Even if it is a "pricier" PS5-like machine, I'd still buy it and I bet I'd make up the difference in less than a year with just the sales games (including older games I can't play on either console).

I think most of the critiques for this are from people expecting this to be aimed at PC gamers.

I don't think it is. I think it's aimed at people that actually DON'T want to bother with building, buying, upgrading PCs, but still want to play cheap games, older games.

To this day, I can't make my PC turn on with a controller (and I've tried). Making a PC wake up as fast as a Steam Deck from sleep? Impossible.

Those little things will all add up to make this a very nice option for the non-hardcode PC game crowd.

Valve is going to steal a lot of users from console, mostly Xbox. Not PC Gaming enthusiast.

However, for years the design has been going the completely opposite direction of what I expected. The focus on more random content instead of my own articles is not what I wanted to see.

Pocket is probably one of my oldest online accounts. I'll be sad to see it go but I guess it was already kind of dead for a few years now.

Amazing opportunity here for a really simple and focused read later app to take the reigns.

For me the key was just to see this huge change as a series of small steps instead of a big "flip the switch and change" move.

Before I changed, I took courses in programming to see if I'd like it, I build projects in my own time to see if I would actually pursue it.

Even once I decided to change careers, I still kept my old things ready in case I needed a fallback plan (this happened over 10 years ago but I still pay my Bar fees every year).

For me, doing it slowly, with a plan, and a backup, removed a lot of the pressure and risk of the change and it worked fine.

I haven't done anything in Law in over 10 years now and am fully "converted" to development.

My advice for you: buy some Cybersecurity courses on Udemy. They might not be perfect, but they are usually cheap (always wait for sales) and see if you will like it. You can take dozens of courses for 9-12 USD there to see if that is the path before committing to something more serious and expensive.

This is just to see if this is really the area you want. The day to day of most careers tend to be heavily romanticized from the outside.

Once you've done that, try to see if you can find some education with a Co-Op component. I found it is much easier to get your foot on the door of a new career as a Co-Op in a course, though that is not required.

It did happen in the past and was fixed after around 1 month (which was good because it decreased my time spent on YouTube).

Right now YouTube is almost unusable for me on Firefox with uBlock Origin on M1 Max and has been for the last couple of days for whatever reason. Even disabling uBlock Origin doesn't help.

Something is definitely going on between the two.

So my "copying current version changes" to Excel was kind of like git merging to the main branch.

People from outside the craft usually get super confused and frustrated by how many hours those contracts could take ("it is just a Word document!"), but that used to be a huge reason why.

Back in the day, to solve that (before I became a software developer and knew about git), I basically re-implemented "git in Microsoft Word version control".

On bigger contracts with dozens and even hundreds of pages, with more than 2 parties involved, we could have 3-4 parallel changes to the same documents at once. And a lot of times some of the parties didn't want to advertise their changes.

What I did was construct a table in Excel with each "version" of each clause, and where they were present omitted. It was a lot of work to maintain, but back then I didn't know any better, but worked.

By saving the versions of the files we sent for revision (think of the main branch), I could receive the version from other parties (kind of like feature branches), turn on Word changes, paste the complete content of the original, and we could visualize the difference.

It's been more than 10 years since I stopped dealing with contracts, but I imagine a lot of lawyers, specially older folks, are still either doing very convoluted processes to track changes like me or not even bothering most of the times.

The article is spot on on the issues (even brought back some memories). This is something I'd use for sure in my previous life.

Once we get going, I start asking how can we change the code to do what I need to do, etc.

After the history gets too long and Claude starts bugging me about limits, I ask it to summarize the context of the whole conversation, and add that to the Project and start a new chat.

After paying for ChatGPT and OpenAI API credits for a year, I switched to Claude when they launched Artifacts and never looked back.

Claude Sonnet 3.5 is already so good, specially at coding. I'm looking forward to testing the new version if it is, indeed, even better.

Sonnet 3.5 was a major leap forward for me personally, similar to the GPT-3.5 to GPT-4 bump back in the day.

Still makes a lot of mistakes, but it gets things "more right" than any of the others in a much more consistent basis.

I've now cancelled my ChatGPT subscription to Claude and also mostly stopped using the APIs (I use Msty to compare most models, you can give the same prompt to multiple models at once and compare the results).

Sonnet 3.5 is amazing.

Basically I enter the transactions and it shows a dashboard of my contribution rooms, how much is left, how much I have already contributed, etc.

Nothing fancy, but it just just a Remix frontend with MantineUI backed by an SQLite db inside Dropbox. Took me about 6 hours, and I only made it after I botched some changes I made into the Google Sheet that broke a bunch of formulas.

I thought about making it into a public app, but it is so tuned to what I want that it is probably not really that valuable to others.

I use it every time I save. Used to be weekly but lately monthly.

That is usually what I already do. Good to know I'm on the right path.

When possible I disable root login as well (though Coolify seems to need it on, even if without password).

