Developer 👨🏻💻| Coding Enthusiast | React, Nextjs, Tailwind CSS | Cricket 🏏 | Vibe coder 😂
Ahmadabad City, India
Joined April 2022
- Tweets3.2K
- Following1K
- Followers770
- Likes5.8K
Hey @thsottiaux my codex desktop app (windows) is behaving strangely. When I am in a chat in a project it suddenly show the loading screen and comes back to new chat screen. This happens every 1 or 2 minute. Current version is 26.1023.50603 — build 22391.
Hey @thsottiaux codex agents are being blocked by Amazon when I ask my luna to find me specific product on amazon using browser, it getting blocked by Amazon human check. This was working fine till last week but now this is being blocked. Any solution to this?
Luna 5.6 is draining limits much faster in codex @thsottiaux . Earlier it was basically unlimited and now it drains like I am using astra or something
Hey @thsottiaux my codex limits are getting used suspiciously more today. Just 3-4 chats with gpt 6.1 sol and then only luna xhigh quickly consumed my 5 hours limit today. Usually it is sufficient for me but Today it is draining much faster
Tejas ✌️ retweeted
🚨 #CRUDEOIL SETUP COMING 🚨
Risk: ₹3K–₹4K
Potential Profit: 2X–3X+ 🔥
❤️ 150 Likes
🔁 100 Reposts
Hit the targets above and I’ll DM the setup ONLY to those who participate!
#StockMarket #OptionTrading #CrudeOil #Trading #Jackpot
Tejas ✌️ retweeted
huawei ships next level innovative phones. Yet, not allowed in open market like USA.
EV cars are another example.
So Starlink is another one such example, that’s it.
- tweeting from Jio internet.
Tejas ✌️ retweeted
HE DIED BECAUSE OF ALIMONY!!
3 CRORE DEMANDED
HAD ALREADY PAID 1 CRORE
THAT WITCH CAN ENJOY THE MONEY NOW
Tejas ✌️ retweeted
This is an Indian Hero Pilot Captain Smit Mac. The follow pilot tried to kill him, but he showing alertness called out to the crew and passengers. The open door of the cockpit helped. Salute Captain Sumit Machchhar.
Tejas ✌️ retweeted
Correction: the heroic pilot who tried to fight off the Islamic attacker midair, and opened the door despite being stabbed and wounded so Israelis could help, is an Indian man.
He is a hero.
Tejas ✌️ retweeted
someone asked if the 7/10 would hold with a bigger batch. so I ran 50 each
GPT-5.6 Luna: 50/50
GPT-6 Luna: 40/50
it called the croissant "other" 10 times. no API errors. it's also slower, 2.26s median vs 1.47s for 5.6
that sucks. GPT-6 Luna seems like a nerf compared to 5.6 Luna
still one logo, not a full vision benchmark. but it held. @moteefai stays on 5.6 Luna
GPT-6 Luna is cheaper per token than 5.6 Luna, so I tried switching @moteefai's marketing chat over to it
all 270 tests passed. conversation and tool calls worked fine
then the image check. same logo, same prompt, 10 runs each. a clearly drawn croissant
GPT-5.6 Luna got it right 10 out of 10. GPT-6 Luna got it 7 times. the other 3 it just called it "other"
so apparently the new one isn't as good as 5.6, which is very disappointing. you expect a newer model to at least match the old one
it's a small test, not a full eval. but it's a repeatable miss on something the old model never got wrong, so production stays on 5.6 for now
funny part is the saving wasn't even worth much. Luna was about $0.78 of my $49.53 API bill this month. image generation was $42.94
Gpt 6 luna seems very dumb, slow and worst at instruction following or reserach task as compared to 5.6 luna
Gpt 5.6 sol in chat gpt interface seems very dumb since few days.. can't even remember list of 30 items and keeps getting hallucinated. Is open ai nerfing it?
