Arvin

17 y/o
Backend dev
im learning AI/LLM engineering

Founder of @rarex_app

[Arvin @Arvin_vx]

Day 0 — this is my public commitment.

I'm documenting my journey to becoming a Full-Stack AI Engineer from scratch.

Where I'm starting:
- CS student in Istanbul
- Know JavaScript, Node, Express, SQL
- built a few AI product (vibe coded)

Where I'm going:
- Build and ship AI products
- Eventually: contribute to the infrastructure that powers AI apps

The path:
Scrimba AI Engineering
Karpathy Zero to Hero
IBM Generative AI Engineering
O'Reilly AI Engineering book
Build for 1-2 months straight
Nvidia AI certification

Following this thread = watching someone go from 0 to AI Engineer in public.

Day 1 starts tomorrow.

[Arvin @Arvin_vx]

It’s been over a week since Claude Fable 5 got globally shut down by the US government.

Launched June 9 as Anthropic’s most powerful public model → killed June 12.

Anthropic says they’re confident it’s coming back “in the coming days”… but we’re still waiting.

Question for the timeline:

When do you think Fable 5 returns?
Drop your prediction + why 👇

#Claude #Fable5 #AIAgents #Anthropic

[Arvin @Arvin_vx]

@SamarthPagaria @X Hi im working on to become an ai engineer

[Arvin @Arvin_vx]

While learning AI agents the hard way, I started @rarex_app to put real businesses on autopilot.

AI systems + automations that actually work with their own data.

What kind of repetitive task would you love to kill with an agent? Tell me 👇

[Arvin @Arvin_vx]

@dani_avila7 son of j

[Arvin @Arvin_vx]

Day 10 of Becoming an AI Engineer 🔥
Today I finally understood the real
plumbing behind AI agents:

• Tool calling: Model decides → calls functions → gets results back

• Token management: Slice old messages when context fills up

• Structured outputs: Force clean JSON schemas + streaming

Key realization:
Building AI products = wiring models + tools + APIs + UIs.

Not just training models.
What clicked for you this week?

[Arvin @Arvin_vx]

@alexandresauv13 Yeah 👍 i agree

[Arvin @Arvin_vx]

I just shipped my first real full-stack AI agent yesterday.

No frameworks for the agent loop built ReAct from scratch.

Biggest lesson: The agent is blind unless you explicitly feed tool results back as messages.

Took me way too long to realize that.

Who else is building agents right now? What's the hardest part for you?

(Reply with your current project)

[Arvin @Arvin_vx]

Day 9: Built and deployed a full-stack AI support agent in 2 days.

The agent answers customer questions from uploaded business docs, books appointments, and escalates when it can't help. No hallucination answers only from the business's own data.

Stack: Express, Supabase + pgvector, OpenAI, LangChain, Next.js
Deployed on Render

First real full-stack AI project. Every concept from the courses / books that i read i applied in one build.

[Arvin @Arvin_vx]

@FarzaTV cool idea man !

[Arvin @Arvin_vx]

who said fable 5 is coming back ?

[Arvin @Arvin_vx]

@DanielSmidstrup Their focus is coding

[Arvin @Arvin_vx]

@Simi_2210_ 21

[Arvin @Arvin_vx]

@isha_singh06 python itself

[Arvin @Arvin_vx]

@codewith55 google chrome for sure

[Arvin @Arvin_vx]

@ravikiran_dev7 Tech skills aren’t everything. Dude knows how to be a CEO that’s why he scaled up at Y Combinator.

[Arvin @Arvin_vx]

Day 8 of becoming an AI Engineer:

Today I turned a chatbot into an agent.

Built the ReAct loop by hand, no LangChain, no framework. The model picks which tool to call, my code runs it, the result gets pushed back into the conversation, and it loops until it has a real answer.

The part that finally clicked: the agent has no memory between steps. It calls a tool but can't see the result unless you hand it back as a message. Forget that one line and the whole thing goes blind.

Mine pulls my real location from an API, checks my interests, and chains them into a recommendation, one tool feeding the next.

If you're building agents, what tripped you up the longest?

[Arvin @Arvin_vx]

Anyone else notice ChatGPT starts lagging hard once a chat gets really long?

Typing feels delayed, scrolling stutters, and the whole UI slows down

How do you handle super long chats? Summarize + new chat, or some extension hack?

[Arvin @Arvin_vx]

trillionaire joke

[Noah @antibearthesis]

Me realizing I’m closer to becoming a billionaire than Elon Musk

[Arvin @Arvin_vx]

Just finished "AI Engineering" by Chip Huyen 🦉

This isn’t another hype book about LLMs.
It’s the practical playbook for building real,
reliable applications on top of foundation models.

If you’re tired of cool demos that break in production, this is the book that shows you how to ship actually useful AI systems.

Highly recommend already changing how I approach projects.