Need better control of each bot's "initialization tree" to better manage context and memory over time for each bot - similar to how claude.md files work with Claude Code (in conjunction with commands). You can really rigorously control the context the coding agent boots up with and in conjunction with commands you can enforce strict gaurdrails and operating hygiene which ensures remarkably consistent and reliable outputs. As is, Grokbot's OS is kind of nebulous and opaque as far as how it manages context and what it remembers and what it consistently boots up into....also you don't really know when that grok session 'started', where's context window is at, what it remembers vs what is compacted away, etc... We need better handle of the bot's 'session boundary' and ability to ensure it is consistent. Memory is everything with these tools. EVERYTHING.
Also the ability to choose what model any given bot uses (by default and per routine)...I suspect it often is dropping to a really cheap/poor model and the output is NOTICEABLE. I will see concerningly inconsistent or bad output. I'd be willing to pay more $ if it means I can ensure higher quality operations.
@DavidAn1611@whitesocksAF Congress can’t sanction killings of people. That’s shit intelligence agencies do and it’s outside the bounds of the law. The point is any legislation congress tries to pass can be vetoed by Trump. It’s a lame duck period, that’s Fuentes’ point. 2028 is all that matters.
Need better control of each bot's "initialization tree" to better manage context and memory over time for each bot - similar to how claude.md files work with Claude Code (in conjunction with commands). You can really rigorously control the context the coding agent boots up with and in conjunction with commands you can enforce strict gaurdrails and operating hygiene which ensures remarkably consistent and reliable outputs. As is, Grokbot's OS is kind of nebulous and opaque as far as how it manages context and what it remembers and what it consistently boots up into....also you don't really know when that grok session 'started', where's context window is at, what it remembers vs what is compacted away, etc... We need better handle of the bot's 'session boundary' and ability to ensure it is consistent. Memory is everything with these tools. EVERYTHING.
Also the ability to choose what model any given bot uses (by default and per routine)...I suspect it often is dropping to a really cheap/poor model and the output is NOTICEABLE. I will see concerningly inconsistent or bad output. I'd be willing to pay more $ if it means I can ensure higher quality operations.
I was there. It wasn’t the counter protest. The event was just poorly organized (not throwing shade it was thrown together last minute). It was a combination of not being garaunteed a spot (no tickets), not being allowed to camp the night before, zero parking anywhere near the event meant people parked far away and stayed up all night sitting in cheap chairs waiting to get a spot. Also there were zero trash cans anywhere to discard anything, even at security checkpoint - you weren’t allowed to bring anything in and they were just stacking confiscated stuff up on the sides (crazy). Offerings tickets would’ve guarantee people a spot and would’ve meant fewer people hanging around the arena for so long before the event. Also there should’ve been trash cans all over the place. Simple oversights that created that situation. That said the amount of litter that was left behind by over 200k people was astonishingly low. That spot in particular is just where many of the over nighters were waiting.
@thinkfree101@JoshMandell6 Josh made more than most on the MSTR train, more than you or I will ever see in our lifetimes (he’s published receipts along the way). Contend with his ideas, not his alleged motivations.