Yacine

21.
No capital. No roadmap.
Learning difficult things and seeing what they can buy me.

[Yacine @yacinesm1]

$30,000 spent. 9 months wasted. Still no MVP.

This is what a SaaS founder told us last month.
His "agency" kept saying "2 more weeks" for 9 months straight.

We rebuilt everything in 5 weeks.
He launched. Got 47 users in week 1.
First paying customer on day 9.
Here's what actually works:

-No feature creep (MVP means MINIMUM)
-Battle-tested stack (React/Node/AWS)
-Security from commit #1 (not "we'll add it later")
-Fixed timeline (6 weeks max, no excuses)
-Clean code (documented, tested, scalable)

We've done this 3 times in the last 90 days.
If your MVP is stuck in dev hell, we can un-stick it.
DM me. Let's ship.

[Yacine @yacinesm1]

How can we integrate AI in blockchain?

[Yacine @yacinesm1]

@thatguyrahim The goat

[Yacine @yacinesm1]

Nothing beats late night focus sessions .

[Yacine @yacinesm1]

Finished a simple character-level transformer using a single head.

1 week ago i felt like it was a very hard thing , yet here i am

[Yacine @yacinesm1]

@Henryf1w my loud keyboard

[Yacine @yacinesm1]

@matthias_mrc 1h ago , when you are learning you must minimize the use of AI

[Yacine @yacinesm1]

overthinking is the enemy , just start and everything will flow

[Yacine @yacinesm1]

Anyone here already built a Transformer from scratch ?

[Yacine @yacinesm1]

@karthikponna19 It depends on your usage bro

[Yacine @yacinesm1]

@Alex1here Uninstall windows

[Yacine @yacinesm1]

@ronisarkar_exe Doesnt mean they dont exist (i have seen them its scary)

[Yacine @yacinesm1]

@bilalc_1 Goated house

[Yacine @yacinesm1]

@samirande_ DSA in C for exams were the best days in my life

[Yacine @yacinesm1]

What was the first research paper that you have read , and how hard was it at first ?

[Yacine @yacinesm1]

This is the kind of problem I'm interested in:

Taking a company's GPU infrastructure → choosing the right model → deploying it with vLLM → making it actually usable in production.

If you're trying to run an LLM privately on your own hardware,
feel free to DM me.

[Yacine @yacinesm1]

How I would choose a model for a private LLM server

[Yacine @yacinesm1]

The real decision I would evaluate:

-Model quality

-VRAM requirements

-Context length

-Concurrency

-Latency requirements

-Input modalities

-Infrastructure cost

Then choose the model.
Not the other way around.