Kenji Phang (@kenjiphang) 's Twitter Profile
Kenji Phang

@kenjiphang

Building autonomous companies @Fern_labs by day
Building agentic office UIs by night

ID: 1559550797058277376

calendar_today16-08-2022 14:41:15

181 Tweet

77 Followers

132 Following

Guillermo Rauch (@rauchg) 's Twitter Profile Photo

We shipped an OSS 'vibe coding platform' (like v0) built with Vercel AI SDK, Gateway and Sandbox. We worked with @openai to tune the GPT-5 agent loop. It can write/read files, run commands, install packages, autofix errors… Demo oneshotting a multiplayer Pong in Go ↓

v0 (@v0) 's Twitter Profile Photo

Introducing the first step towards design systems in v0: •Create unique color schemes for your v0 generations •Adjust the colors for a design system right from Design Mode •Preview in light & dark mode Learn more directly from v0 team member and creator of tweakcn

Kenji Phang (@kenjiphang) 's Twitter Profile Photo

I realized that Cursor tab was always suggesting the same url whenever I ask for one, so I got curious and clicked on it you should click for yourself youtube.com/watch?v=dQw4w9…

I realized that <a href="/cursor_ai/">Cursor</a> tab was always suggesting the same url whenever I ask for one, so I got curious and clicked on it

you should click for yourself
youtube.com/watch?v=dQw4w9…
omni_georgio (@omni_georgio) 's Twitter Profile Photo

We brought 20 developers from LDN to NYC to build with agents for 1 week. This is day 4. What’s becoming clear: London is proving itself as a powerhouse for AI and talent. - Top universities are creating and bringing in talent. - A dense network of investors and accelerators

Kenji Phang (@kenjiphang) 's Twitter Profile Photo

A lil late to the content game >> But me and Yicheng have just been flown out to NY 🇺🇸 >> To compete with 10 of London's 🇬🇧 most competitive hackers >> For a prize pool of $100k 💰🤑 The stakes are high, what will we build 🤔 cc Unicorn Mafia Coral Protocol Solana

机器之心 JIQIZHIXIN (@synced_global) 's Twitter Profile Photo

Wow, a new post-training method. SFT = efficient but capped 🚦 RL = powerful but slow 🐢 Now enter: Guess-Think-Answer (GTA) GTA fuses guess (SFT), think (reflection), and answer (RL-shaped). Result: ⚡ Faster convergence than RL 📈 Higher ceiling than SFT 🛠️ Gradient

Wow, a new post-training method.

SFT = efficient but capped 🚦
RL = powerful but slow 🐢

Now enter: Guess-Think-Answer (GTA)

GTA fuses guess (SFT), think (reflection), and answer (RL-shaped).

Result:
⚡ Faster convergence than RL
📈 Higher ceiling than SFT
🛠️ Gradient
Kenji Phang (@kenjiphang) 's Twitter Profile Photo

My type of holiday: > get flown out to different parts of the world to build AI agents and just doing unimaginable things