ttt3ts

I work with a bunch of brands and the marketplace itself is a huge problem. Stolen goods, return fraud, Amazon directly competing after finding a profitable brand/product, chinese knockoffs/junk, mixed inventory, ads, many more. The tools amazon provides to combat these are largely useless.

Diving into one I think is detrimental to both sellers and customers, Ads. If you are not a huge brand this is a ~mandatory additional cost on top of FBA (or you won't show in results). As a customer it sucks because many of the top results are trash. Why bother having a decent product if you can just buy a ranking and some fake reviews? Amazon is pushing this so hard right now as it is a quickly growing revenue stream for them :(

Amazon FBA takes a big percentage, abuses sellers/brands, and significantly limits business models (e.g. international sucks) There is a market here. I hope they pull it off.

Man, I miss my TJ. I now have a family car with excellent crash ratings. Take the top off and enjoy it while you can :P

Interesting. I guess I agree on the idea just not the specific number of 48GB. Amd & Nvidia both offer prosumer GPUs with 48GB, the W7900 and A6000 respectively. I don't see how they would not increase that in 5 years?

5 years?! You think the enterprise hardware hype will last that long? When it slows companies will look at other less valuable markets.

Unless something fundamentally changes with the capabilities of produced tech I think demand will be saturated before 5 years.

You have to pass the context between GPUs for large models that don't fit in VRAM. Often ends up slower. Also, tooling around AMD GPUs is still poor in comparison.

What are you talking about people do notice and care?! e.g. https://www.theverge.com/2023/9/7/23863258/bmw-cancel-heated...

Windows UI has gotten considerably worse for casual users who just want a web browser. There are so many ads and pop ups in the OS now.

I installed Ubuntu on my grandma's machine after she installed malware on her windows machine. I get less questions because it doesn't update with "features" constantly.

Actually it is physics and the market https://www.statista.com/statistics/220174/total-us-electric... a huge percentage of us electrical power is from natural gas. Natural gas to electricity to heat is a lot less efficient.

9to5mac.com ... I will hold my judgment for a unbiased review.

Same,fyi, chrome's --app combined with --user-data-dir is amazing for this without having to install a bunch of different electon apps.

They have a way to run thunderbird by running a local bridge. I don't really want their desktop client anyway. I do use their android app which is decent.

I only use Linux and I am happy with proton.

Wanted to play with typesense.

The tech behind building something like this isn't hard. Marketing and traffic is. No point in monetizing with no users.

It works fine. Just issue a HEAD request when you are unsure and rotate proxies a lot l. Takes a bit of infra but definitely possible.

Built the same thing a while back while collecting a lead list for sales. Not bothered to keep data updated but was a fun thing to build in a couple days. (disclaimer mobile experience is meh cause it was a fun project)

https://zensear.ch

How did you find list of all Shopify stores? I ended up just checking every .com, .net, etc as I didn't find an easy way to figure it out directly from shopify.

And meta doesn't? They released LLAMA2 and many other models.

Microsoft has a play, meta has a play and google has? A fear of failing? I am not sure....

My 10 year old Jenkins build still runs. My CircleCI build.... less so...

Stem != Job. Ignoring software, finding a traditional stem job directly out of college is very challenging. I know many mechanical engineering, bio enginers, chemical engineers who had/have a rough start.

I really wish people would stop equating stem to a job. Unless you are willing to move, it is not the case in many job markets.

Most medical degrees require clinical hours and demand means you can usally get a job wherever you do clinicals.

To give you ballpark I colo metal servers at a few different data centers. A gig up is a few hundred dollars a month.

Let's say you're really pushing the connection and your p95 is 900megabit up. That is $200 at colo vs ~$8200 for amazon.

I just run searxng. There is no reason to waste time checking multiple places.

I did find I used Google's widgets a lot and it took some time to break the habit but for most flows I don't have a first step of opening a browser anymore so actually ended up faster.

> With product/UX/new features it's the other way around, coasting is not possible.

That very much depends on the application. I maintain some enterprise solutions for customers. Clients get upset if UI has flow changes. Also, changes don't make me money unless they are required for a new customer. I will do it on request and invoice for the work but no one is interested in change for the most part. I think this is VERY common in enterprise software.

I would argue that a coasting grey beard could easily be better than a mid level dev with maybe 5 years experience.

I run a contract shop and all my best DB guys are greybeards... coasting. It doesn't matter, after so many years with postgres & MySQL they are amazing.

I used next 12 for a frontend with a Django backend and I was fairly pleased with it. Little slow on the build but tooling I could give to a fairly novice engineer and be pretty hands off.

Then recently, for another client, I tried nextjs 13. What a POS. A lot of the patterns I had learned were tossed replaced with something more complicated.

My first client is still happy but I will never use the stack again. Waste of time.

It isn't about the money. It is about getting the payment method. Getting tons of payment methods that are not obviously linked is not easy. Payment companies will often give up addresses, phones, etc. A bunch of information that can be used to connect accounts/bots.

Thanks! Ya, I opted for dual 3090 for my workstation (keeping full LLM in VRAM is crit) was wondering what lift was for M2.

OP implied that there were workloads where it out competes renting in terms of cost. Was hoping it was true for something than a single user interactive session (which can be done a lot cheaper)

Thanks for correction.

How many tokens a second? Really trying to figure out viability.

4x NVIDIA A100 at lamda labs is $4.40 an hour and I really have not had an issue getting them.

I find at AWS I am always wasting engineer time optimizing dumb things. Do you know how much a TB of ram costs? Or 10TB of blazing fast NVME costs? Less than $5K. How much does that cost at AWS?! This is not even considering bandwidth which AWS overcharges soo much. Yet I waste time.

Also, maintaining servers is not hard at a proper data center. It is often more hands off than the migrations cloud providers force on their customers.

With a 70B param model how many tokens/second?

Did the math and assuming 100% util and equal performance (which is certainly not the case) payback on your Mac is 9 months...