AI Developer · Indie Hacker · AI Enthusiast · Full-stack Dev | Tech Stack: Python / Go / React / AI Agent / RAG / LLM | Focus On: Vibe Coding · Build in Public
Joined October 2019
- Tweets1.6K
- Following1.6K
- Followers1.9K
- Likes2.1K
Hey everyone! 👋
I'm an AI developer from Hong Kong, passionate about AI and emerging technologies.
If you're into AI and tech, let's connect! 🤝
AI agents can do the work. But are we slowly forgetting how to think?
The real danger isn't AI making mistakes. It's humans losing the ability to recognize them.
Delegate execution, not judgment.
Automate the work, not accountability.
Think critically. Verify with evidence. Own the outcome.
AI can take the work off your hands, not the responsibility.
Wow, this offers a new approach to model training — using a powerful model to edit and optimize the weights of a smaller one.
We asked GPT-6 Astra to edit Qwen3.5-0.8B weights to improve math.
It directly changed only 0.016% of weights, fixing 0.8B's reasoning loops & improving math up to 6x and even instruction following 3x, without gradients or training.
We call this Weight-Space Programming: frontier LLMs can now natively read/write neural network weights! 🧵
Blog: rikiyat.github.io/blog/weigh…
Readers added context they thought people might want to know
LLM was given weights explicitly as text in prompt and used code/forward passes (per authors' logs) to ablate Qwen looping direction, a known thinking-mode issue per model card. Primarily fixed repetition in that mode, not general 'natively' improving weights/math.
rikiyat.github.io/blog/weight-sp…
github.com/RikiyaT/lm_und…
huggingface.co/Qwen/Qwen3.5-0…
huggingface.co/RikiyaT/Qwen3.…
Attention Is All You Need. Agents Are More Than LLMs.
Agent engineering is about the entire stack:
Agent Loop, Context, State, Memory, Tools, Skills, Protocols, Runtime, Sandbox, Observability, Evaluation, Security & Governance.
MCP, Plugins, Subagents, Hooks, and external Session Stores bring it all together.
The future is agent architecture.
Every day is Zeroday. ⚠️
Wiz researchers uncovered at least 9 security incidents across AI developer platforms in October alone.
Fireworks AI, Lightning AI, Railway, Warp, OpenCode… the list keeps growing.
Credential theft. Agent swarm attacks. Model theft disguised as distillation.
The AI boom is creating a massive new attack surface.
Build fast. But never forget security. 🔐
wiz.io/blog/black-october-fo…
Vibe coders, what have you been building? 👀
Show us what you've created! Drop your projects in the replies 👇
Let's discover cool ideas, share inspiration, and connect with fellow builders! 🚀
Your coding agent can write code, but can it use your logged-in browser to get real work done? 👀
From Claude Code to Codex and Gemini CLI, there are 6 ways to connect AI agents to your existing Chrome session.
Here’s how they compare in capabilities, setup, and security — and which one you should choose. 🧵
🚀 What if websites could expose tools directly to AI agents?
That’s WebMCP.
Instead of relying on screenshots, clicks, and DOM guessing, websites can expose structured tools through JavaScript APIs.
MCP connects agents to backend systems.
WebMCP connects agents to web experiences.
The future? Websites built for AI agents, not just humans. 🤖
I've been using it, and it's perfect for everyday use! If you need it, give it a try
300M free GLM-5.3-Flash tokens on ZCode. Every day.
Claim from 11:00 UTC. Use 15:00 – 01:00 UTC.
Update to the latest ZCode, claim in-app, then switch to Global Building and pick GLM-5.3-Flash.
zcode.z.ai
Moonshot AI just unveiled Kimi K3.1 Pro, a 2.8T-parameter flagship model with a 1M-token context window.
Full model weights are expected to drop by Nov 4.
API pricing is down roughly 50% vs. K3:
• Input: ¥13 / 1M tokens
• Output: ¥68 / 1M tokens
Need more speed? The new Highspeed version delivers 2–3× faster output at 2× the price.
Introducing Kimi K3.1 Pro: our most capable 3T-class open-weight model.
🔹 1M-token context, 2.8T total / 104B active params (MoE), Multimodal.
🔹 Full-project coding: games in Unreal, Unity & Godot.
🔹 Excels at parametric 3D modeling in Blender & CAD.
🔹 More reliable at finance work: Accurate modeling with hallucination down by 65%.
🔹 Matching the strongest proprietary models across coding, agentic and knowledge benchmarks.
Kimi K3.1 Pro and K3.1 Pro Highspeed (2–3× faster) are now live on Kimi.ai, Kimi Work, Kimi Code, and the Kimi API.
Open weights by Nov 4, 2026.
🔗 API: platform.kimi.ai
🔗 Tech blog: kimi.ai/blog/kimi-k3-1-pro
What happens when AI goes beyond answering prompts and starts doing the work? Cloud Agents combine models, Harness, tools, and sandboxes to execute multi-step tasks. Here’s how they work, where they shine, and what you trade for convenience.
