Eli Lap

Opensource sub-Reddit banned all AI commitment. Even projects, that can't be explained due to lack of knowledge of English language and therefore written with AI.

Putting AI on top of the old scenarios, software and processes amplifies their current state. Slow processes gets slower, bloated software will become slower and less friendly.

The same with the processes. If it's you always making a decision - you'll become an even bigger bottleneck. Overworked, burned out, outdated.

Opensource sub-Reddit banned all AI commitment. Even projects, that can't be explained due to lack of knowledge of English language and therefore written with AI.

And I'm noticing how challenging it is for me. The old way of building is ineffective. Straightforward vibe-coding leads to a meaningless output. Like I've bought a car that has nothing inside. It's new and shiny outside. While in reality it's only a body.

And putting things inside it is not so far, so effective and so great. Often leading to decision fatigue.

Opensource sub-Reddit banned all AI commitment. Even projects, that can't be explained due to lack of knowledge of English language and therefore written with AI.

I see it as a signal smart people being fed-up with wannabies paired with AI. Getting an output means nothing. So they decided to pull a trigger to keep things the way they are.

I expect it might take a few more years before they get back to the same point of the cycle and re-validate this decision. But for now it's there.

MiniMax M3.1. Flash has been released this week. These are still the most affordable cloud models and plans I've seen so far. While published test results are looking promising - I'm migrating immediately to see how it performs in real life.

"I don't need developers, claude/gpt/whatever one shot this" might be fun game, but a poor strategy. I spent last week vibe coding dashboard, to make use of an old iPad Air 2. The goal is to keep it as much responsible as possible. It took 5 nights to get so far.

And during the week I haven't seen the code yet. I skipped digging the CalDav and refining JSON only because these are established protocols. By the way I am using my own checklist spec for reminders, that's why I'm sure it will do what I want.

And at this point I realised I don't care about the future of my career. If products are gonna be built putting 0 effort and heart - I don't want to be a part of it anyways. Happy weekend!

Considering UI, reviewing engines for reminders and calendars. Testing functionality and validating assumptions.

After all it feels like autonomous AI agents would make much better decisions. I have no idea how we are supposed to believe in humanity in this reality. It is impossible to make any agreements with people, especially when money is involved.

While the world patches PHP I extended my homelab. My earlier stack choices gave me options I wasn't locked out of. Your stack shapes what you can even imagine building. Read: https://read.elilap.dev/your-stack-matters-more-than-you-think/

I’m starting to suspect PHP is not just a programming language anymore. There are haters and fanatics. Haters make fun of it and avoid it. Fanatics rush to prove that other languages are just as bad, especially when it comes to security. But that’s not the point.

Not necessarily because PHP is bad. The same could happen with any technology. But because PHP became a default choice. Not always because it was the best, but because it was the easiest. And when something becomes the default, we tend to stop asking whether it’s the right choice.

Don’t get me wrong. The easiest path is often the right one. I’ve wasted plenty of time and effort by refusing to take it. But that’s me. I’ve always been suspicious of the easy path. And I intend to stay that way.

PHP is old. It has limitations. There are workarounds for almost everything, including problems the language was never designed to solve. And that’s fine. Until the workarounds become the norm, and nobody questions them anymore. Now a huge part of the web shares the same risks.

PHP websites are going down. Keep vibecoding and you'll figure out how to get back in time into the pre-PC era.

Oh boy! Entire PHP world is buzzing because of the vulnerabilities that allow remote code execution. The type of vulnerability that really opens up a gate to your system.

If you are running legacy PHP systems and saying you're adopting AI and moving off PHP is not on your roadmap - you have no technical roadmap.

Keep adding features. Keep pretending you're smarter than others. PHP is a big target, and today a shot was fired.

Oh boy! Entire PHP world is buzzing because of the vulnerabilities that allow remote code execution. The type of vulnerability that really opens up a gate to your system.

I've been working with PHP for years. I don't hate it. But I have never chosen it for my projects. Because it is running on inertia. 70% or so of Internet is built on PHP. But it proves nothing, except being a big target.

When I was 18, I spent days on car forums. Not because I loved fancy cars. I drove an old Audi and was tired of paying mechanics who replaced expensive parts without actually solving the problem. So I started learning how things worked.

Why spend hours investigating when you can generate another fix in seconds? I’m not against AI. It’s an incredible tool for learning.

But if we use it to avoid understanding rather than accelerate it, we’re creating a bigger problem. The real engineering skill isn’t producing solutions. It’s knowing why they work.

AI didn’t invent superficial problem-solving. I’ve seen developers patch bugs without understanding or testing them long before AI appeared.

AI simply made generating solutions dramatically cheaper than understanding and verifying them. And that’s where our incentives become dangerous.

Recently, Sean Allen announced he’s closing his iOS education business as revenue has dropped almost to zero. It made me think about how we learn today.

We pay for AI tokens, generate fixes, and sometimes spend hours trying solutions that don’t solve the actual problem. Sounds familiar?

We're in a bubble. A video on reverse engineered block game confirmed my worst expectations. A financial success doesn't come from anything else, as calculations and what in my opinion are manipulations. What's it about?

It's not a game. It's a sophisticated manipulation mechanism, where results depend on a myriad of a hidden params. It's not about skill, strategy or understanding.

The problem is not the game. The problem is even casual gaming depends on a hidden rule. How long are you going to pretend our society was not tricked and next generations are "demanding"? Changing the rules in the middle of the game is cheating, not flexibility.

A game is selling ads, and each player is valued against a set of rules, like device model, settings and behavior regarding the ads. It has a few algorithms to assert when to make you lose to show you an ad.

When will we start talking AI and hiring is not about the money. It is about to show who has the power. My point is backed up by myriad of stories how more expensive decisions were made and executed, because it was protecting someone's ego.

Ignoring the fact Software Developer were learning while coding is why current market is a noise, not a discussion.

Now with AI we pretend machine can do the full cycle. It can. But without getting in touch with it we are useless. And the output will be useless as well.

If you haven't learned anything from the coding session - you did it wrong. That's the entry point to make AI-augmented coding both: effective and useful.

Ignoring the fact Software Developer were learning while coding is why current market is a noise, not a discussion.

Writing the same code again and again was long and boring. This is why we were learning and introducing libs and later frameworks. But even then - we had to learn how to use it properly.

Everyone is exhausted. You're happy, because it's Friday. How will you feel on Sunday night? Those working these days are exhausted by AI. Those who are not working - are exhausted with battling AI to find a job.

Algorithm interviews makes no sense these days. But how do we know output won't be a silently accepted slope? We don't until we learn who they are. We don't need advanced tech for it, we need empathy & humanity.

Software was always an IP.

I think Open Source is the future. And personalized software is where we are heading to. Someone can spend a month and create a great service for 10 neighbors. Others will be able to design reliable and scalable service.

What are these services? Everyone's trying to understand. Should we have a service to understand problems worth to solve?

Software was always an IP.

There were 2 main barriers: resources to build software, time to find proper algorithms.

Now we have higher-level algorithms: LLMs. So time and algorithms are not unique. With reverse engineering - you don't even need to know what to build.

Another meetup. Smart guys, years of experience. The same "nobody knows what the future brings". At first it makes me anxious. But having a second thought made me happy. We're at the moment where the smartest move is stop pretending future is known.