Long Luong

Founder, CTO at Vexere.
Fullstack Expert - ReactJS, NodeJS, C/C++, C#, Java, php, AI, deep learning, image processing, system & devops

[Long Luong @longbkit]

For the past two weeks, Claude has been helping my productivity way more than Codex. It seems Anthropic has made some solid improvements with Opus 5. With Codex, I constantly get 'Model is at capacity,' and when it does work, it either stops halfway or asks too much

[Long Luong @longbkit]

@MattPRD @Muse I'm interested.

[Long Luong @longbkit]

@0xSero How well does it scale with concurrent sessions?

[Long Luong @longbkit]

@thsottiaux [System Error] Selected model is at capacity. Please try a different model.

Still unusable for the last 3 days. Could you track this error correctly on https://t.co/W4EIzcs6rY

[Long Luong @longbkit]

Experiencing a significant degradation in the intelligence of gpt-6-astra and gpt-5.6-sol. Who else agrees?

[Long Luong @longbkit]

@JacobAllenMedia @thsottiaux Same here. Codex keeps returning the 'Selected model is at capacity' error and has been completely unusable for the last 36 hours. The uptime report should really be updated to reflect this. I've had to switch to Claude in the meantime.

[Long Luong @longbkit]

@JensenHuang @ChaseLochmiller @OpenAI Looks like OpenAI needs more GPUs for inference. I've been experiencing server overloads since yesterday, unusable.

[Long Luong @longbkit]

@KyleHessling1 Awesome work! Could you share how it achieves this without degrading the base model? Also, are there specific trade-offs or domains where it excels vs. struggles?

[Long Luong @longbkit]

Robot hands. This is crazy!

[Space and Technology @spaceandtech_]

Xynova has introduced Prima 1, a new robotic hand with 22 degrees of freedom for advanced movement and gripping.

It combines vision and touch sensors to better understand and handle objects.

With direct-drive motors, it can move quickly and lift objects weighing up to 20 kg.

[Long Luong @longbkit]

@KyleHessling1 This can be a functional product that you can use as a task management tool for human and AI to work together btw

[Long Luong @longbkit]

@KyleHessling1 I prepared this to test Qwen 3.8 27B with different harness (Claude, Codex, OpenCode, Pi), and with frontier model (Claude Code with Fable 5.1 and Codex with GPT-5.6-Sol) to see how close they are (but does not proceed yet) https://t.co/cFdpufKc1H

[Long Luong @longbkit]

Wow so cute! We know the robot generation is coming soon but I didn't expect to be this early.

[clem 🤗 @ClementDelangue]

https://t.co/KXKqDSkNJj

[Long Luong @longbkit]

@ItsmeAjayKV That is huge. 1.22TB/s of bandwidth seems to match the rumored Apple M5 Ultra coming this year.

[Long Luong @longbkit]

@bil0090 Oh really. Definitely will try it next time. Thank you.

[Long Luong @longbkit]

@bcherny Sure, I could just use another account, but staying silent doesn't sit right with me.

We are stepping in an AI-driven world, and this kind of cold, automated judgment and unchecked power is something nobody wants to see happen.

[Long Luong @longbkit]

@bcherny I was banned 6 months ago with no clear explanation, and my account is still locked. How would you feel if you got completely banned from a community over just one mistake?

[Long Luong @longbkit]

@Zadge16 @thsottiaux @abraibrai Just got the bank reset. Thank you!

[Long Luong @longbkit]

@Zadge16 @thsottiaux @abraibrai me too, didn't receive the bank reset. Can you fix the root cause? cc @thsottiaux

[Long Luong @longbkit]

@thsottiaux Can you make this in Codex?

[Long Luong @longbkit]

This is exactly why I love being an engineer.

[Charlie Marsh @charliermarsh]

In the latest Codex CLI release, I redid the lifecycle to make `codex` startup instant.

It's now ~25x faster and immediately responsive.