One thing about RAG that looks simple on paper but gets messy in real projects.
When an answer is wrong, the LLM isn’t always the problem.
Sometimes the right document was never retrieved.
Sometimes the document was right, but the useful part got lost during chunking.
Sometimes the context is good, but the model still makes things up.
sometimes the information simply doesn’t exist in your knowledge base.
You can’t fix all of these with a better prompt.
You need to know where the pipeline is failing before changing anything.
That’s the real work behind a reliable RAG system.
If your RAG app gives a wrong answer, what’s the first thing you check?
RAG is easy to explain.
LLM + vector DB + retrieval + prompt.
The fun starts when you put it into a real system.
Bad retrieval.
Wrong context.
Hallucinations.
Latency.
Cost.
That’s where the real engineering begins. 👀
On your AI learning journey, you must learn how to fine-tune small models.
Fine-tuning teaches you: data quality, overfitting, evaluation, and deployment.
I have said it once, and I'll say it again.
The key to grow on X is posting consistently, on top of that is to engage with other's posts.
Show love to others. You're not gonna loose anything. What you give, comes back to you.
Have a blessed day y'all. 🤝
Sometimes you spend 30 mins writing a post.
You share something you learned, an idea you had, or even a small win.
And then… nothing.
No likes. No comments. No conversation.
I think we can do better than that.
If you see someone’s post, take a second to engage. A like, a comment, even a different point of view.
I’ll do the same for you.
Because we’re not here just to collect followers.
We’re here to learn from each other, help each other and grow together.
A small interaction can mean a lot to the person on the other side. 🤝
Press conference for builders with tiny followings and great products. 🎤
Reply with what you’re building + a link.
Then follow the 3 people above you. Big accounts won’t amplify you. We’ll amplify each other. 👇
I’m building slowly.studio, a creative technology studio focused on building AI products, digital experiences, and experimental technology. slowly.studio
📊 To all colleagues and mutual connections who regularly engage with my posts: please participate in the poll above.
I am well-acquainted with your profiles, content structure, and engagement patterns, as I frequently interact with your pages.
The group being established aims to significantly elevate your current reach and performance without requiring changes to your established workflow.
As this initiative is designed for your benefit, your participation is strongly encouraged. Ultimately, assistance cannot be imposed on those who do not actively seek it, and individual objectives are fully respected.
🌍 Furthermore, language should not be a barrier; although English is the primary medium of communication, you are welcome to submit questions and feedback in your native language. Measures will be taken to ensure clarity for all participants.
Thank you. 🙏
✅ Please consider this invitation, complete the poll, and submit your request via direct message.
imagine spending 3 years building an app.
someone downloads it, gives it to AI, and says “figure out how they built this.”
and AI starts doing exactly that.
i’m starting to miss when AI just made ugly pictures🤣🤣
github.com/morluto/rea
53% of reused AI agent skills are never modified after someone copies them.
A study of 18,463 skills and 23,199 more in GitHub repos found copy, paste, forget is the norm.
If you reuse skills, do you review them first, or just drop them in?
POV: you applied to 200 jobs and got 0 interviews.
Maybe you didn't fail. Maybe you were just applying to the wrong ones.
I'm building Chayan. Upload your resume, see the jobs that actually fit you, with a reason for each one.
Early access opening soon 🙂 Link in the reply 👇
Most people still think AI is slow.
Casual users treat it like a slightly better search bar or writing assistant. They see the mistakes, the occasional weird outputs, and conclude it is overhyped.
A smaller group running frontier agentic models on real codebases is finishing work that used to take days in a few hours.
Agents restructure systems, write tests, and ship changes with minimal hand-holding.
The difference is not magic. It is access to stronger models, larger problems, and heavy use on tasks that can be verified. The gap is already wide, and it is growing.
the hardest question i got this year was five words long.
"why would you do that?"
i'd just left jp morgan. nothing was wrong. i loved it there.
somewhere along the way i'd stopped recognising myself.
try explaining that to a relative.
Title: The Lens Within: Seeing the World Through Our Own Story
Look closely at the image. It's a simple, almost diagrammatic illustration, yet it holds a profound mirror to the human condition. We see one single tree, lush and static in the center. Yet, around this tree, four distinct individuals—a child, a young student, a lumberjack, and a businessman—gaze upon it. The genius of the image lies in what we cannot see on the tree itself, but what we can see floating in thought bubbles above each person's head.
The child, innocent and unburdened, sees the tree as a sanctuary—a place for a bird's nest, a symphony of chirps.
The student, perhaps thinking of his future or his environment, envisions a functional object: a wooden desk.
The lumberjack, defined by his craft and livelihood, sees potential: raw timber, logs ready to be split.
The businessman, driven by commerce and capital, sees the tree as liquid assets: stacks of money.
This stark visual contrast perfectly illustrates the core truth of the quote below it: "We don't see things as they are, we see them as we are."
This isn't just about how different professions view the world; it's a universal commentary on perception. It reminds us that objective reality is an illusion we construct. Every person we meet, every challenge we face, every sunrise we witness, is filtered through the unique and complex lens of our own experiences, beliefs, desires, fears, and biases.
Think about a disagreement with a friend. You are looking at the same "tree," but one of you might see the threat (the lumberjack's logs), while the other sees the hurt (the child's disturbed nest). We are not arguing about the event itself; we are arguing about the meaning we have projected onto it.
This image offers both a humbling realization and a powerful key to empathy. It humbles us because it shows we are never truly objective. We are subjective interpreters, constantly rewriting the world's narrative to fit our own.
But it also empowers us. If we can recognize that our current perception isn't absolute truth, but a product of our inner state, we create space for understanding. When we encounter conflict or misunderstanding, we can pause and ask, "What lens am I looking through right now? And what lens might they be using?"
The tree remains the same, but the world is transformed by the observer. The power lies not in changing the world, but in consciously choosing how we view it—and realizing that the most important thing we see is a reflection of ourselves.
🔒 I'm planning to launch a private DM group, but not like the others.
🧠 Focus: understanding the X algorithm, growing by the rules, real guidance.
🌍 English-first, referral-based, for active verified accounts only.
📩 Interested? Vote, and if it's a yes, DM me 👇
68%Yes, I'm in. DMing now
24%Interested, tell more
7%Maybe later
1%Not for me
71 votes • Final results

















