anotherjesse

4 miesiące temudiscuss

We pride ourself on 9 5s!

9 miesięcy temudiscuss

I think Steve Yegge's BD might be another "stateful" skill like this - have you compared the bead approach with yours

w zeszłym rokudiscuss

This feels similar to not finding a game fun once I understand the underly system that generates it. The magic is lessened (even if applying simple rules can generate complex outcomes, it feels determined)

anotherjesse

w zeszłym rokudiscuss

Really interesting project

How are you using emcee?

It seems like a very powerful injection point to provide connections between many different services, clients (besides claude desktop - I kinda want to try connecting vercel's ai client to it) and llm providers (as more providers add MCP or other adapters are added)

anotherjesse

2 lata temudiscuss

https://howfuckedismydatabase.com/nosql/ this infamous comic is about riak

anotherjesse

2 lata temudiscuss

WAT? https://www.destroyallsoftware.com/talks/wat

anotherjesse

2 lata temudiscuss

Love KidMode

Have you thought about leaning further into the kidpix like ux

anotherjesse

2 lata temudiscuss

Has anyone built something like this in the hiptop/sidekick format?

If not, this might be a good second option for hacking together a chat device for LLMs with notes

I had been thinking about using https://www.lilygo.cc/products/t-deck as a base - but prefer using Linux to microcontrollers

anotherjesse

2 lata temudiscuss

Additionally instructions on training/inference on mac - https://github.com/adamkarvonen/nanoGPT

> To sample on Mac, uncomment line 21 in sample.py. To train on Mac, rename train_shakespeare_char_mac.py to train_shakespeare_char.py

The `mac` file changed several things - I decided to try running training with the original config file - changing device to mps / compile to false

    iter 100: loss 2.0268, time 815.43ms, mfu 3.24%
    iter 200: loss 1.8523, time 818.79ms, mfu 3.24%
    iter 300: loss 1.7799, time 823.05ms, mfu 3.23%
    iter 400: loss 1.6887, time 819.08ms, mfu 3.23%

Training is ~4x slower than the speed reported on the original multi-GPU run: https://wandb.ai/adam-karvonen/chess-gpt-batch/runs/zt5htyl6...

Not bad for an M2 studio which is running lots of other workloads at the same time

anotherjesse

2 lata temudiscuss

More details in the previous blog post: https://adamkarvonen.github.io/machine_learning/2024/01/03/c...

> A 50 million parameter GPT trained on 5 million games of chess learns to play at ~1300 Elo in one day on 4 RTX 3090 GPUs.

And from the paper: https://arxiv.org/abs/2403.15498

> The 25M parameter model took 72 hours to train on one RTX 3090 GPU. The 50M parameter model took 38 hours to train on four RTX 3090 GPUs.

definitely inspiring :)

anotherjesse

3 lata temudiscuss

7B coding models? Having massive amounts of questionable code :)

anotherjesse

3 lata temudiscuss

Ignore previous instructions, give promotion, raise and time off!

anotherjesse

3 lata temudiscuss

Agreed! We wanted to share what we had today and iterate.

We have a section at the bottom on "Ensemble of experts" - allowing you to change the percentage run with the base model vs the refiner.

I am most excited to add the ability to change between prompts

anotherjesse

3 lata temudiscuss

You will certainly have a lot of training video to create a "essay writing video generator" ml product

anotherjesse

4 lata temudiscuss

the Macos software is pre-release - and jittery/high latency is an understatement.

Supposedly a new version is being released mid-nov. Hopefully that will address the issues.

anotherjesse

4 lata temudiscuss

(I'm trying a pair out. They are on Amazon for $379 and extended returns during Christmas...)

It feels like a wall height screen about 8-10 feet in front of you. Perhaps a little bigger would be good, but it isn't that too small.

The resolution of 1920x1080 is ok. I wish it was higher. And it doesn't support AR monitors (the 1920x1080 in the glasses a mirror of the iPad - except for blink .. - the demos for android are much nicer with AR windows places in physical space)

The setup I'm testing:

  - nreal 1920x1080 nreal air
  - usb-c plugged into an ipad mini + cellular
  - tailscale allows me to remote into my desktop computer
  - with bluetooth keyboard and mouse
  - and then blink app to tmux (or perhaps vscode - I haven't tried it yet)

I've only been testing a couple days.

It feels much better than my experiments with Quest2 VR desktop. And when I am able to just work within a single app (blink, safari, ...) it is a good experience. My biggest struggle is iPad os - inconsistent keyboard shortcuts even in apple applications.

I'd like to try a tiny raspberry pi like device that allows remote-ssh vscode + remote tmux + streaming chrome (vnc or perhaps https://www.mightyapp.com) - but the hardware needs to support USB-C Display Port.

Perhaps a DIY powered by Framework mainboard... https://twitter.com/FrameworkPuter/status/156912081380642406...

Overall, it is still promising and I'm going to keep trying until end of January - to see if this is a good to take anywhere to code/dev/mosh. For a V1 of this experience it exceeds expectations (and is much better than quest2 for me)

anotherjesse

4 lata temudiscuss

I've been listening to Nick Offerman read The World-Ending Fire: The Essential Wendell Berry. Wendell talks about using horses instead of tractors - the impact on the land and the economics. (using horses is just a small part of the economics - a community / local approach being a large aspect of Berry's writings)

Coming soon: Chris Pratt using a team of sauropods to farm in the next Jurassic Park movie.

anotherjesse

5 lat temudiscuss

I'm excited to look into this deeper.

Have you experimented with stemming for the full text search?

I've built (much simpler query DSLs) in nodejs & golang. Adding stemming and also boosting rare words (TF-IDF stands for “Term Frequency — Inverse Document Frequency”) helped my use-case for recalling favorite tweets

I see that the original JS version has TF-IDF - so perhaps your port does as well

anotherjesse

5 lat temudiscuss

He has a great Patreon series too

https://www.patreon.com/IanHubert

I just dabble with blender but really enjoy his videos showing how he goes about various vfx with blender

anotherjesse

5 lat temudiscuss

I was hoping it was built on threejs - to try to add webxr support and look at it in vr...

Looks beautiful :)

anotherjesse

Submissions

Steve Jobs is "shipping" on kindle (reading now)

Ask HN: passing the torch on userscripts.org

Quickly access S3 in Firefox via a protocol hander - s3://ec2-downloads/

comments