ea016

I've been using Tanstack AI (https://tanstack.com/ai) for this. It's been really easy to setup and sturdy. They offer the foundation for tool calling, handle reconnects, persistence, computer use, voice and more. If you're on the web I really recommend it

Instead of this I use Leechblock to cut the /feed access after 2 minutes, to prevent doomscrolling (no DOM change)

For our mobile app we ended up doing YEAR.WEEK_NUMBER.BUILD. It makes it much easier for our team to know how old a version is (support, analytics, engineering). And regular users don't care about the version

Price comparison:

GPT Image 2

  Low     : 1024×1024 $0.006 | 1024×1536 $0.005 | 1536×1024 $0.005

  Medium  : 1024×1024 $0.053 | 1024×1536 $0.041 | 1536×1024 $0.041

  High    : 1024×1024 $0.211 | 1024×1536 $0.165 | 1536×1024 $0.165

GPT Image 1

  Low     : 1024×1024 $0.011 | 1024×1536 $0.016 | 1536×1024 $0.016

  Medium  : 1024×1024 $0.042 | 1024×1536 $0.063 | 1536×1024 $0.063

  High    : 1024×1024 $0.167 | 1024×1536 $0.25  | 1536×1024 $0.25

Didn't work for me either, I asked Cursor to export the session from ~/.claude/projects/

Well, as Jeff Atwood famously said [0], "any application that can be written in JavaScript, will eventually be written in JavaScript". I guess that applies to embedded systems too

[0] https://en.wikipedia.org/wiki/Jeff_Atwood

We're on the enterprise plan, so far we're seeing Dashboard degradation and Turnstile (their captcha service) down. But all proxying/CDN and other services seem to work well.

No relations to them, but I've started using Happy[0]'s iOS app to start and continue Claude Code sessions on my iPhone. It allows me to run sessions on a custom environment, like a machine with a GPU to train models

[0] https://github.com/slopus/happy/

2 interesting facts about it:

- it was completed ahead of schedule and with no budget overrun. The construction company (Eiffage) had a strong incentive to do so: the deal was that they supported a most of the cost but in exchange got to collect the tolls

- they have small mirrors all over the viaduct used to measure its movement - a bit like real-life telemetry

Google Cloud Storage seems to be down or very slow

Vercel heavy user here. They have a very misleading pricing. It's "starting at $150", it varies depending on the region. I end up paying $400 / TB as we have a very international website.

A really great article. However they tested on nextjs.org only, so it's still possible Google doesn't waste rendering resources on smaller domains

I couldn't share any pricing data since the discussions with providers are private. Instead, I added a graph of prices from gpulist.ai. For an Infiniband cluster, median is $2.3 per H100 hour, average is $2.47.

I think Google Slides supports video, the downside is that you need to upload them to Google Drive first

Steps to do so:

- open the network console

- right click on the request

- click "copy as curl"

- visit https://curlconverter.com/ to convert to Python/Node/any language

Thanks, fixing the typo :)

OP here. Here's a comparison with an input image vs gray levels vs dithered: https://imgur.com/a/osE57eh

When displayed on the frame you can barely see the dithering. Interestingly you see the jpeg compression artifacts a lot more

OP here. Your project is really cool, thanks for sharing. I was considering the same idea, but it means updating the drivers to work on an ESP32

The best option I found so far is a timer shield that can wake up the Raspberry periodically: https://www.pishop.us/product/sleepy-pi-2-power-management-s...

Eliot here, I work with Vincent on PhotoRoom. Our current focus is shipping an incredible app on the web and on mobile, but we'll tackle video very soon!

There's a good video on the topic here: https://www.youtube.com/watch?v=zR0M7KjnJTE

I also really enjoyed the beautiful "Meet the Machinists Who Keep the New York Times Running":

https://www.youtube.com/watch?v=tGHStfuLdyY

If you're going to travel a lot, I would recommend using a lightweight laptop and use a remote server with GPU.

You can either access GPUs on demand or rent a server by the month. Hetzner[0] offers a GeForce 1080, 64 GB ram and Quad-Core i7 for ~$115 / month (+ setup fee).

[0] https://www.hetzner.com/dedicated-rootserver/ex51-ssd-gpu

Great opportunity to remind people: never ever plug an unknown USB device in your computer. Even USB-C chargers can infect your machine as outlined here: https://twitter.com/_MG_/status/949684949614907395

Some macs have the ability to load OSX from network or from an external hard drive. It might be the same on the iPhone

For those of you who want to play with transfer learning, Keras has a nice collection of models for computer vision available here: https://keras.io/applications

Shameless plug: And if you want to try them out in your browser, I made a website that allows you to try pre-trained models: http://pretrained.ml

First YC company to go public ! Congrats

It's always fun to look at the comments at the time the MVP was posted here on HN:

https://news.ycombinator.com/item?id=8863

The title is misleading: they announced a ban of the sales of new petrol vehicles by 2040 [0].

[0] (in french) http://www.lemonde.fr/planete/article/2017/07/06/nicolas-hul...

I always wonder why people always use libraries when building software but never use templates when making presentations. Graphics and transitions are extremely important when presenting, so why not build on top of other people's work ?

I highly recommend using templates. You can get some really good ones here (from 5 to 30 dollars)[1] or for free using google.

[1] https://graphicriver.net/category/presentation-templates/pow...

Could we please add (2012) to the title ?

To get more background on this "research", you can read the Wikipedia article: https://en.wikipedia.org/wiki/S%C3%A9ralini_affair

For those interested, episode 2 is out [0]. There is also a playlist containing all episodes [1]

[0] https://www.youtube.com/watch?v=tNa99PG8hR8 [1] https://www.youtube.com/playlist?list=PLOU2XLYxmsIIuiBfYad6r...