Thien Do

AI engineer focused on optimization, intelligent systems, and hands-on experience with LLMs, VLMs and Stable Diffusion

[0xMarioNawfal @RoundtableSpace]

THIS GUY JUST DROPPED A 16 MIN TUTORIAL ON USING GEMINI 3.1 + SEEDANCE 2.0 TO BUILD CINEMATIC $10K WEBSITES

[toyxyz @toyxyz3]

LTX 2.3 IC-LoRA EditAnything test #AI #AIイラスト #comfyui

[AlphaSignal @AlphaSignalAI]

A peanut-sized Chinese model just dethroned Gemini at reading documents.

GLM-OCR is a 0.9B parameter vision-language model.

It scores 94.62 on OmniDocBench V1.5, ranking #1 overall.

For context, it outperforms models 100x its size. 100% open-source.

It works in two stages.

1. A layout engine detects every region in a document.
2. Each region gets read in parallel.

The model predicts multiple tokens per step instead of one.

That's what makes it so fast at small size.

It handles things most OCR tools struggle with:

> Complex tables and nested layouts
> Handwritten text and stamps
> Math formulas and code blocks
> Mixed image-and-text documents

You can run it locally through Ollama.

It fits on edge devices with limited compute.

Every expensive OCR API just got a free competitor.

[andy nguyen @andynguyendn]

Memory Skill for OpenClaw with 26k+ users in 1 week🚀

OpenClaw's memory system is broken by default.
It requires curating massive MEMORY.md files or relying on duplicate-heavy generation. Hours are wasted tuning, and massive amounts of tokens are burned.

It's time to stop. So we built the memory skill to solve that prob
Here is our superpower ⚔️
🎯 Top #1 market accuracy (92.19%) after 8+ months of intense architecture iteration
🧠 The ultimate solution to keep the timeline, facts, and meaning perfectly in place
☁️ Local & Cloud + Version control
⚡ Super easy setup

[Look at my shining eyes @GothicOrnate]

Seedance 2.0 复刻周星驰《大话西游记》经典情感戏片段...

[Code With Antonio @codewithantonio]

figma MCP + playwriter + opencode is everything you need

[Thien Do @dotieuthien]

While serving LLMs with vLLM in production, I ran into a pretty weird issue where some requests had TTFT spike to 2–3 hours.

I spent some time digging into the scheduler and KV cache behavior, managed to reproduce the issue, and wrote a short blog post explaining the root cause.
Would really appreciate any feedback or corrections 🙏

https://t.co/A3P00IkPfy

[Tom Yeh @ProfTomYeh]

llm.c by hand ✍️ C meets Transformer

This combination is perhaps as low as we can get to explain how the Transformer works.

Special thanks to @Andrej Karpathy

--

Part 1: gpt2_forward

Karpathy's llm.c implements the transformer forward step as the following sequence.

(skip)layernorm_forward
1. matmul_forward
2. attention_forward
3. matmul_forward
(skip) residual_forward
(skip) layernorm_forward
4. matmul_forward
5. gelu_forward
6. matmul_forward
(skip)residual_forward

--

Part 2: matmul_forward

B: Batches (of Tokens)
T: Tokens
C: Input Channels
OC: Output Channels

--

C programming and matrix multiplication are two of the most important topics. But it is often difficult to get people excited by these topics.

I hope this exercise can help people see further into the LLM black box, and appreciate C programming and matrix multiplication by hand. 😀

---
100% original, made by hand ✍️
Join 56K+ readers of my newsletter: https://t.co/fFt8roc8D9

[Md Ismail Šojal 🕷️ @0x0SojalSec]

The cleanest, most satisfying visual explanation of convolution I've ever seen.✨

- Kernel flip
- edge detection
- Gaussian blur
- complex plane glow-up.
- Math = dangerous levels of satisfaction.

Credit by : 3Blue1Brown, energy meets retro Mario & Kirby

[Md Ismail Šojal 🕷️ @0x0SojalSec]

Chinese scientists have developed,

The best shortest-path algorithm in 41 years!

A team from Tsinghua University has broken Dijkstra's "sorting barrier" - the first improvement since 1984.

Just use for a world-map 🤯

Paper - https://t.co/0AhR5O7vl4
https://t.co/a9KMVRuYGx

[Thien Do @dotieuthien]

@bluequbit sorry @bluequbit , I have already updated something and changed the link, please use this link https://t.co/2zK5vHU3dW

[Thien Do @dotieuthien]

Debugging mini-sglang, step by step for a single request.
Please read and give me feedback.

https://t.co/WgnPdrzBi1

[PatternedAI @patternedai]

Big news 🎉 Pattern Recoloring is here!

Explore endless colorways of your seamless patterns in minutes without breaking tileability or spending hours manually tweaking colors in Photoshop.

Try it now → https://t.co/A8jc6bZtuw

[Code With Antonio @codewithantonio]

💥 Introducing Polaris, an AI-powered code editor like Cursor or Replit.

Stop using AI coding tools. Start building them. 😎

✏️ Full code editor with syntax highlighting & minimap
🤖 AI-powered code suggestions & quick edit
🧠 AI agent with file manipulation tools
📚 Live docs scraping with @firecrawl
▶️ In-browser code execution with @webcontainers
🖼️ Live preview with terminal output
🐙 GitHub import & export
⚡ Background job execution with @inngest
🗄️ Real-time database with @convex
🔐 Authentication with @clerk
🐛 Error tracking, monitoring & logs with @getsentry
🐰 AI PR reviews with @coderabbitai
🌐 Next.js 16 + TypeScript
🎨 @shadcn UI + @tailwindcss v4

[Tongyi Lab @Ali_TongyiLab]

1/4 We’re releasing MAI-UI—a family of foundation GUI agents. It natively integrates MCP tool use, agent user interaction, device–cloud collaboration, and online RL, establishing state-of-the-art results in general GUI grounding and mobile GUI navigation, surpassing Gemini-2.5-Pro, Seed1.8, and UI-Tars-2 on AndroidWorld.
To meet real-world deployment constrains, MAI-UI includes a full-spectrum of sizes, including 2B, 8B, 32B and 235B-A22B variants. We are publicly releasing two models: MAI-UI-2B and MAI-UI-8B.

[Thien Do @dotieuthien]

@SohomScales yoooo thank you so much @SohomScales , You were one of the very first to drop congratulations on my own product

[Greg Brockman @gdb]

Prompting GPT-5.2 Codex for long-running tasks:

[Thien Do @dotieuthien]

@codewithantonio @inngest @stackblitz @convex Legend, I bought your course without a doubt

[Jintao Zhang @Jintao_Zhang_]

TurboDiffusion: 100–205× faster video generation on a single RTX 5090 🚀

Only takes 1.8s to generate a high-quality 5-second video.

The key to both high speed and high quality?
😍SageAttention + Sparse-Linear Attention (SLA) + rCM

Github: https://t.co/vT3nfax8H9

Technical Report: https://t.co/LEgLyhdPXh