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 ;)

For a single tab? I wonder how much the telemetry takes up...

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 :)

my two best friends from BA and MA moved away, so I can clearly explain why someone might not have friends from university time (-:

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

but if you put into perspective what it would cost you a month for a graduate's salary - and you compare that to the max plan - the value for money is still astronomically incomparable

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 :(

Don't vibe code your critical business logic and infra.

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.

> AI amplifies. It amplifies the success of the good professionals and amplifies the failures of the bad ones.

It also worsens the problem in general by making it way, WAY easier for the bad ones to performatively appear good. They'll have the better-sounding promises but if you listen to them you'll crash and burn in a few years. This doesn't even have to be intentional, just someone technically ignorant channeling AI sycophancy while simultaneously playing politics (i.e. promotionmaxxing while delegating ideas to AI) will have the problematic effect.