able to solve the long term memory problem at all. Not with the current technology.

An LLM context is a pretty well extended short term memory, and the trained network is a very nice comprehensive long term memory, but due to the way we currently train these networks, an LLM is just fundamentally not able to "move" these experiences to long term, like a human brain does (through sleep, among others).

Once we can teach a machine to experience something once, and remember it (preferably on a local model, because you wouldn't want a global memory to remember your information), we just cannot solve this problem.

I think this is probably the most interesting field of research right now. Actually understanding in depth how the brain learns, and figuring out a way to build a model that implements this. Because right now, with backtracking and weight adjustments, I just can't see us getting there.

in my experience, as long as you set up a decent set of agent definitions & a good skillset, and work in an already pretty clean codebase with established standards, the code quality an agent outputs is actually really good.

Couple that with a self-correcting loop (design->code->PR review->QA review in playwright MCP->back to code etc), orchestrated by a swarm coordinator agent, and the quality increases even further.

that would certainly be a very interesting thing to add, but in the current iteration that is not possible.

The biggest issue here is that there's not really "one" microtonal system out there. The entire fretboard of a conventional guitar is mapped to work in 12tet - and the libraries I'm using to do all the musical operations also only supports 12tet.

to accommodate all the microtonal temperaments out there would be a pretty daunting task. But I'm not saying never!

still working on https://stringscales.com - fun sideproject to visualize guitar scales on a configurable fretboard, with interactive note highlighting to a backingtrack.

The backingtrack is what I'm actively improving right now. It's just a pad running now, but it will turn into a full track with bass/drums/piano/... and will feature a comprehensive chords based editor so you can add and save your own progressions with a logged in account.

I was thinking the same. The author makes a great point when it comes to a portfolio, but most of the work I do is for corporate clients for example. They value efficiency and ease of use (I.e. predictability) of a ui/ux solution over any creative outlet. I don't think I should start doing more of that.. I'll be out of a job pretty fast

I've relied on Tonal heavily to build stringscales.com - it was a very pleasant experience. Much is already present, and extending the lib with more scales and functionality was easy as well. Definitely recommend

Oh nice, that looks really cool! How did you get the fretboard to sync up with those songs? Ai or some manual encoding work somehow?

Well done!

Oh right! I see what you mean. Got it, awesomeme idea: Chord charts like these are as of now a planned feature. Something to look out for in the future.

Thanks for the suggestion!

Very cool! I like the non standard fretting as well. Neat feature

Thanks a lot for sharing!

Yeah, I love the tuning option myself!

I totally agree, in fact feature number one is already planned. It'll probably be in the form of a chord schema generator and player that also shows you which scales and roots/thirds... you can use in a given chord.

For 2, it's possible to flip the fretboard to a lefty guitar. Is that what you are looking for? You can find the option under settings. Or do you mean flipping it vertically?

Definitely go for it! And let me know how I can help you to make it a little bit less overwhelming

thanks!

Thank you!

This might be a biased opinion, because it's how I do it and it works for me (ymmv), but I think starting with for example the major scale patterns in 3NPS is a great foundation. Once you know this by heart, and especially have learned the relative degrees of all the frets in the pattern, it becomes easier to learn a different scale formula relative to this major scale.

For example, when learning the myxolidian mode, I might go: "oh, but that's just the major scale with a lowered 7th degree!" And then I can apply that pattern easily on the fretboard.

Now, as for why this tool is useful (to me at least): it's really quick to visualise a position in a certain scale in a certain key, and show the note degrees (it's the "roman notation" setting). Especially for really exotic scales that have a lot of differences to the major scale (e.g. messiaen's modes), it is very useful to have a visual aid in front of me while playing.

thanks! I found that even just lazily continually looking and playing the scales and the positions while building the tool actually also really helped with learning more myself :)

can confirm, but also on google chrome

Ever taken a look at how bootstrap's CSS is written? It's horrible and even worse when you start to try to override certain things :(

Point taken!

What I would love to see in a design system is something in terms of "white-label" components. I'm talking about components with great UX, but basically no styling applied.

I'm not sure if what I'm talking about is possible, and I'm not sure what it would look like -- but I'd love a framework that just makes everything work really nice - except it doesn't force you to do really quirky css hacks to re-style the thing.

If that makes any sense :)

I love how ant design invents a new 'design language' and then it turns out to be yet another bootstrap.

I wish we'd get a real revolutionary design system instead of these frameworks that try solve all cases, and ultimately end up so generic you can't use them for anything -other- than prototyping

I'll be continuing work on Jam Together.

Jam Together is a web application which you can use to play some music together with your friends, remotely. Right now it's possible to play with a midi keyboard as well as just your laptop keys.

I've got a piano, a synth and a drum kit working (if you create a free account. Piano only without an account).

I'll be adding more instruments, work on reducing latency (anyone here with multiplayer game development experience that would like to help out?), different jamming modes, and most importantly work on some marketing to get more feedback on the app so I can evolve it in the right direction.

Check it out on the link below. A couple of disclaimers though... Chrome only, optimised for people residing in the EU for now, otherwise latency is too much of an issue to reliably play.

http://jamtogether.eu

rm: command not found

:(

> Rather than being tightly linked to structure, these are very loosely coupled - this is why one can often wireframe an entire application out before styling it!

Here's one of the premises on which the article was written - and I must wholeheartedly disagree.

If you're looking to write clean, maintainable CSS - you MUST consider DOM-structure of the thing you're styling as well. Too often, quirky CSS selectors are written, only because someone failed to foresee how HTML and CSS work together to create a visual.

This decoupling of the HTML that's written, and a CSS "theme" being written/applied after the fact, is also where things like negative margins (ugh) and the like tend to come from.

Now, I've been working on a very large component library in React, and keeping our CSS maintainable has not been an issue. We've got one (S)CSS file per component, and it resides in that component folder.

And while the file is separate, we absolutely do not allow anyone to develop any component anymore, without having any knowledge of how to style the thing.

We used to do exactly that: other devs would develop component, we would try to style them afterwards. This always results in either us writing messy CSS with crazy selectors or - and more often - we simply opted to re-write the entire HTML, to suit our styling needs.

At this point in time though, I would love to write some CSS-in-JS here. Too often we need to use variables based on JS logic; And it ends up as an inline style right now, anyway...

All in all, if the writer believes HTML and CSS can be written separately, without resulting in a terrible mess, I'm having a lot of difficulty accepting that they know what they're talking about. Sounds like they never had to work on, or maintain, a large application.