Siraj Raval @sirajraval

Taught 1M people AI before ChatGPT. Building what tutorials won't teach you. The OG vibe coder sirajraval.com Joined May 2014

Tweets

4K
Followers

63K
Following

1K
Likes

12K

Siraj Raval @sirajraval

2 days ago

This interactive LLM visualization is an excellent learning tool. Press spacebar repeatedly to explore nano-gpt in all its tiny glory. bbycroft.net/llm

2 0 8 952 4

Download Image

Siraj Raval @sirajraval

a week ago

Best thing you can do for your X sanity: switch from For You → Following. For You programs you. Following = you programming yourself.

3 1 6 2K 0

GPT-5 is NOT the best AI for coding. I tested GPT-5, Claude, and Gemini using "Vibe Coding" (building apps with 100% prompting, 0% typing). The results were definitive and shocking: 🥇 Claude: 10/10 🥈 GPT-5: 8/10 🥉 Gemini: 7/10 Watch the showdown: youtu.be/277l16eJphI

1 0 1 2K 1

Siraj Raval @sirajraval

a week ago

Running uncensored open-source AI at scale is brutal. I’ve tried it. Most people can’t. Venice feels like the first option that actually works. venice.ai/chat

0 0 1 685 1

Siraj Raval @sirajraval

2 weeks ago

How long until nobody types code anymore?

0 0 2 1K 0

Siraj Raval @sirajraval

2 weeks ago

Do I use Claude Code or Cursor? Yes.

0 0 3 1K 0

Download Image

Siraj Raval @sirajraval

2 weeks ago

I took the summer off to spend time with my newborn son. While he learned to grab things, I learned to code without hands - Meta Raybans, Claude Code, voice only. Built 3 apps during tummy time. This is Vibe Coding. Tutorial drops Friday.

0 0 13 2K 2

Siraj Raval @sirajraval

4 months ago

I just open‑sourced hermes3‑fast‑inference: 4‑bit Hermes 3 on vLLM. 90 tok/s on an A100, 28 tok/s on a 4090 with no quality drop. vLLM is perfect for high-throughput, low-latency apps. Code + benchmarks → github.com/llSourcell/her…

3 1 11 3K 3

Polyhedra @PolyhedraZK

4 months ago

In fintech, adtech, and sports betting, bad model outputs mean fines, lost trust, and lawsuits. @sirajraval with over 1M cross platform followers, spitting knowledge on The Polyhedra Show.

75 19 240 7K 1

Download Video

Polyhedra @PolyhedraZK

4 months ago

Great to see @sirajraval applying zkML in real-world AI builds. We're excited to collaborate — and to have him join us as part of the zkML Festival, kicking off May 5. #zkMLfestival

Siraj Raval @sirajraval

4 months ago

Great to see @sirajraval applying zkML in real-world AI builds. We're excited to collaborate — and to have him join us as part of the zkML Festival, kicking off May 5. #zkMLfestival

8 6 24 12K 14

69 27 234 12K 1

Sam Altman @sama

5 months ago

o3 and o4-mini are super good at coding, so we are releasing a new product, Codex CLI, to make them easier to use. this is a coding agent that runs on your computer. it is fully open source and available today; we expect it to rapidly improve.

1K 3K 37K 3.2M 10K

Siraj Raval @sirajraval

5 months ago

Most AI products don't fail because of bad code. They fail because the founder hated waking up to them.

3 0 9 2K 0

Siraj Raval @sirajraval

5 months ago

You don’t need OpenAI anymore. Llama 4 Maverick beats GPT-4o on coding, reasoning, and vision. It runs on one GPU. It’s open-source. It’s fast. It’s free. The API tax is officially dead.

Ahmad Al-Dahle @Ahmad_Al_Dahle

5 months ago

You don’t need OpenAI anymore. Llama 4 Maverick beats GPT-4o on coding, reasoning, and vision. It runs on one GPU. It’s open-source. It’s fast. It’s free. The API tax is officially dead.