G-Code is officially OPEN SOURCE
o-research.tech/mac
The full Mac agent. The engine, the tools, the permission system, the local MLX/GGUF runtimes. No SaaS. No gated runtime.
Come and get it 👇
I built it. I open-sourced it. AI subscriptions, paid out of my own pocket. Hardware, bought myself and backed by friends. I emailed the government for help. Replies: zero. @mymdec when you roll out the red carpet for foreign founders, do any of your policies cover local solo developers? @anwaribrahim have you ever given your own citizens a second glance?
Malaysia will let you move your startup there with a pass valid for up to 10 years.
No local co-founder needed.
The Malaysia Tech Entrepreneur Programme (MTEP), run by the government's digital agency:
• New founders: a 1-year pass, renewable for another year
• Established founders: a 5-year residence pass, renewable for 5 more
• Established founders can bring their family
• Apply online from anywhere
• Decisions in about 4 weeks
• AI startups are a priority
What you need:
• A tech product, at least in beta
• RM50K (~$12K) in the bank for 3 months
• A plan to register your company in Malaysia within a year
No equity taken. No program to finish.
Just a door into Southeast Asia.
mdec.my/programmes/mtep
1
56
Rapid-MLX: ‘up to 4x faster’
TensorFold: ‘exact decoding’
G-Code: took the best from both, rewrote in Swift, 3.2x faster, bit-exact, session restore in 2.2s, 60 tests passing.
Sometimes you don’t need to pick a side. You just take both. 😈
G-Code is open source. This engine isn’t.
github.com/leetn9468/G-Code
1
27
Qwen3.8-Flash-Next 120B: native Swift implementation. No Python. No external dependencies. 3,434 weight arrays verified.
Max delta from reference: 0.000001.
#LocalCoding #LocalAI #OpenSource
39
Dual format support: MLX and GGUF native.
12.4 t/s vs 5.3 t/s, 2.3x faster (M1 Max 64GB)
One click.
Local. Free, but no I don’t feed you shit.
Nov 15.
(speed test on M5 Ultra 256GB soon)
#LocalAI #LocalCoding #OpenSource
1
57
G-Code Lite is live on the App Store.
A real AI coding studio on your iPad. Python, Node and C/C++ run on the iPad itself, offline. The agent writes, builds and checks. Nothing lands until you approve it.
Your API key. No account, no servers, no subscription. $3.99 once.
apps.apple.com/app/g-code-li…
36
[G-Code test with local model]
M1 Max 64GB, Qwen3.8-27B-Q4_K_M (Extra): 287minutes
M5 Ultra 256GB, Qwen3.8-27B-Q4_K_M (Max): 139minutes
Compare with Deepseek V4.1-Flash-API: 28minutes (previous post)
Open source. 15 Nov. Mark the date.
#LocalAI #LocalCodingTool #OpenSource
96
TN Lee retweeted
We're giving away a Hogwarts Legacy custom wrapped NVIDIA GeForce RTX 5090 to celebrate its official release with DLSS 5 🪄
To enter:
🟢 Share this post
🟢 Comment #RTXON
Terms and Conditions: nvidia.com/en-us/geforce/con…
A fully local coding agent. Tick zero-retention on your provider’s platform and your privacy is maxed. Open source, thousands of edge-case tests. 15 NOV, mark the date 😈
#DSHplugin #compatible #LocalGGUF
1
1
34
Why do these people keep screaming to ban AI, halt progress, and preach the AI apocalypse? Are they looking out for humanity? No. They are protecting their class.
Through AI, ordinary people are learning at near-zero cost, building at near-zero cost, and leaping across class lines that took generations to cross. This terrifies the elite.
“Only use AI for science and medicine” sounds noble, doesn’t it? It is technological segregation. Keep the most powerful tool in history locked inside institutions the ordinary person will never enter, and the existing hierarchy is preserved. The elite stay on top. Everyone else stays where they are.
This is not democracy. This is elite capital exploitation wearing a democratic mask. Gender awakening and racial equity are the excuse. For these people, democracy is simply the most convenient vehicle for dictatorship.
They do not fear AI destroying humanity. They fear AI liberating it.
— TN LEE, 7 NOV 2026
1
37












