maciekpaprocki

finally, after two years of hype I have some usage for llm.

we import descriptions of products from a seller. the problem is they are mental ( probably written by chatgpt :)) and are way too long. we need only small blurb.

I put our style guide and given text to chatgpt and I get somehow reasonable description back. then editors still need to check it, but it's way less work.

Lol. Most of the open-source communities will have a nazi and a fury working in tandem :)

You run a community it's up to you what standards you enforce. And there's plenty of other projects you can contribute to if you disagree with them. You can also fork things and create your community.

Yes, some organisations take a piss, but at the same time, you have loads of other options.

Additionally, except for anecdotal cases, you can submit PR, which will be reviewed and merged no matter your political stance.

SEO very much does count. And on site optimisation is the easiest return on investment as it's generally not a lot of work. few hours of work can bring you tons of clients.

You dont want to exclude your images. That can very much affect your results as it will remove you from image tab, but also content of articles that contain them might be affected.

For those that loved the book. Hasek also wrote an autobiography, which is even funnier than the life of Svejk. It's definitely made up quite a bit, but the stories in it are great. I cannot find english title ( it might have been not printed ), but transleting from polish it's called "3.5 beer".

Spoilers!!! Stories include:

- One time he wrote to paper called animals and unfortunately run out of animals to write, so he just started inventing them. It ended up in year long trip to Australia paid by some wealthy women who really wanted to get her own "Tasmanian Vampire"

- Russian general invades city. Hasek knows the best way to stay alive is to have a drink with him. They get very drunk and general ask Hasek what's his biggest problem. Hasek says that the fact that people cannot read. Next day he wakes up to bunch of posters around the city saying that whoever will not be able to read by noon tomorrow will be executed.

not diss on OP more on SF Chronicle

How is that newsworthy? Some poor kid died and his mum is a retired CEO of YouTube. Just seems like invasion of privacy in quite terrible moment for the family.

If I would tell any of my tutors/teachers that I am not able to find time for it, I would be thrown out of the school or at least didn't pass. I am like 30, so i don't think that much changed.

The f*** is that? Some crazy scam on hn.

That's a security feature. If you would have that in any environment it would be massive security issue. If you actually own the service you can still do that.

wait. why not iframes?

This article is a bit bs. the first example of how much code it requires is misleading. you can accomplish this same thing with exactly this same amount of html nodes. You just gonna have more classes.

Calling tailwind tightly coupled Vs semantic loosely coupled is also a bit of strange. I would call it composition over inheritance.

Just finished the project like that.

The website was terrible. Mixed encodings which messed everything up. all the routing in htaccess going to hundreds separate files. PHP 5.4, no version control and pretty much everything wrong that could be.

Company had a hot potato that was generating quite a few milion pounds. They actually didn't had a programmer for a year, but because of pandemic fortunately old one decided to come back.

I had quite a lot of trust from owners as I developed two other side projects with them already, in crazy times. they knew they had to invest in technology or die and they said they want to do that.

We had to start from scratch, there was nothing salvageable there and files where 100k lines long with no explanation which one is run anywhere. I wanted to replace piece by piece, but because db structure and encoding, I really couldn't find a way. whatever we would end up with, wouldn't be decent thing, it would just be a frankenstein that then would have to be rewritten again ( although drastically better ).

I told owners how much it takes to develop similar project and said that this is not estimate as no one would know how long it will take ( they tried to redevelop this twice and didn't manage to ).

The project struggled with a lot of issues. The other developer couldn't really contribute anything even that I tried to pull him to new code. He ended up taking over product owner job on my advice as he was actually really useful to company, just not as coder. We couldn't find anyone to work even that we paid pretty well and allowed people from any place in the world. We found two developers which were pretty skilled, but seriously didn't do anything. I often deal with that, but in such a small project it's just killing any productivity, including mine.

We managed to publish the project with large delay. It ended up being super rushed, but we generally managed to sort out most of the issues quite quickly.

We failed on one thing though. SEO. Even that the site increased in all the stats in webmaster tools, the reindexing wasn't kicking in 5 months later. We hired SEO agency, but frankly they didn't help anything. The issue had nothing to do with new site. Just we were in google bad graces with previous site and google just ignored our new links while removing old ones. I knew that will be the case with new site, but the benefits should drastically out-weight the cons.

At this stage company literally refused to pay me my shares and some money they did owe me ( they were broke waiting for new round of funding so I gave them a bit of leeway ). I had to stop working and I am suing them now.

The moral of the story is:

- Everything will take way longer than you expect.

You cannot divide and conquer such a large project. Any amount of planning outside of basic one will be just a waste of time as no one will be aware of all the features, some features are lacking and some are just plainly stupid. Any scope will change thousand times. Rewriting partially is way better, but it wasn't possible in my case.

- Business will say they want to invest the money, but they don't understand tech.

All the time, the solution to too slow progress was hiring more devs. Man hours are almost never a solution. Time is way more important investment. There also have to be contingency, as something will go wrong. In my case SEO issue will be solved, but it might take 3, 6 or 12 months, in which time the business will have to loose some sales.

- Communication with business is very hard.

At this same time, you need to explain the stuff will go wrong ( unless you have unlimited amount of time and resources ) and will be delayed. In this same time you want them to invest money in it. Frankly I failed on that the most. What I would make very clear now, is that those issues are caused by lack of investment over the years.

- Only go to it with good team.

I managed to build a very good team at the end, but it took a lot of bad apples to get there that wasted a lot of my time. People had the skills, but half of them tried to rewrite every piece of code not developing anything useful and the other half did just not do anything for weeks.

My view on that is: Unless business understand the need to change, have the money and time to do it, and you have a good team, don't do it. Seems like you are 0 to 4.

Some businesses cannot be saved. It's their fault they didn't invest any money over the years and if you want to do tech, you need to have tech people in management or on board seats.

Ok. I would kind of like it, but environmental impact of that needs to be insane. I guess 50 years ago people were not that concerned about it. And then how much energy they will have to spend just removing sand. Seems like a pretty pointless job for someone.