FaithFellowshipFun

[FaithFellowshipFun @FellowshipFun]

@CanadianPM Ask Venezuela how that whole 'We call it Gulf of Mexico' worked out for them...

[FaithFellowshipFun @FellowshipFun]

@MapQuest It's ok - we are!

[FaithFellowshipFun @FellowshipFun]

@superalesha @superalesha
From your series of tests - would you say that Q3 is pretty much equivalent to Q4 overall on equivalent thinking modes??

[FaithFellowshipFun @FellowshipFun]

@CrazyRetroGame1 I started a 'solo-mmo-project' - not to launch a game, but to 'learn how to scale' - so that means I try to do it all via AI with minimal direct code/input from myself.

Learned a TON about ai-coding properly (validation means EVERYTHING) - and now trying to do iterations cleanly

[FaithFellowshipFun @FellowshipFun]

@0xBakeer @Alibaba_Qwen @QwenDevs I don't have the hardware myself - but sounds like a COOL experiment - and would be IMMENSELY useful to folks in the <32gb +64Gb ram local enthusiast scene.

(Shameless plug hehe)

[FaithFellowshipFun @FellowshipFun]

@0xBakeer @Alibaba_Qwen @QwenDevs @0xBakeer - saw your work here: https://t.co/dhQQJm8Z0A

Have you considered combining that with:

Flash-Next tensor census + layer-wise REAP calibration + 25/37.5/50% pruning sweep - and seeing if quality holds?

Possible to draft a 24-32Gb derivative w/ 95%+ retained?

[FaithFellowshipFun @FellowshipFun]

@ForwardEditor I have found Sol-Light to be extremely effective. I personally have a /codex-swarm skill that uses Luna-Max as orchestrator -> spawn Luna-Max workers -> Use Sol-Light as validator, plan steering, and 'intelligent sage'

Really good results so far in efficiency and performance.

[FaithFellowshipFun @FellowshipFun]

@thsottiaux @hqmank The real pain will be that when we know a reset is coming (pre-announce - or even when it happens) - our ability to use our limit to maximize reset effectiveness is going to be stopped.... Unless all resets moving forward are banked

[FaithFellowshipFun @FellowshipFun]

@Im_IrushiK The Luna-Max suggestion is the best to work with these $20 plans (or business).
Created a skill called /codex-swarm - and it uses Luna-max as an orchestrator and tells it to spawn other luna-max workers - and if it uses my workflow with validation Sol-Light -> ReadOnly validation

[FaithFellowshipFun @FellowshipFun]

@superalesha Good luck brother - love seeing this type of innovation and optimization attempt!

[FaithFellowshipFun @FellowshipFun]

@filip0x1 @tripoai Humanoid animations are almost always great - and is what all demos show.... But quadrupeds and anthropods etc... No one has good stock animations for them sadly ...

[FaithFellowshipFun @FellowshipFun]

@EsfandTV hahahaha! You are so awesome brother!

[FaithFellowshipFun @FellowshipFun]

Are you telling me that we not only have 5h limit back - BUT that 1% of my weekly use is 7% of my 5h?

What? Why?

[FaithFellowshipFun @FellowshipFun]

@Xeratricky Any game that is 3rd person but I can't 'center' that and have to operate from a shoulder is 100% no go for me - never feels anchored right - and always makes a default 'corner' advantageous... But I'm not a shooter fan anyway so not sure I count =)

[FaithFellowshipFun @FellowshipFun]

@thsottiaux @thsottiaux is it correct that a multi-seat BUSINESS account is just a glorified Pro account? I'm seeing the 5h limit - and that will necessitate pulling this and going with an alternative (like Cursor/Grok)

[FaithFellowshipFun @FellowshipFun]

@DogukanUrker @ornith_ I made a small mod to my llama.cpp and proposed the change in this PR: https://t.co/SDEZoeoU1M

I am running it locally - and it helps with the overthink - would love to have a solid set of tests like this w/ this mod. =)

[FaithFellowshipFun @FellowshipFun]

If you use Codex - use Luna-Max with explicit instructions to spawn other luna-max sub-agents for work + Sol-Light for validation/planning (but never execution - must defer those back to Luna workers) - to be a more efficient strategy than Sol Orchestration w/ luna workers.

[FaithFellowshipFun @FellowshipFun]

@ornith_ 35b MoE is a champ! Thinking is great, but @ornith_ can 'overthink' - HOWEVER - llama.cpp custom mod can help 'blunt' that without the hard cut weirdness from `reasoning-budget` if you do a small mod to adopt some of the changes here: https://t.co/i3yaCAEduE

Should be good!

[FaithFellowshipFun @FellowshipFun]

@Beno10_MFC Top angle: 60
Given angle: 50
Triangle is 180 - (60 +50) = 70

Inside angle near k: 70
Line = 180
k = 180 - 70 = 110

I believe it is 110 degrees

[FaithFellowshipFun @FellowshipFun]

@0xLuciferCore my thoughts went: 59 -> 39.... BUT....

-54 by putting pencil in FRONT on the number is technically smallest I think...