iyda (@notiyda) header
iyda (@notiyda) profile picture

i post about games, software, shipping, and whatever breaks along the way. building Justified, a GMod DarkRP server. @OpenAI @AnthropicAI discord: iyda

United Kingdom
Joined December 2012
Photo by iyda (@notiyda)Photo by iyda (@notiyda)Photo by iyda (@notiyda)Photo by iyda (@notiyda)Photo by iyda (@notiyda)Photo by iyda (@notiyda)Photo by iyda (@notiyda)Photo by iyda (@notiyda)Photo by iyda (@notiyda)Photo by iyda (@notiyda)
Pinned Tweet
Codex and ChatGPT having separate limits is actually so slept on. like with Anthropic everything is just one bucket. you use Claude Code, you use chat, you use cowork, it all eats the same limit. you rail one thing and the whole plan is just dead. last time i tested, cancelling even removed Claude Code which is actually insane for a $200 sub. OpenAI at least split it. i can sit in Codex for hours and ChatGPT is still there. or i dump a load of usage in the app and Codex doesnt just die with it. sounds small until you actually live on this shit every day. not saying Codex limits couldnt be better but tibo resets + separate ChatGPT and Codex pools is a godsend. two separate pools is just so much nicer than one shared one you are scared to touch because you might need it later for some basic question. and then with Claude what do you do? you just go to another provider to ask that basic question... Anthropic please just do this. one plan two meters. it is not that deep. Thank you very much to the OpenAI team for these limits being separate on both platforms @thsottiaux @ajambrosino @AndrewGinns shoutout to you folk and the rest of the team
1
1
1
13
11,263
Yup, /btw will help but we are also rolling out a way to see intermediate progress that should land in a couple days. People have been liking it a lot internally. Fair callout as the model is able to go for longer!
6
3
117
2,563
Make Google Antigravity feel like home. 🛠️ Customize your keyboard shortcuts to match your muscle memory. Just head to Settings (or type /keybindings in the CLI) to rebind any command.
45
31
3
695
21,543
Gemini 4 Neon is 100% going to make this a lot better and I can't wait for it but I have a feeling Stitch Loop is going to be waitlist only for a bit and then maybe even USA only then it will be like 2-3 months until we get it :( I am already in love with gemini 4 neon though it's so good
Stitch Loop is coming....... for anyone missing Jules and that autonomous background coding agent that just handled the reviews and security audits in cloud Stitch Loop does that its not available yet, they better ship this asap Gemini 4 Neon would make it even better
1
1
157
I would really love to see Jules or something similar get released so we can automate codebase reviews and security audits automatically utilizing some cloud platform like Jules used to do :( This would be super cool to exist with Gemini 4 Neon RIP jules.google
1
2
939
Using the Gemini Ultra 20x Plan I do not use Claude/GPT Models at all but somehow the usage on them goes down ? unsure how or what is using it but would be interested to know if anyone else has these kind of issues? 🤔
58
Reset everyone’s Gemini quota on Antigravity. The caching improvements have increased the amount you can get done in quota by 22%. Enjoy building with Neon over the weekend!
998
515
488
10,790
696,722
Antigravity feature request: Can we get it so you can middle-click links or right-click them to open in ur default browser? i hate needing to manually copy it and paste it :( @_mohansolo
34
For Antigravity can we fix/improve the steering system to be similar to how Claude Desktop is? I don't wanna be clicking the -> button and it always feels like i am interrupting the model/stopping it when I do it it feels a lot better/more natural if i just type what i want and the model will pick it up after next tool call similar to Claude does it I do know the auto send thing exists but idk still kinda feels like i am interrupting the model? would be nice for it to send and it sends on next toolcall/when it can and it has a button on the message saying "send now" similar to Claude Desktop lol @rodydavis
1
28
Gemini 4 Neon sets a new gold standard for parsing tables and visual layout (bounding boxes) ParseBench-2, our comprehensive benchmark for document parsing. We compared it on low thinking mode against other recent model releases, from Opus 5.5 to GPT-6.2 Sol. ✅ It is the best model across all of ParseBench-2 on parsing complex tables within documents! ✅ It outputs bounding boxes more precisely than any other frontier model ✅ It is quite expensive (6.2c per page) - so would reserve for the heaviest workloads. Come check it out on ParseBench-2: parsebench.ai/
35
12
2
263
17,752
Oh? Antigravity has a new /swarm UI 👀 if you are using Gemini 4 Neon right now PLAESE TEST OUT /swarm antigravity.google/docs/swar… It's amazing
1
69
Please let me turn this on :( it keeps turning off on Antigravity
34
We just added Gemini 4 Neon to the Gemini 4 family. It brings frontier performance across coding, science, and knowledge work at roughly 50–80% lower cost per task than top competing models. See how Neon performs across benchmarks →
95
126
34
3,121
193,923
uhm.... Antigravity? i wanted speech to text not... this Also it kinda seems inaccurate at picking up my voice but yeh...
