sminchev

What currently happens looks really scary. The level of technical skills needed to make and deploy app in production dropped a lot. Everybody are trying to make something. This can be good and bad. Good because increasing the level of concurrency should lead to better end results. At the same time, a lot of people, not knowing what they are doing, they create AI slop. Of course, they have good intentions, and do their best, just the quantity of low-quality apps becomes too big.

Open, for example, Reddit, and check the SideProject channel. I counted 1 application per 2 minutes! This is huge! And a lot of it is bad quality, because people have no technical and UI/UX skills. They don't know what architecture, MVC, code review, retrospective, technical documentation, target group, performance, hashing of password, and a lot a lot of things really EXISTS.

Even if you make something cool and stable, you just can't popup! And that's the current state. Too much noise!

I am working as contractor for the last 15-16 years. And each new project starts with interview. I have been on around 20+ . Most of the time, I just read again the most basic staff like, what is class, interface, overriding, overloading. Just to remind myself what's the proper/modern terminology that the person in front of me would like to hear. To use it, does not mean that you can explain it, and this is what they are searching for.

All else is confidence, experience, nice professional stories, curiosity, good soft skills. People need to like you as a person, to feel unconsciously that working with you will be safe, cool, fun, productive process.

I will :) Sometimes, I really need someone to tell me that it makes sense. I hit so many walls trying to make this idea publicly visible, that I almost become desperate. Most of the time, I am just treated as 'one more intruder'.

I will continue to push forward :)

I guess it is related to the model as well? Is it Sonnet or Opus? Sometimes the old Opus 4.6 gives more appropriate results.

I am working on two things in parallel. https://howareu.app/ is the first one. Of course, struggling with the making it publicly known and trusted. ;)

And the other thing I just started is a MCP server that will connect to Java remote debugging entry point. This way I will try to allow Claude to start debugging session by its own. And I hope that in pure back-end logics, it will easily trace and fix bugs. I now do it with log messages, but it is kind of annoying and dirty and error-prone to do it that way. I would like to try an approach that does not mess up the code so much. Let's see how it will go :)

This is always the struggle. To implement it is not like promote it. Making the product publicly visible and known is actually the harder part.

Be consistent, try on multiple places, be active, make good site, good videos, good training, talk to people, listen what they say. Post it in reddit, write articles in places like here, dev.to, medium, explain how you did it, why, what you learned...

Cases where somebody implements something and it becomes viral for a one night sleep are so rare, that I am not really sure that they actually exists ;)

Sometimes, I get bored. Writing code was so much fun. Trying to figure out, how to resolve the issue, and they feel the excitement, the joy when the task is done. This is now gone. I case of issue, ask the AI, tell the AI to write it. Where IS the JOY?! Some people find the joy in making the product that they were not being able to do, because of lack of the needed skills, and this is not bad. :) But for me, with my 20 years experience as back-end Java developer, I sometimes feel this big gap.

AI can produce garbage, it can produce excellent results. As long as you know what you are doing, and provide perfect specification, it can come with results, that you alone can't think of, or they can take a lot of time for 'manual' research and implementation. Some people still find issues, and don't trust, but the topic is big here: what's the model, how good is the specification, what's the used process/workflow, what are the agents, what's the technical background of the person leading the agents. Not that easy to say: AI does not work, why people want to use it after all, when it produces garbage

Life changes, customer demand it. We have no choice, and we need to adapt. Of course, there will be places that code will still be written manually. Banking, military, and other highly confidential areas have no other way to do it.

How did people feel in the beginning of the industry revolution? Have they felt the same way back then ?

At 40+ , especially if you have kids and other responsibilities, just going back t school is really, really hard. I may also say, impossible. I tried to learn German a few years ago. My brain does not work the same way, as when I was before 36.

I read it in a report, and I wrote it here in HN many times in the last few weeks: AI amplifies... It amplifies the success of the good professionals, and it amplifies the failure of the bad ones.

Good professionals are needed so that AI is used the proper way. I think that the way we do our job will change, but there will also be place for developers, PMs, POs, Team leaders, etc.

The problem is not the AI/LLMs technologies. It is the greed. Earn more money, and the money to be more important than the people. It is not a capitalism thing. It is bad culture, attitude, and egoism, IMHO.

Everything is based on the requirements and available resources. One of our clients decided that calling the AI so often takes time, and money, and this does not work for him.

AI can give suggestions, not decisions. IF you want decisions and responsibility to be taken, use real people.

There is Javascript interpreter, caches and other memory consuming things ;)

I would be happy to see some prove that this works. A project that started and went go. How a friendlier specification looks like and why agents understand it better.

This is cool. The plan written as algorithm. Pro-activity is the key. Usually, people like to stay in their comfort zone. This guy was searching for his, and found it.

I wonder, why he did not have any friends from the years of studying. Usually, this is the place friendship forever happen :)

I am happy for him :)

If you love it, do it. It is not that easy to find a job as before, but if you are good and pro-active, and with passion, money will come... no matter what you actually do.

You have good approaches. I created a slash commands that I run, and explicitly wrote to be executed with Sonnet. I also compact manually, and try to keep the session short.

But let's look at the big picture. When using a tool, it must make our lives easier and more convenient. To stop troubling our minds with redundant things. Our focus must be on the product and its quality, and features, and don't waste energy on things outside the scope. What's the point of using a product that pushes us to think about redundant things.

The whole point of using Claude Code just disappeared. I am thinking more on tokens, limits, and optimizations, making compromises on the tasks, rather than product vision and requirements.

