sittingmongoose

[sittingmongoose @sittingmongoose]

@theo @FrameworkPuter There are no chargers that I’m aware of that can do more than 140w currently.

[sittingmongoose @sittingmongoose]

@theo Have to start somewhere, it all trickles down. I’m working to fine tune models that run on 24gb for specific processes in my orchestrator. I imagine in a few years that runs on 8gb.

[sittingmongoose @sittingmongoose]

@theo @tenobrus I think opus 5 is horrible, but it has its strokes of brilliance. There is a good model in there, if only Anthropic knew how to bring it out. It caught a few impressive things today for me…as it introduced countless bugs then fixed them :|

[sittingmongoose @sittingmongoose]

@theo My only complaint is qwen 3.8 max isn’t in it, which I would put as b tier. It’s good all around(a little below k3) but it’s by FAR the most creative with front end. Way beyond fable. It’s also cheaper than k3.

[sittingmongoose @sittingmongoose]

@aerion586915 @SPAC89 Opencode zen and go have it for free for the next 6 days.

[sittingmongoose @sittingmongoose]

@hadesdevs @opencode Most of the Chinese models are distilled from the American models, so they will act or say things exactly like them.

[sittingmongoose @sittingmongoose]

@hadesdevs @opencode This does not work. Models don’t know their own models.

[sittingmongoose @sittingmongoose]

@theo I think part of the confusion is most harnesses have editors built in. CC, Codex, etc. many of your “competitors” have harnesses built in like opencode. You don’t modify the prompt at all right? You just dump directly into the cli of the harness.

[sittingmongoose @sittingmongoose]

@ccb881096 @thsottiaux Use fast mode now then

[sittingmongoose @sittingmongoose]

@richiemcilroy @cap Have you looked into slint? It’s a light weight front end with native rust support. It works natively on Mac, windows, Linux and has wasm support, I’m using it now for a project, it’s really well supported.

[sittingmongoose @sittingmongoose]

@theo @TheAnirudh Btrfs has a lot of issues. It’s been pretty well documented for a while. Oddly, if you look into Unraid, you will see a lot of it clearly. For a while they were defaulting their cache drives to btrfs, and that caused some mega problems for people.

[sittingmongoose @sittingmongoose]

@theo @bhanu4417 They recently added a lot to headless mode. It should be doable now, it’s not acp but you can still control it.

[sittingmongoose @sittingmongoose]

@LoganHowe @abacaj Fine too. I would be fine with 5.6 sol if it was faster and cheaper. But waiting forever and burning all my tokens is just too much combined.

[sittingmongoose @sittingmongoose]

@thsottiaux @rezoundous Curious, do you try the competing models out? I tend to try all of them. Opus gets a lot right, but it also seems to forget, or find issues and not fix them, or just stop. It’s been doing all those things a lot and I’ve seen similar online.

[sittingmongoose @sittingmongoose]

@GohilHardy Because Claude and codex are subsidized and give you far higher limits for their models. Cursor is just using api pricing. Though they do give you a subsidized pool for grok and composer.

[sittingmongoose @sittingmongoose]

@OmedVibeCodes Have you tried Kimi k3 or grok 4.5 on it?

[sittingmongoose @sittingmongoose]

I love how ChatGPT Work creates sites so easily to share!

[Greg Brockman @gdb]

Was very cool to hear about the reasons people love Sol. We're doing the promotion again, except this time for ChatGPT Work:

Tweet what you love about ChatGPT Work, claim $100 in free credits, get more work done.

First 10k get the free tokens: https://t.co/w7QrPBYkrh

[Greg Brockman @gdb]

Was very cool to hear about the reasons people love Sol. We're doing the promotion again, except this time for ChatGPT Work:

Tweet what you love about ChatGPT Work, claim $100 in free credits, get more work done.

First 10k get the free tokens: https://t.co/w7QrPBYkrh

[sittingmongoose @sittingmongoose]

@ohjeyburner @qwen_cloud When you sub, and login, it shows which models are available and with what capabilities. Glm and deepseek do not have vision.

[sittingmongoose @sittingmongoose]

@Otto_x090 They don’t need fresh water. They are using it because it’s faster and easier to spin up. They are taking a lot of short cuts like using generators instead of really power sources and not insulating HVAC systems which makes them super loud.