npodbielski

Really hard game for someone that is non native. Or maybe I am just at scrabbles?

Sure I understand though, I have honestly no idea what tauri is really. Sure there is a link and it looks like some kind of web framework. Looks nice but it is hard to picture how chat application written with that would look like. I am not trying to force you to do want I want. I just think it is in a good taste to put some screenshots or even a gif. But I understand your reluctance.

Interesting but could you put few screenshots there? Of both desktop and mobile? It is really hard to invest time into installing something that you cant see anywhere prior, and it will be really easy to do for someone that is using it daily. Sorry for complaining. Seems like nice project.

Ah mobile app is not ready yet. I am looking for some alternative to matrix because running it with bots is a bit convoluted, i.e. you have to have limit of edits of message for model streaming or you will kill entire room. Or I never seen robots in matrix sending encrypted messages. Why bother than? Anyway if mobile will be a thing this seems like perfect thing to have for your family and friends.

Ah past renewal... So probably in next contract there will be a lot of thinking if we still need that.

Curious: Anybody have comparison what is the difference when company is changing to token based billing? How many times it is bigger than 200$ subscription?

Strange. This is what they got at my company. I do not remember anyone mentioning paying for tokens. Maybe because it is fairly small, couple of hundred people in IT.

So just a mobile wrapper on agents running on some machines?

I added repetition penalty of two. I do not know. Maybe it is not applied correctly somehow by llama swap, that I am using, but I do not consider it that much of a nuisance to so I did not tried to fix it yet.

Do you have this DRY docs?

Yeah but 4 bits very often loops needlessly. Which is not that bad because you do not pay for tokens. But you paid for hardware and you want use it for something useful. Q6 is better but then you have like 40t/s prefill. Which is really tiring. But at least it says sorry when you ask it what is wrong! I heard there is some extension for PI preventing that. I need to look into it. Otherwise I am quite happy.

I like vibe and pi. Vibe just looks nice and is good enough. But pi extensibility is just another level. There is also Dirac that is quite OK but seems like full of bugs. Zerostack is the simplest harness I saw. OpenCode is OK too. Rest I did not try.

How? If something like that would happen robots would be disabled in 5 minutes.

What is wrong too? The plan to create a lot of cheap models? That it is what is happening but it is bad plan? Or the opinion that it is what is happening but it is wrong opinion because they do not plan to do that but i.e. China is just bigger with more people so just more companies are releasing more models?

Maybe both is happening a lot of choices will make more people go with Chinese models, which means less money to the US which means more problem to US companies. But at the same time they are cheaper and less of quality because this is China, this is what they does.

Personally I think this much more complicated as everything in life.

Interesting take. One question though: is it even possible for US companies? I am not so sure. Investors put a lot of money in those companies and they expect return. Doing nothing for 6-12 months waiting for China would probably make a lot of important people very unhappy.

And how does it relate to Medium?

What?

Sure I am not disputing that. But we can't say that z.ai is profitable because they charge the same as other providers per token. There is also training cost. And reaserch cost. Just inference is not all. Maybe there are profitable with all of it, I am not saying they are not, I do not know but at the same time I highly doubt that.

Unless somebody will find a way to actually do training by just updating weights with new data you have to train from scratch and this is very costly process.

Saying that some lab is profitable because they are profitable doing inference while disregarding training cost is not fair assessment.

And how you do know that GLM is not subsidised? Maybe Chinese government is burning money to destroy US economy by making them lose money in AI companies that are just furnace burning dollars. Unless you have some compelling argument it is just a speculation.

This would probably became very true if we would stay with the current prices of 'AI'.

Will we? Open AI is not profitable. Anthropic says that they may have profitable quarter. If they will raise prices will it still be the case? If you can 'vibe code your taxes app' but it will require constant fixes every month and those fixes will cost you 50$ in tokens and it will not be bullet proof, does it makes sense anymore? Maybe just pay 50$ for subscription to similar software? Maybe Chinese companies will keep low prices and it will cost 2$ dollars instead of 50, but that only works if you are doing that to 'vibe code' your scheduling/to do app. If you are any serious company you have regulations and GDPA and ISO and you cannot sent you financial and customer company to Chinese deepseek provider.

And software need constant upkeep. OS update, API changes, libraries get obsolete, build system does not work anymore... etc. This is very apparent for me every time I am doing changes to my Flutter mobile app need an update: I basically need to spin up environment from scratch, then update half of packages, then update all APIs fro those packages because of changes and when I finally do the change, pipeline breaks and I cannot sign the android release. Last time I just gave up on that. Non software people think you can just install Claude and prompt your app. Which is true. But then things break. Data disappear. You do not have backup. Licence changes and you can't use new version of some tool. Binaries got renamed. APIs disappear. Domain is not reachable anymore. And so on and on and on...

Software companies are forced to use 'AI' too so speed of breaking changes will increase and you either have to pickup on those or pay someone to do that for you regardless if this will be 'AI' provider and tokens usage or SaaS.

In 90 there were people in my country selling PC parts on every corner. No there is maybe one or two in entire city and I did not visit none in maybe 10 years. There were a thing because you could just buy parts and build your new system. You still can but now you can just order online.

So sure 'vibe coding' is a thing now but I am not convinced it will be a thing in 10-20 years. Maybe it will be online service that will automatically write an application for you based on specification for few $ but as a user you expecting an outcome and do not want to be bothered by npm and node version necessary for that.

Who they employ then? AI?

unsloth version works.

Touche!

I tried to run it but seems like it is either broken or it does not work on dockerized llama.cpp:

0.01.865.326 E llama_model_load: error loading model: missing tensor 'blk.40.attn_norm.weight'

Do you really believe that? If I would be employer or manager of developers saying that they need to carefully design whole system for the agent to implement, so not a technical person, it would look like their are buying their time.

I believe that it will be the case. 'What design?! Just feed Jira ticket and business documentation to Claude and quit stalling!!'

Only the Sith deal in absolutes

Maybe they have DeepSeek subscription for that occasion?

Me?

I am using Mistral Vibe.

480GB? Probably like 100k$ each? :D

Yep, I left one project because I think I was underpaid there. I could not find another job for like 3 months. I finally found one and I was also very stressful. No it is better, at the same project and I think I make more money than I would ever be able in a project I left. But on the other hand it was very stressful for my family. Was it worth it in the end? I do not know because even if money is better and company is bigger I could damage my kids development by being angry, sad and depressed for two years.

With manual ROM in switches probably not.