Vadym Kov

How do you make sure AI does the right thing… when it writes your features, tests, and infra? Before you check it all manually

One pattern I keep seeing while doing delivery assessments: teams can’t optimize processes they haven’t mapped visually. Once the workflow is drawn end-to-end - bottlenecks, unnecessary handoffs, and automation opportunities become much easier to spot.

@bridgemindai Yes, if we use it for car wash

Claude Opus 4.7 with max intelligence inside of Claude Code still does not pass the car wash test. Are we cooked?

Thank you @claudeai $100 for 2 hours. Nice. @OpenAI - your time

@jgreze too much winning?

I am having an argument with my team about whether winning is a unit of work.

@icanvardar I like how others try to support claude structure (skills, agents, etc). So it will be easier for people to switch

stop using claude code before you end up vendor locked and incapable of thinking without it

@Amank1412 while waiting for an LLM - I am writing another prompt in another LLM terminal session 😃

Someone built a transparent Mario game that runs OVER IDE so can play while waiting for Copilot to write code.

@asaio87 Can we call software engineers "vibecoders"?

I just finished vibecoding a Saas app we pay $99/mo for. It took me just 5 full working days. I worked 14 hours a day for just 5 days but it was all worth it. It has a few bugs I still cannot get rid of, but I am going to spend a few weekends fixing them. My family will understand. And the cost of running it is just $75/mo in AI tokens But its so worth it vibecoding is pure awesomeness and saas is dead... almost forgot

Assuming we’ll have more and more AI systems, consider creating products for AI systems, not just for people.

@hiiinternet AI Operator

The future of all jobs is everyone will become a software engineer

@NoahKingJr Think about creating a product that can be used by other AI products (not people)

Will AI create more jobs than it destroys?

So why all the above can’t work with another step up the semantic expression ladder? We had it already multiple times in our lives, and we were ok with it. And this is what @unclebobmartin trying to say, from what I understand. And he clearly said in some other posts, that we should own structure, architecture, tests, etc. For me it seems like we are talking about a little different things.

@vadym_80 @unclebobmartin Accountability and ethics are the answer to all the questions. It is not a matter of being better vs not better, more reliable vs less reliable. Engineering goes beyond typing words. Engineering encompasses things which can’t fit the “I don’t review” mindset

AIs are just another step up the semantic expression ladder. We initially expressed our semantics in binary, then assembler, then Fortran, then C, then Java, then Python, etc. AI is just the next step up that same old ladder. And when you take that step, nothing else changes. You are still expressing behavioral semantics. You still need to express structural semantics. All the old principles still apply. You still have to be concerned about design and architecture. And even though the syntax allows informal statement, you cannot abandon formalism. When you express behavior you need a formal way to enforce the behavior you want. I use Gherkin for this. It seems to work pretty well. Consider that Gherkin is written in triplets of Given/When/Then. Each of those GWT triplets is a transition of a state machine. A full suite of Gherkin triplets is a formal description of the finite state machine that represents the behavior of the application. Other formalisms that matter are things like module dependency graphs, testing constraints, complexity constraints, and many others. This step up the semantic expression ladder provides you with an enormous amount of options. But you'd better choose those options wisely!

@EdmilsomCarlos @unclebobmartin It seems like you might be making assumptions about what we think regarding engineering, which could be a form of cognitive bias. Unfortunately, I didn’t receive answers to my questions

@vadym_80 @unclebobmartin You guys think engineering is just typing a bunch of words on a specific sequence using a keyboard - that’s a fundamental misconception

@EdmilsomCarlos @unclebobmartin do you trust it more? Why? Why do you think some John will have enough coding knowledge according to your expectations? why do you think some John has enough knowledge and it is better write in some Python/Java, but not C or assembler?

@vadym_80 @unclebobmartin Are you you putting human knowledge and accountability to check?

Claude 12 green bars out of 30. which is 40% @trq212 😕

@ThePrimeagen and fix GPT, it is also down

mythos, opus is down, fix immediately, no mistakes, ultrathink ultraplan

@EdmilsomCarlos @unclebobmartin Why do you trust some John (you even don’t know) who wrote the code manually?

@unclebobmartin @unclebobmartin would you fly on an airplane whose navigation software was never reviewed by engjneers, but entirely done using AI via Gherkin instead? Would you trust your family life on a medical machine whose telemetry software was also done the same way?

@NoahKingJr skype

Which app has the most retards?

@icanvardar Endless debate

i think it’s time we all admit it codex is just better

Have you calculated how much you’ve already spent on LLM subscriptions/tokens vs cost to run your own LLM? Would it be significantly worse than continuing to pay for subscriptions? Is there somewhere you could sacrifice quality in the long run?

so true 🤣