tdhttt

Cool project. Call me pessimist but all of these workflow/scaffolding/hooks are being absorbed by the model makers like ChatGPT and Anthropic. In fact, my personal setup with Claude Code is somewhat similar to this setup, except I bend it in CLAUDE.MD

Interesting story! During my short stay at CERN, I lived in an apartment in Pays de Gex as well. Cycling across the border twice a day is fun! (The border is not very bike-friendly though)

I am doing this and it's fun! The only hassles I had are a. setting up DDNS (since I don't have a static IP) and b. getting a certificate from Let's encrypt with only port 443 open (80 is blocked by my ISP). But other than that, it has been great! Hopefully my ISP won't have a problem with that ;)

> “Biological objects can have a complexity that cannot be simulated by a digital computer,” Austin says.

I couldn't find more details in the article. Any idea what those details that couldn't be simulated with digital computers (but could be with the simple robots) might be?

Same. Also the author mentioned peer review, is that available on substack?

Please check "Are paywalls ok?" at hacker news [FAQ](https://news.ycombinator.com/newsfaq.html)

Also, previous discussion about WSJ: https://news.ycombinator.com/item?id=13434938

The [paper](https://www.nature.com/articles/s41567-020-1008-4)

The first issue was published today (Sep.3) so I thought this might be interesting.

Sorry for not being clear, I meant when all 21 million bitcoins are mined. The transaction fee is pretty small compared to the reward from the new coin, right? Also, since many cryptos have addressed the incentive issue. Maybe let's just focus on bitcoin specifically?

> The network hash rate is only so high because the ROI is there, but without it, people would still be mining.

Can you expand on that? What would be the incentive for the miners then? What happens when all blocks are mined?

They featured WSL in the Surface Book video! And terminal is in the dock!

522 in Southern California

The rear facing camera seems to have a nice upgrade. However, apart from AR applications, I found it weird to hold a 12 inch tablet in the air to take pictures while everyone could see my screen.

~~Why does one have to send an email just to ask about license?~~

Edit: As people had pointed out, that they were asking for chat to integrate it into the email client. Sorry for the confusion.

I just implemented something very similar last week! I store the note in a Raspberry Pi which I can access anywhere. Also with iOS shortcut, I can easily send note to my Pi from my iPhone's clipboard.

Previous discussion:

General: https://news.ycombinator.com/item?id=18336202

Version Control: https://news.ycombinator.com/item?id=21661013

Does DDG have anything special that could metigate the current SEO madness?

When using ML models to deal with natural science such as physics, working well+being interpretable actually implies new discovery. Working well only, on the other hand, cannot win you a Nobel Prize.

One more benefit: an interpretable model could at least give some hint when the model is not working as intended (e.g. debugging).

[Here](https://www.peak-labs.com/docs/en/magi/intro) is the English version of the intro.

Qiskit[0], the framework that IBM Q uses, also has a awesome community and the introductory textbook called "Learn Quantum Computing using Qiskit" [1]. The "Quantum Algorithms" chapter is thoroughly documented both from theoretical and coding perspective.

[0]: https://qiskit.org/

[1]: https://delapuente.github.io/qiskit-textbook/preface

Also for a more comprehensive listing of the quantum software, see the awesome-quantum-software [0] on GiuHub.

Personally I have used Cirq[1] and Qiskit[2] to implement some classic algorithms like Deutsch-Jozsa[3] and Shor's[4] algorithm and I would say the two frameworks are very similar to use for those implementations.

So, it would be interesting to get some insights/opinions from someone who has used multiple frameworks. Is there any clear win/loss of those frameworks?

[0]: https://github.com/qosf/awesome-quantum-software#experimenta...

[1]: https://github.com/quantumlib/Cirq

[2]: https://qiskit.org/

[3]: https://en.wikipedia.org/wiki/Deutsch%E2%80%93Jozsa_algorith...

[4]: https://en.wikipedia.org/wiki/Shor%27s_algorithm

Summary:

+ Amazon didn't build any quantum devices.[0]

+ AWS provides a software stack called Braket. [0]

+ AWS provides classical simulation or access to quantum computers from D-Wave, IonQ, and Rigetti.[0]

+ Many companies (e.g. IBM, Rigetti) have already provided similar software stack/cloud quantum devices. For a more comprehensive listing, please see [1].

[0]: https://aws.amazon.com/blogs/aws/amazon-braket-get-started-w...

[1]: https://qbnets.wordpress.com/2019/11/23/list-of-quantum-clou...

Citation needed?

I am using a decent gaming rig, so I don’t think it is the performance issue. I think it might be related to the touchpad. Are you using mouse for scrolling?

What drives me away from Chrome on Windows is the very short freeze you have after you let the page idle for a while and try to scroll again. It was introduced earlier this year, I think.

Side note, the Edge (not the ChromeEdge) has a nice feature where you can set aside all the tabs so that you can view them later with pretty much the same state, much like a tmux server. I couldn’t find similar features in other browser.

I apologize for the slow loading... I am trying to optimize...

I have several questions about Karpathy's presentation: 1. How to label automatically using fleet learning? An example he gave is detecting cut-ins. However, I am wondering if it could be applied to more general and basic cases, for example, labeling lane marking/vehicles. 2. How do you know if someone is a good driver (which is essential for imitation learning)?

So I work in experimental particle physics and our daily tasks involve using particle collision simulation tools(mg5+pythia+delphes), distribute them on clusters, construct jet images using ROOT and finally experiment with neural networks on those jet images. So as people pointed out, what we do had a lot in common with the industry (where we learn from as well), which explains why many people involved in this kind of research could land a data scientist job. An interesting thing I've noticed is that many phD students in our group (with little CS background) waste lots of time on getting the software running (either locally or on cluster). You guessed it, we are trying to exploit the power of docker to make it easier for researchers to run those simulation tools.