0vermorrow

Latest Haiku (smallest Anthropic Model) is version 4.5, they haven't released a new version, hence the comparison to that.

I'm eagerly awaiting for Qwen 3 coder being available on Cerebras.

I run plenty of agent loops and the speed makes a somewhat interesting difference in time "compression". Having a Claude 4 Sonnet-level model running at 1000-1500 tok/s would be extremely impressive.

To FEEL THE SPEED, you can either try it yourself on Cerebras Inference page, through their API, or for example on Mistral / Le Chat with their "Flash Answers" (powered by Cerebras). Iterating on code with 1000 tok/s makes it feel even more magical.

It's funny how we went from using iPod Nano as a watch with a third party case, to using an Apple Watch as an iPod Nano with a third party case.

Well it's not broken, I'd assume they limit the `max_tokens` to minimize costs.

Well see, there is your problem, because the author says this in the immediate next sentence

> In other words, it was as unsuitable an environment for psychedelic experimentation as I could have designed [...]

Are there any users here who actually ran Brave ad campaigns?

I'm wondering what is your overall experience, how much of your ad campaign traffic were bots and how did it fare compared to other ad platforms.

Are you missing an /s or is this a serious comment?

https://www.merriam-webster.com/dictionary/crux

Elon Musk has his hacked as well... https://twitter.com/elonmusk/status/1283495825998520320

What just happened?

The pushback from users made them rethink this change, here's a pending merge request to the blogpost: https://gitlab.com/gitlab-com/www-gitlab-com/merge_requests/...

""" UPDATE: Thanks for the feedback. There were many more concerns than we expected. We’re going to process the feedback and rethink our plan. We will not activate product usage tracking on GitLab.com or GitLab self-managed for now. We'll make sure to communicate in advance on our blog when we do have a new plan. """

So after the authors disclosed this issue to AWS it was fixed and CloudFront no longer caches 400 Bad Request by default, also from the paper linked on the website [0]:

""" Amazon Web Services (AWS). We reported this issue to the AWSSecurity team. They confirmed the vulnerabilities on CloudFront. The AWS-Security team stopped caching error pages with the status code 400 Bad Request by default. However, they took over three months to fix our CPDoS reportings. Unfortunately, the overall disclosure process was characterized by a one-way communication. We periodically asked for the current state, without getting much information back from the AWS-Security team. They never contacted us to keep us up to date with the current process.

"""

[0] - https://cpdos.org/paper/Your_Cache_Has_Fallen__Cache_Poisone...

You mean like https://www.gremlin.com/ ?

One of the founders of Gremlin is an Engineer that worked in Netflix and probably worked on Chaos Monkey as well :)

While we wait for the delayed stream to start, it's worth taking a peek at their job board which has some interesting bits of information in the job descriptions: https://jobs.lever.co/neuralink

For example:

> "As an Optical Engineer at Neuralink you will develop custom optics and imaging systems used directly in our surgical robot, and associated consumables manufacturing systems." [0]

EDIT: I guess the article from The Verge linked in the other comments sheds even more light on that particular bit:

> "[...] the company has developed ”a neurosurgical robot capable of inserting six threads (192 electrodes) per minute [automatically],” according to the white paper. " [1]

Similarly described in the Bloomberg article which was just released [2]

[0] - https://jobs.lever.co/neuralink/c98c011c-0f49-497e-a504-59a5...

[1] - https://www.theverge.com/2019/7/16/20697123/elon-musk-neural...

[2] - https://www.bloomberg.com/news/articles/2019-07-17/elon-musk...

It's probably because i use xterm-kitty.

Awesome work on aerc, looking forward for continued development.

Yeah, the problem is I am on OSX, and I actually had to compile scdoc myself without the -static flag (missing libraries).

Also, I had to run it as:

  TERM=xterm-256color aerc

On OSX, otherwise got:

  panic: terminal entry not found

Love the idea, and I really will give this a go as a daily driver.

