I said it before, AMD should be training models specifically on all of their documentation and IP, and using that to optimize things.
Working with @AMD, our models identified up to 14% PPA improvement on flagship GPU IP in less than five months—building on designs AMD’s engineers had already spent years optimizing.
The same approach helped improve Jalapeño, our first custom inference chip, with gains verified on real silicon.
New evaluations also show our models getting better at certain chip design tasks.
We’re sharing these results and how we’re working with @Synopsys to bring our model capabilities and their trusted EDA design and validation tools to more chip designers.
openai.com/index/advancing-c…
Witizenship retweeted
You: Want a PS5 to play GTA 6.
We: Are giving 6 away.
Just repost this, reply with why you want it, we'll pick the best replies to win.
Paid out in upgraded money (USDC). US only.
Witizenship retweeted
ai won't replace you
somebody using ai won't replace you
robert pattinson will replace you
Bring this to the UK man. Why is the UK never on your release schedule?
Knowing things → doing things → finding ways to save you time and money
We’re expanding Tasks in Gemini free of charge to hundreds of millions of people in six countries. Powered by Gemini 4 Hydrogen, Tasks lets you hand off everyday life admin directly:
- Forward emails to [email protected] to RSVP and add it to your family calendar
- Have Gemini call local businesses to compare price quotes and check inventory (with full transcripts + approval before dialing)
- Check out securely via Google Pay (uses the Agent Payments Protocol - AP2)
- Seamlessly sign in via Google Password Manager
We wanted to make this helpful + secure, we’re also partnering directly with major websites and security providers so it works well for site owners and agents.
10
Witizenship retweeted
$200 of GLM 5.3 usage for $10 with opencode go!!!
Up from $60. Deepseek v4.1 flash still gets $100
- glm 5.3 flash $1.40/$4.40
- deepseek v4.1 flash $0.15/$0.60 off peak and $0.30$1.20 peak
good thing is that peak hours are only Weekdays 01:00-04:00, 06:00-10:00 UTC for deepseek
some sick inference engineering going on at open code
If this is true then GLM5.5 or GLM6 actually has the opportunity to take a big leap ahead. They are doing their own data this time round.
You will notice that Gemini Carbon thinks differently than Claude and GPT. It often comes up with entirely different but more clever solutions than competitor models.
This is precisely because we never distilled other models. Distilling other models creates a poor imitation rather than unique intelligence.
226
Can this be used with DeepSeek Harness?
OpenCode Go at $10 just became the best deal.
you now get:
> GLM 5.3 for planning/orchestration
>DeepSeek V4.1 Flash as the workhorse
a few things i’ve noticed:
1. GLM 5.3 is the strongest Chinese model for serious agentic work. it feels deeper, more reliable and better at following guidance than DeepSeek.
2. GLM is slower, but smarter. i’d rather use it for architecture, planning, difficult decisions and orchestrating other agents.
3. DeepSeek V4.1 Flash is still the better workhorse. it’s cheap, fast enough and very good at finding and fixing issues once the scope is clear.
4. GLM is better at writing code cleanly and following a plan. DeepSeek can overthink badly, get stuck in loops and burn 2-5x more thinking tokens on similar tasks.
5. the combination is more useful than either model alone. GLM decides what should happen, DeepSeek does most of the repetitive implementation and debugging.
for $10, that’s a pretty strong setup:
GLM 5.3 = brain / orchestrator
DeepSeek V4.1 Flash = execution / workhorse
35
Deepseek is phenomenal at long Horizon tasks omg it's not even close. I have used OMP and Opencode and Codex and Claude Code and Kimi Code. How did they do this?
1
79
Witizenship retweeted
I think a lot of people are sleeping on Googles AI Models. Here is a list of reasons why its use of custom TPUs will help beat the competitors:






































