Answerth

[Answerth @AnswerthRS]

@oldstackjournal I log out because unauthorized SSH access to a server is illegal. If I'm granted access, I run`codex` and `claude` to check for an AI cli to do a directory, network scan and assess what's in the environment.

[Answerth @AnswerthRS]

@StefanoGPT @theo No shot they have 5M users at $200/month. I'm not even sure if those 40M users are paying, or if they are bundling in mobile app users via chat/work toggle; my guess is 90% on the $8/20 tier, 6% on the $100 and 4% on the 200

They have over $1 Trillion in spending commitments...

[Answerth @AnswerthRS]

@ApesFinance @thsottiaux Yeah they should just queue up turns. People would probably be mad about it but it's better than just stopping the run

[Answerth @AnswerthRS]

@jiang_kev I need it to bypass robots.txt, will it do it?

[Answerth @AnswerthRS]

@CrabDude @rynorhn I refuse to try dots so that I implement my own proactive agents 😅

[Answerth @AnswerthRS]

@CrabDude @rynorhn Sol 6.1 was a W though. New cost:intelligence ratio

[Answerth @AnswerthRS]

@hhawk @thsottiaux This :/

I would have been fine with my rationed weekly allowance, but I sent off some bigger tasks to (astra ultra, fast) to use it up and now I'm stalled out on regular work

[Answerth @AnswerthRS]

@skyzer4ever @AdamHoltererer @thsottiaux They might be running non-relational like Mongo, but also, it has to distribute across ledgers globally
¯\_(ツ)_/¯

[Answerth @AnswerthRS]

@RamaPuvvula @randomi777 @thsottiaux I would but I'm out of limits and also have 5h30m left until work is over

[Answerth @AnswerthRS]

@CaliberPSC @thsottiaux And it takes what, an hour to roll out on average?

[Answerth @AnswerthRS]

@thsottiaux Managed to burn through the last 70% of my week in a few hours :-) paused work waiting for reset to roll out

[Answerth @AnswerthRS]

@MysteryEthos @thsottiaux Don't worry, you can get the $500 plan for 25x usage

It's only 2.5x the cost and 20% more usage for the plan you used to be able to get for $200 last month

[Answerth @AnswerthRS]

@bertobully @thsottiaux @scottstts I do that, and then my usage quota gets yeet'd and deleted

[Answerth @AnswerthRS]

@tomiyasu16 Waiting on openai to expand voice support for remote devices in the app but 🦞 is a viable workaround for now

[Answerth @AnswerthRS]

@tomiyasu16 Trying to build Jarvis as well. My current approach has been provisioning a server & use openclaw as the voice gateway to orchestrate anything on that server or talk via A2A to other 🦞 instances which themselves orchestrate codex.
Omnidirectional, from any device to any device

[Answerth @AnswerthRS]

@RyzenFire_ @testingcatalog @cerebras True but the 20x plans were 50% off compared to 20/100 plans per token, so they probably looking to avoid that

[Answerth @AnswerthRS]

@bygregorr @Codex_Changelog Grok says
Preserves
• Full conversation hist.
• Drafts + queued prompts
• Clean new thread, original stays untouched
Doesn't preserve
• Guaranteed prompt-cache hits on inherited prefix, though it's been discussed
• Full live tool/shell runtime state

[Answerth @AnswerthRS]

@ChrizBogota @MmuzTW @OpenAI @thsottiaux @sama The government actually requests that they quietly do this if they suspect distillation, which, if somebody is rotating multiple accounts on a proxy for example, could trigger that. The intentionally don't disclose it because if you tell it distiller, you give up the game

[Answerth @AnswerthRS]

@MiddleOfMayhem Multiple psyops can coexist with the presence of off-world originating species and craft

[Answerth @AnswerthRS]

@thsottiaux @stemonte FWIW, there are a lot of traps that burn extra tokens.
If I delegate to tons of sub agents and those are reporting back to Sol, that causes a lot of polling / turns on Sol when each reports back.

I reduced polling freq & use exec -> orchestrator+reviewer -> implementers