1
62
Dam... Google Team already added some of the stuff ♥️ Thank youuuuu!!! Antigravity slowly getting better 😃
Replying to @macsdev @rodydavis
It's just browsers feel a lot worse than using it directly inside of an app like the Claude App it just feels clunky/worse 1. Remove 5 hour limit for higher plans 2. Allow us to opt into Beta updates so people can try out new versions of Antigravity and provide feedback 3. Add more connectors (PlanetScale) 4. Have a proper mobile app please for Antigravity 5. Keep limits between Antigravity and the website separate 6. Still some errors/weird UX 7. Just remove Opus maybe even Sonnet from the $20 plan it's pointless as you literally do not/cannot use it for making anything since you just mostly will hit the ratelimit on it and I think personally Neon is more worth to push people towards but yeh 8. Code Mode ? if not already a thing 9. please show the context window?? why is it hidden 10. please make it so when the model is compacting it actually says/mentions it in antigravity since right now it doesn't 11. please make the price when upgrading/buying actually prorated at checkout rather than refunding me later 12. when using the thumbs up/thumbs down it resets if you just go out/in the chat again so it feels like you did nothing 13. Allow us to disable models we don't wanna see this menu just feels a bit clunky ngl with the amount of random models also Orchestrator should just be 1 single feature that you can toggle and then select which models you want to do it with cause i want to do Carbon + Neon with a little bit of Hydrogen 14. Add a proper "Auto Continue" system so when you run out of usage it will auto start after you get usage back 15. for the love of god for @theo please make it so you can hide both ur NAME and Email and any private information (mine is literally showing my FULL name) (same with the QR code) 16. turning on Enable Telemetry literally doesn't work and unselects it's self 17. Welcome back *doxes my full name* please fix this.... it makes it impossible for people like Theo to actually use the app without doxxing his name/email maybe not name but you get the point 18. Auto Archive system to archive old threads/delete them after X days 19. Auto Cleanup Worktrees as well 20. Ability to set the context window that you actually want and set where/when it compacts 21. Please make it so the model can watch/pay attention to PR's for comment and add it so it can auto fix/solve issues that get found by things like @coderabbitai so you can tell it to make a PR and watch it fix all issues and auto merge when finished Probably more that i can find over time
1
13
419
Google Team AntiGravity >> Search Bar Also please make it so you can disable showing ur FULL name and email @theo will love you (INSIDE OF THE MENU NOT IN A CLI COMMAND OR SOME VAGUE STARTUP PARAMTERS) @rodydavis Thanks love you ♥️
1
97
Gemini Neon /swarm mode is fucking amazing Just using usage to test it's capabilities at this point Infinite recursive optimization from Gemini Neon O_o
2
65
Hi Gemini Neon in T3Code 😉 I have my Claude Desktop using the T3Code MCP to interact with it lol
2
146
Couldn't agree more... the team at @GeminiApp @GoogleDeepMind cooked hard with Gemini Neon 4 And to see them constantly iterating and improving it based on user feedback is the best thing Google could be doing right now I really hope it doesn't fall flat and just stop randomly ^-^ Interactions with the community and users being able to see things get improved will always win users over :)
Yes, Gemini 4 Neon is *that* good. I subscribed to Ultra tier. Well done @Google
5
171
iyda retweeted
Gemini 4 Neon, while working on something else just said: "Want me to also investigate the top-right toolbar overlapping Claude's buttons (visible in your screenshot)" I have never seen a model so naturally pickup on little things like that, it's a very good model
74
42
8
2,816
65,415
More info: The hit rate was particularly bad for workloads where there was more time between model invocatikns (cache was evicted). This should fix that and they should see a much higher improvement. We will continue to improve this metric globally, sorry for the inconvenience!
30
9
1
826
63,561