davepoon

Build AI/non-AI stuff | | | Peel Banana ( | OpenClaw (

[davepoon @davepoon]

Inspired by the incredible real-time AI vision demos and ngxson's live SmolVLM + llama.cpp project, I wanted to see what was possible on Apple Silicon! 🍏

Super excited to share my SmolVLM Real-Time Webcam Demo with MLX-VLM! 😙

[davepoon @davepoon]

@Ansh_096 @0xRapid @TimJayas @xai ChatGPT Plus still seems to offer better value than that.
Great info, thanks!

[davepoon @davepoon]

How's ChatGPT Pro so far? 😄
5× Codex usage now, not 10× anymore. Worth it?

[davepoon @davepoon]

@koltregaskes That's right.

https://t.co/Ds8PbLrLh3

[davepoon @davepoon]

Claude Design now shares usage limits with https://t.co/LJTHBppuvg and Claude Code.
It's no longer a separate usage limit.

[davepoon @davepoon]

@mattpocockuk Haha, I made the same mistake a few times today.
I should chant it like, "Mirror, mirror, on the wall, enter the workflow mode, and…"

[davepoon @davepoon]

@nesquena It’s incredible, thanks for sharing it with the community. 😃

[davepoon @davepoon]

Spent some time packaging Nous Research's Hermes Agent as a one click Railway template. A few things I didn't expect to learn along the way.

[davepoon @davepoon]

@jesselarose interesting, thanks

[davepoon @davepoon]

Claude Code remote control keeps quitting on me.
The worst part is once it quits, I can't reconnect remotely anymore.

[davepoon @davepoon]

@pbakaus Loving it!! Thanks for all the hard work.

[davepoon @davepoon]

Claude Design now shares usage limits with https://t.co/LJTHBppuvg and Claude Code.
It's no longer a separate usage limit.

[davepoon @davepoon]

@ivanzugec @pbakaus @impeccable_ai Yeah, give it a try when working on frontend tasks.

[davepoon @davepoon]

It gives you 24/7 cloud uptime without leaving your computer on, and it's super cheap ($8-10/mo) while scaling up to 48 vCPU/48GB RAM.

Best of all, you can pair it with your other Railway services so you can more services to integrate with the Hermes Agent, and also you can allow can Hermes Agent autonomously manage your web apps, automate workflows, and handle customer emails.

[davepoon @davepoon]

@zentalksai It’s been available for a while now.
Or try https://t.co/VBoCkqJNdn. It works really well too.

[davepoon @davepoon]

If you want to poke at it:
https://t.co/gyXnuQ7D52

[davepoon @davepoon]

Headless OAuth was the other one. Browser callback flows can't survive in a container.
Device code (RFC 8628) running in an xterm panel does. Same flow you'd use signing in on a TV.

[davepoon @davepoon]

@fayazara Great work!!

[davepoon @davepoon]

GPT-5.5 feels nerfed. Less sharp, more mistakes, and sometimes it just gives up outside plan mode