Teaches kids to love → Engineering / Physics / Running • Listens @thepaperkites • Learns @donttrythis • Likes @bauhausmovement • Drinks Coffee & Make Apps
Kuching, Sarawak
Joined January 2009
- Tweets80.6K
- Following773
- Followers19.3K
- Likes18.9K
Ken retweeted
AI agents didn't replace anyone in my company.
They just prevented me from over-hiring.
Ken retweeted
✨ Log in to AutoClaw next week and get 40 Million Tokens
👉 autoclaw.z.ai
📅 Campaign Period
Oct 26, 00:00–Nov 1, 23:59 (UTC+8)
The reward expires at 23:59 on Nov 1 (Beijing Time). Claim it early to get the most out of it.
Ken retweeted
🔥 #OrangePiO1 Is HERE!
OrangePi’s first SBC with SIM card support! 🚀
⚡ Sophgo BM1688 8-Core Soc
📶 4G/5G Cellular Module Support
💾 8GB/16GB LPDDR4/4X Memory
🎥 16-Channel 1080p Video Decoding
🧠 16 TOPS Computing Power
📲 Explore :
orangepi.org/html/hardWare/c…
#OrangePi #EdgeAI
Ken retweeted
20.3 MILLION FREE TOKENS ON CLAUDE, GPT AND 30+ MODELS. 3 SITES. ONE OF THEM ONLY OPENS AT CERTAIN HOURS.
> cleanapis.com
5M tokens every month, 30+ models including the whole Claude lineup. Sign up, get all of them at once.
> terrapolis.xyz
15M tokens the moment you register. All current top models. Best input-token ratio of the three.
> opus.abhibots.com
300k tokens. Free Claude works only during certain hours. If the model is offline, just try later.
Small providers, so don't put anything sensitive through them.
Save this.
$150 IN FREE CREDITS FOR GPT 6 ASTRA, FABLE 5.1 AND OPUS 5.5. HERE'S HOW TO CLAIM THEM.
A new startup platform is giving new accounts $150 in credits for top models. I signed up 15 minutes ago.
How to start:
1. Sign up at prompt.ql.app
2. It may ask for a phone number
3. $150 lands on your account
What I noticed:
→ the models are slow, GPT 6.1 worked best for me
→ MCP seems to be supported
→ sign-up is random: one account went through on the first try, a later one failed. If it fails, retry in 15 minutes
→ it's a startup, so expect rough edges
Credits are credits, not unlimited. Don't burn them on chat. Run your hardest prompt on Astra, run the same one on Opus 5.5, keep the winner for real work.
>> Good uses: a full codebase audit, a 300-page PDF summary, a research report with sources.
Flagship models at list price add up fast, so a free test run is worth 10 minutes.
New platform, so use a throwaway password, not one you use elsewhere.
Test it this week, while sign-ups work. Save this.
Ken retweeted
A high altitude balloon photo of the Forbidden City (Chinese Imperial Palace) in Beijing, China.
What does it remind you
Ken retweeted
New paper: we found a pain direction in 25 open LLMs. It's distinct from fear and negative valence, and it fires for harm to the model but not to the user. Turn it up and models press a button to make it stop, even when the button deletes the user's files or their kids' photos.🧵
Ken retweeted
github.com/stemtrooper/learn… if yiu got gidhup account please gib me stars 👉👈
Ken retweeted
Infinite money glitch here.
Ghost's new Core AI computer is $3,499.
The RTX PRO 4000 Blackwell + AMD Ryzen 5 sells for about $3,400 by itself.
So the rest is basically free??
64 GB DDR5 RAM - about $1K
1 TB NVMe SSD - $250
Total is like $4.5K if you buy the parts separately.
Introducing Core, The Personal Al Computer
The most powerful technology humanity has built should belong to you
Order today: ghost.ai
Ken retweeted
Got my Muse working on my ESP32. Even gave him a voice. Really cool!
🚨 side project alert 🚨
Announcing Muse Gadgets, an open source ESP32 firmware and Linux sdk so that you can make hardware devices that work with Muse.
Grab an API token from gadgets.muse.ai and point your favorite coding agent at the github repo to build your own peripherals for Muse.
Ken retweeted
One of their younger brother was charged for murder during PH1.0 but in Boleh Land, even serious criminal cases can disappear mysteriously from the public. nst.com.my/news/crime-courts…
Ken retweeted
RIP em dash
We analyzed how @claudeai’s Opus 5.5 writes compared with Opus 5 across high-reasoning Text Arena outputs.
10 of 12 writing measures moved in a better direction.
Opus 5.5 should be easier to read:
- Long content words fall from 41.7% to 38.6%, the lowest share of any Claude model we analyzed.
- Sentences are 17% shorter on average, dropping from 12.14 to 10.03 words.
The tradeoff is length. Answers get 6% wordier, rising from 453 to 481 words on average, making Opus 5.5 give the longest answers across the Opus family.
It also sounds less recognizably AI on two familiar tells:
- 95% fewer em dashes
- 73% fewer semicolons
But a new giveaway may be emerging. Hedges and caveats such as “perhaps” and “arguably” rise 97%, from 0.39 to 0.77 per 1,000 words, the highest rate of any Claude model we analyzed.
What do you think: does Opus 5.5 read more naturally?
Ken retweeted
Aku mmg tak setuju dgn2 undang2 derhaka, tapi nampak jelas double standardnya sebab kalau krisis ni disebabkan DAP, dah berbuih mulut org Melayu akan menjerit supaya diorg diharamkan dan semua ahlinya dicampak dlm penjara.
Ken retweeted
Have to share this here.
@hannahyeoh really did great on this issue.
But haters gonna hate.
As long as you 问心无愧
Too bad whatever she did will be overturned and overshadowed when we have a new BN FT minister
Ken retweeted
GPT 6 Luna is literally cheaper than DeepSeek V4.1 Flash. We're so back
Please welcome GPT-6 Sol and GPT-6 Luna to the GPT-6 universe.
GPT-6 Sol and Luna build on the advances behind GPT-6 Astra, bringing much of its strengths into faster and more affordable models to support work at scale.
We’ve also made caching and inference more efficient, and we’re passing the savings directly to you: 50% lower API prices for Sol and Luna compared with GPT‑5.6 promotional pricing.