It is actually amazing how fast and thorough the connection attempts happen as soon as you put anything online.

I've been playing around Hetzner and Coolify recently, and notice that, as soon as port 22 is opened, it is bombarded by those attempts. Several per second. It might be due to Hetzner IPs being reused, but happened to me every single time. Same with Postgres default port (those were the ones I've seen).

I have defaulted to use Terraform and bash to only open those ports in the Hetzner firewall (and more common ones like 3000 or 8000) to my own current ip. It does mean I'll get drift and need to reapply the Terraform code if I change ips, but seems to be at least one way to defend.

I fear that a lot of devs jumping into the "you only need a VPS" crowd on Twitter will end up with a huge attack surface on their apps and machines and most won't even know they are being targeted like that most of the time.

To this day I still find it hard to find a comprehensive security guide for those newer Linux fresh boxes (and the ones you find are all so very different with different suggestions). If anyone knows of a good one, please share with me!

Agile, testing, design patterns, best practices can all tank and bury a project if applied excessively "by the books" without consideration of the actual problem to solve.

I've worked in teams that had about 10 people actual doing dev work that implemented the full suite of Agile "principles" as rules. Daily standups, grooming, retros, pointing as "poker", 1:1s every week. The result was that we had barely time to actual work since the week had 10-20 hours of meetings. Most retros and standups were literally just us saying "same as yesterday, only had a few minutes to work on this" the whole week.

Testing is the same. If applied without consideration for the actual problems, reaching that 90%+ code coverage is easy if nobody cares about how hard and time consuming it will be to change code later. Specially when a feature is in very early development.

I think all those things are good, but what I see sometimes is that they are applied as absolute rules that cannot be deviated from, which inevitably leads to poor results.

I'm now working in a "light Agile" environment with just 2-3 meetings a week, barely 1 hour total, and much less strict PR/testing requirements (we focus on testing the important functionality, not line coverage count) and it is so much better. Some of the same co-workers that were under the more strict rules are now twice or more more productive then before.

Instead of trying to do everything into a single chat or chain, add steps to ask the LLM to break down the next tasks, with context, and store that into SQLite or something. Then start new chats/chains on each of those tasks.

Then just loop them back into LLM.

I find that long chats or chains just confuse most models and we start seeing gibberish.

Right now I'm favoring something like:

"We're going to do task {task}. The current situation and context is {context}.

Break down what individual steps we need to perform to achieve {goal} and output these steps with their necessary context as {standard_task_json}. If the output is already enough to satisfy {goal}, just output the result as text."

I find that leaving everything to LLM in a sequence is not as effective as using LLM to break things down and having a DB and code logic to support the development of more complex outcomes.

All 3, sadly, have seen the quality nosedive in recent years.

Today I mostly rely on Udemy. I know it has a bad rep and tons of garbage, but there are a handful of instructor that deliver really high quality content there. For example, see Fred Baptiste Python courses.

Even with AI taking charge, I still believe there is a place for a high quality, guided course like PluralSight used to offer back in the day.

I don't trust Google with barely anything these days anymore (except Gmail just because it has been so long, and Maps), but Google Meet is the one thing that I prefer Google's solution over anyone else's.

Meet is just so much better than Zoom, Teams, FaceTime, WhatsApp Video, etc.

I'm so glad they are tackling this specific issue. Pretty amazing feat if it works well.

But anything more complex and it is very hit or miss. I'm trying now to use GPT-4 Turbo to write some integration tests for some Go code that talks to the database and it is mostly a disaster.

It will constantly mock things that I want tested, and write useless tests that do basically nothing because either everything is mocked or the setup is not complete.

I'm settling in using it for tests for those small, pure functions, and more using it as a guide to find possible bugs / edge cases in more complex cases, then writing the tests myself and asking it in another prompt if they would cover those cases.

As most people that actually use AI heavily these days, I think the usefulness of AI for coding increases a lot if you already have a pretty good grasp of the subject and the problem space you are working on. If you already know roughly what you want and how to ask, they can be a huge time saver on the smaller and simpler things.

"Hi ! I'm trying to buy something in store X but my card is getting declined, can you make a PIX (Brazil's real time rail payment) for for YYYYY (4, 5 digits amount) reais for me? I'll pay you tomorrow"

Or the out of fashion "This is X and I have your daughter, pay X so we can release her".

I have never understood how those people can do this for that long with the amount of tracing and spying the government has. Specially using payments from the Real Time Reail (PIX) that is tied to a CPF (equivalent to SIN/SSN).

Even in the current state of streaming, with sports spread between several providers, you can get, in good faith, in YouTube watching a pirated live stream of a major competition where the broadcaster fakes a famous bank or financial institution pretending to make lottery and faking draws on screen. It is so rudimentary you can see the person literally typing random names.

This is probably one of the easiest crimes to automate enforcement against in the age of AI, and yet, it happens for months, every day.

Good for Spain for actually doing something about it. I wish Brazil would start doing the same.