Just a small feedback: I had trouble with one of the requirements -> scdoc, didn't know what exactly that is (Googling didn't help), so might be worth to provide a direct link in the README.md to https://git.sr.ht/~sircmpwn/scdoc/

Again, it's mainly for online payments, the others are an addition, and a welcomed one at that if you forget your card (or don't have Google / Apple Pay) when you're out somewhere.

It is quite well accepted, as in:

- the biggest Polish banks (whose initiative this was) support it, and the ATMs (Euronet included) usually have that as an option.

- ALL of the Polish payment providers / gateways support BLIK.

- Poland is quite infamous for supporting contactless payments [0] and most of the POS terminals support BLIK as well.

- For merchants it also makes sense as there are no chargebacks.

I do concede that for BLIK vs. contactless payments in the brick & mortar shop the latter always would win for quick payments, but why does it seem high friction to you?

[0] - https://zalewskiconsulting.pl/mastercard-poles-love-contactl... and also Android / Google / Apple Pay

Also on a related note woth mentioning is Poland's BLIK[0] it's a very nice system based on generated one-time codes that are confirmed via your mobile device.

It works for online payments, ATM withdrawals and even in brick & mortar shops.

The flow for online looks like this:

1. Go to the Payments / billing page.

2. It is almost prevalent now that you just see a BLIK input box immediately, and the other standard "pick your payment options" box below.

3. Fire up your bank's mobile app, navigate to the BLIK tab (normally a swipe away), grab the 6 digits (or copy)

4. Put them in the box (or paste)

5. Press OK on your mobile device (you can even 'trust' your browser after succesful payment, such that you don't have to confirm the next payments on your mobile device with some monetary limits).

6. Done.

It usually takes less than 30 seconds to do all of this. No need to put your CC details or anything in.

[0] - https://polskistandardplatnosci.pl/en/

[more-info] - https://www.ppro.com/wp-content/uploads/dlm_uploads/2018/04/... https://www.finextra.com/pressarticle/71555/blik-becomes-the...

I completely agree with this.

I feel him taking some time off and easing off would benefit everybody. He is known for his super focused and driven way of work, and as he himself said:

“If other people are putting in 40 hour workweeks and you’re putting in 100 hour workweeks, then even if you’re doing the same thing … you will achieve in four months what it takes them a year to achieve.”

But he is only human, not a robot (to the best of my knowledge), and sleeping under the desk in the factory overworked will stress out everybody and the people in their close environment no matter how logical-thinking person you are. Building your company in a constant fight-or-flight mode is a disaster in the making.

Balance is important. For the sake of sanity, health and actual productivity.

Although this is completely anecdotal, I had a similar "100 hour never stop working" approach with my initial start-up ventures (obviously not of the scale of Tesla or SpaceX) which ended up me destroying my personal relationships and I cannot say my productivity was off the roof.

Then I slowly transitioned into having a proper balanced day, with exercise, meditation and PROPER time off. Also having healthy work habits inspired a lot by Learning How to Learn (popular MOOC on coursera) and work by Cal Newport ( Deep Work ) with a sprinkle of org-mode here and there.

I have to say that first time in my life, I feel happy, fulfilled and proud of my "weekly output"

The paper itself doesn't focus on Machine Learning / AI or anything too specific for that matter (it reads like a review / praise of LISPs), but I am currently researching Symbolic Artificial Intelligence as well as let's say the history of "Good Old-Fashioned Artificial Intelligence" and the role LISP took in all of that.

Any books / papers / articles that one should definitely read on these topics?

It's straight up in the article...

“Please refer anyone you know who you think meets the Tesla bar for talent, drive and trust” he wrote. “Between Fremont and Giga, Tesla will be adding about 400 people per week for several weeks.”

I love this. Also a super fan of the UI. Seamlessly led me to a download button with the source. They've coupled it with an awesome hosting setup process. I can promise you I will use this service for the very next project of mine.

Commoditisation of design cannot come soon enough! ( at least the web design part :) )

My initial feedback would be that the music is way too loud and that I really like that you used Python to show how to construct the bits and pieces.

I am in Germany as well, and it redirects me to blogspot.de but the actual content is in english.

Not yet.