Alvin Toms

I build AI agents and products, and test how they become businesses. Hermes, BitForge, RoofPrecon, and notes on what works, what fails, and what pays off.

[Alvin Toms @Atoms977]

@daniel_mac8 finally catching up to open source eh? thats good.

[Alvin Toms @Atoms977]

@tonysimons anything cool like this comes up , few days later it get pushed into the main repo. Would you wait or just go for it?

[Alvin Toms @Atoms977]

@alex_prompter all the time, including antigravity. Gemini has a generous usage limits, yet to reach its capacity. So i use it for all kinds of documentation and research work

[Alvin Toms @Atoms977]

@mattpocockuk you might be one of the best person to follow for non coders getting in AI coding.

[Alvin Toms @Atoms977]

@treg_ai @iamlukethedev its a powerful combo

[Alvin Toms @Atoms977]

@DaveShapi I belive thats why we have Buzz for that.

[Alvin Toms @Atoms977]

@Santoshbiswall @rileybrown Its cheaper so it a viable alternative for now

[Alvin Toms @Atoms977]

@thsottiaux @astralarkitekt Look like the issue was codex itself, maybe we need another reset :D

[Alvin Toms @Atoms977]

Hot take: GPT-6 Luna is currently the only model that works as an actual daily driver. What’s the general consensus so far?

[Alvin Toms @Atoms977]

@NousResearch @tonbistudio Man everyone's desktop UI looks much better than mine , what am i missing :/

[Alvin Toms @Atoms977]

@orgo @ycombinator When Omarchy OS on Orgo?

[Alvin Toms @Atoms977]

@tonysimons Finally, something we can use to actually work with, right?

[Alvin Toms @Atoms977]

@orgo when Omarchy on Orgo?

[Alvin Toms @Atoms977]

All I do now is come back to Codex every few hours and just type "Proceed" because Astra keeps my usage going to zero. I have to wait a couple of hours, go click, press, and type "Proceed," and it keeps going again. Five minutes later, I'm out of usage.

Rinse and repeat until my job is done or my entire weekly allowance goes to zero.

Yeah AGI is here

[Alvin Toms @Atoms977]

@_oliver_morley_ yup it should be scrapped or changed, makes no sense for new models like Astra or even Sol

[Alvin Toms @Atoms977]

Astra burned through 100% of my rate limit in a single prompt. The $20 plan limits make long-horizon agentic models almost impossible to use for practical work—having to pause for 5 hours mid-task kills the entire workflow. We need either higher Plus allowances or faster open-source alternatives

[Alvin Toms @Atoms977]

Looks like @OpenAI is being generous with their resets, not gonna hold back on testing Astra.

[Alvin Toms @Atoms977]

Running Astra on some of my existing workflows, let’s see how it goes.