And this is a pity. Total waste of time and energy

Anthropic's whole subscription plans, models, performances, and token usage are lately a big issue for everybody.

They started so good, now a lot of people are pissed off. You are the next, unfortunately :(

Is it really possible to review and choose a product that works for you these days?! :)

I read it in a report: AI amplifies. It amplifies the success of the good professionals and amplifies the failures of the bad ones.

In all cases, whole enterprise solution can't be made with pure vibecoding. Specification is needed, a basis of predefined rules, coding styles, security considerations.

For development Android can be a nightmare. Especially, if you try to make a long-running app. Each OEM (Original Equipment Manufacturer) like Samsung, Honor, Xiaomi, etc has its way to keep the OS up and running, in good shape, performance and with low battery usage. To do this, they have mechanisms that stop applications that they might think are redundant: - if you use too much battery - they might kill it; - if you use too much CPU - they might kill it; - At night the devices goes to 'doze mode' - they might kill it; - It the app is not opened for a long time - they might kill it;

Some of the 'optimizations' you can disable/enable. Each device has its own menu and naming. But not all of them. And even if you make all configurations good, at some point it is possible an update of the OS to reset your configurations silently.

I don't know how it is in iOS, but in android, long running apps are not something that is 100% doable.

After I struggle for months with my app, I have my preferences for iOS right now. It might be difficult as well, but at least the manufacturer is only one. ;)

gmail :) Not kidding, Most of the time, when I want to send something from my device to my laptop, I share it with an email. In case of friends, directly in Viber, and Teams.

Yes. When things get too complex Sonnet misses some things. For example, it creates all the components, but does not link them. Or it does not go deep enough in the code and misses certain usages and possible regressions. In other words, it does not, pro-actively, search for things that I have forgotten to tell the model about.

I don't know about the medicine. I have not been in any medicine related project, so far.

Regarding the banks, they can't use AI. This is a big security risk for them. They need to keep really, really high level of confidentiality. They need to cover all kind of audits: internal, government, external, etc. No audit will pass, if any of the documentation and implementation go to the cloud.

Highly regulated!

I guess, if the medicine follows the same high level of confidentiality, AI won't be used there as well.

No AI, no push to use it -> normal expectations of time and effort ;)

I fully agree with the garbage. It is extremely difficult to release a product these days. Everybody are trying to vibecode something, without any knowledge of basic software development, validation, reviews, specifications. Even if there is something good, and written with focus on the details, it just can't popup and be seen, because the overproduction of software.

This is like saying: nobody can do woodworking with with manual tools anymore these days, because there are machines since 30 years that do woodworking.

There will be always a room for good developers.

Banking software. No place to hurry there.

Also, changing a job is two-side process. they need to like you and you need to like them. If you find that too many developers are going out, too many job offers from their side, that might mean that the company is toxic. Stable good companies, don't have a lot of job offers. They are not so aggressive searching for people ;)

Always ask why they hire, try to find out why people leave.

There are enough history novels out there ;)

One of the reasons the AI fails constantly is that it has no context of the entire code base.

I had the same issue, take your time and describe the context. Create a documentation out of the project. I used BMAD for this. It created a specification for me, by checking the code I answered a lot of questions, clarified a lot of things. And once I had this done, I now start each session with: read the documents, prepare yourself for implementation, have a good context so that you don't make any errors.

Afterwards, I do my things, fix, test. At the end I do an update of the documentations: update lessons learned, add new rules, if necessary, update architecture, etc.

With the time, I got enough basis so that most of the fixes happen with only 2-3 iterations, which is good enough for me.

I use Opus.

Sometimes, managers are lazy and don't work, and they think that the employees don't work as well.

And something it is pure bureaucracy. I once worked for a bank, as a contractor. A team in the bank responsible for the auditing ask my manager in the bank, how can he be sure that I worked today, and for the money spent I actually do something meaningful. He had no written prove that I was at work. Yes, I push every day, yes, I am on the daily meeting with my team everyday, but no official document that a non-technical auditing team can understand.

The solution?! The most ridiculous solution ever! Every morning I was writing a letter to a secretary, and she was adding a check in an excel sheet. And this was the official document!

All happy, except me, because, I guess I was the only one thinking that this does not prove anything.

And another example, from 15 years ago. I was, again a contractor, but this time in an office. They were checking when I check in for work and when I check out. And one day they saw that I worked 7 hours and 45 minutes, instead of 8 hours, and they did not want to pay those 15 minutes. Again I can literally, go in the office, stay there 8 hours, walk the stairs all day, and do nothing.... ;) Of course I am not that kind of professional, but the point is that, sometimes, the management request does not make sense and create more burden and pressure, rather than solutions ;)

Now it is the same as the remote work. I can be at the office whole day, being seen, and at the some time I can do nothing :D

Reading a book, I do for fun. to get the stress away. Reading a novel is a whole new experience. To feel the story, imagine the characters, imagine how they feel, how they think. To really hate the bad guy....

AI can't replace that!

Even changing the technologies can be really hard. I usually use Java and GoLang. I decided to do an Android app with Kotlin. I hit all the Android/Kotlin specifics. Because of my technical skills, I managed to overcome most of the problems, but I did not catch them on time. If I knew that they existed beforehand, I would have saved a lot of tokens, time.. And probably, I would have approach the whole problem differently.

Having that said, non-developer can produce commercial products, but the learning curve will be so steep, and the expense so much, that probably does not make sense to do it.