Developer Experience @openai
Based in United States
- Tweets 161
- Following 207
- Followers 468
- Likes 3,966
konstantine retweeted
I am very uncomfortable about people trying to ascribe religious force or a surrender of human judgment to AI models, and think it is a real safety issue.
5,525
1,804
26,556
5,591,780
konstantine retweeted
Pretty sure there are more dots than bots already in this little world.
Cool thing is that we’re improving them everyday and they learn directly from your feedback.
698
74
4,389
608,256
If talking to AI, I'd really prefer it NOT look like a human
Which is why I've named mine after my dog and he looks like a frog
Introducing Griffin, the first model to pass the video Turing test.
48% of people who talked to it live thought it was a real human. Previous systems have had a pass rate <3%. It is #1 on NVIDIA's benchmark for full-duplex AI video.
It’s the first Human Interaction Model (HIM).
Community note
The 48% figure and "video Turing test" claim are from Tavus's own study of 54 one-minute calls, not independently verified or using a standard protocol. Griffin-Lite leads NVIDIA's VideoFDB benchmark on their public leaderboard.
cellcog.ai/blog/tavus-gri…
research.nvidia.com/labs/amri/proj…
tech-ish.com/2026/10/02/tav
9
14
2,769
Just noticed today that (⌘ + /) opens keyboard shortcuts in codex without having to leave the chat.
1
136
Building this with @Dimillian was so fun but seeing people play it yesterday was even better.
It sucks though that I’ll no longer be able to justify having a steering wheel mounted to my desk at the office.
We had so much fun building the ultrafast game with @codestantine.
Once again, blown away by Astra’s capabilities. People could make their cars live by just prompting them using GPT-Live and getting custom 3D models in less than 10 seconds.
5
20
1,303
GPT-6.1 Sol performs even better than Astra at my low resource language translation bench. What a nice surprise.
And it’s an even bigger jump than from 5.6 Sol to 6 Astra.
1
14
1,044
konstantine retweeted
GPT-6.1 Sol matches Astra on DeepSWE at ~1/5 the cost!!
$2 input, $10 output and just $0.10 cached input per million tokens.
Enjoy!!
31
18
412
15,574
i got to try openai dots a week ahead of the launch as part of their early access community, some thoughts now that it's out:
1/ it's closer to Muse than Grokbot.
one bot holding all your context, not multi-agent coordination. stateless, but with full Codex + ChatGPT context
2/ openai team said dots is way more ambitious than just a "personal agent".
it runs in the cloud, so you get 24/7 tool use and auto-review, and you can build complex workflows on top of it
3/ the interface feels like iMessage: chat or call your dots. SMS only for now, RCS and iMessage coming soon 👀
4/ biggest shift for me: agents used to need computer use just to see your context. now you invite your dot into Slack or Teams and it's already there
the best personal agent lives inside an interface millions already use,
GPT is one step ahead of 99.9% here.
s/o to the best devx team! @paw_lean @codestantine @victornunez @danizeres and so excited about dots, ultra fast, and decision apis!
12
2
90
6,444
konstantine retweeted
You should come see @codestantine and myself at the Ultrafast booth. We got a really cool game!
5
2
33
2,949
You should come see @codestantine and myself at the Ultrafast booth. We got a really cool game!
5
2
33
2,949
konstantine retweeted
72 hours to OpenAI DevDay.
We’ve been building. Time to show our work.
636
519
10,119
1,464,865
konstantine retweeted
o yes… we’re back in action and we’ll reset usage limits for all paid users across codex and ChatGPT work
sorry about the brief disruption!
(and yes we have a special spare codex when things are down to help us out)
4,230
812
17,749
5,175,607
konstantine retweeted
Sorry Adam, it should be back online now! 🙏
Codex should be coming back online for you now!
Team is monitoring it diligently, apologies for the inconvenience. 🙏
status.openai.com/incidents/…
1
14
1,968