Lee Moore

AI Dev Guy, Principal Engineer, Enabler of Agentic SDLC nonsense

[Lee Moore @leegmoore]

@ryanflorence You're almost there. Here's how you do it properly.

/goal Do my job for me. Don't make mistakes.

[Lee Moore @leegmoore]

@thdxr Sounds like the teaser tweet for the new 8 Day Agents product about to drop.

[Lee Moore @leegmoore]

Feels like a primitive is being made the paradigm by a lot of folks. That's been my primary issue with proper ralph loops/control loops. They are a great paradigm for very specific types of projects. But for broader software construction using effective functional and technical specs, they are 1 primitive among a number of orchestration primitives that are generally needed to make up a proper general agentic building process with ai.

So I guess my issue isn't so much with the loops, but the way they loop back around to be the silver bullet every 3-4 months.

[Lee Moore @leegmoore]

@threepointone add me to the list of fuck yeah, please

[Lee Moore @leegmoore]

@jjacky composer 2.5?
gpt 5.4 High fast?
gpt 5.5 Low Fast?

All generally produce better results for less or same cost per task

[Lee Moore @leegmoore]

@daradoescode @maria_rcks @theo about 99.7% of those are being said with deep hatred and contempt. Theo spends far more taking talking about shit he hates than likes.

[Lee Moore @leegmoore]

@trq212 Nice! Looking forward to to it!

[Lee Moore @leegmoore]

@brandon_galang @kimmonismus You spelled scrubs wrong

[Lee Moore @leegmoore]

@DavidKPiano based on what some describe when they say loops they seem to be referring more to state machines than loops. Claude's dynamic workflows seem to be about state machines. I think the phrase state machine sounds too old-timey for folks

[Lee Moore @leegmoore]

@EMostaque Not much. This space is so dynamic that that already falls in distribution of the reasonably possible for me. I'm currently assessing glm 5.2 to see if it can fully replace Opus for me, and so far it's looking damn good on performance and vibes.

[Lee Moore @leegmoore]

@ptremblay Is what true? I'm saying openAI is the best choice if you just have 1 plan because of how good the models and rate limits are so I'm guessing we aren't disagreeing on that

[Lee Moore @leegmoore]

@ZackKorman I'm auditioning GLM 5.2 as well. It's currently slotting in nicely where I use Opus. It's not as chatty or neurotically self referential and doesn't have as many annoying RL tics.

[Lee Moore @leegmoore]

This is a good position for this class of things. And while I don't share Dario's position currently on the risk level, I think it's a reasonable position. And I do think there are cases where "a few people protect the humans from their worst impulses" is justified, it should not be the default. I think the key point here, is is AGI/ASI AI one of those things? I think it's reasonable to believe that it is and to use what influence you have in the world to steer towards the best outcome. Even if you do it badly at times.

[Lee Moore @leegmoore]

I wasn't attacking, just attempting to make sure I understand that you did make a strong point that I didn't acknowledge in my first response and making sure I got it right. Yes the argument is applicable to any tech, it's just that it's currently being applied to AI in this context so that's why I focused on it with that framing.

I'm not saying the position is a master plan to prevent extinction. Just trying to make sure I understand the position as I dont want to strawman or talk past you

[Lee Moore @leegmoore]

@ZackKorman You did say that. So your primary criticism is not the opinion or the implications but that the narrow way in which you think Dario thinks it's best (controlled access by a few) is not the best way to mitigate the concerns. Do I have that right?

[Lee Moore @leegmoore]

@Shinnzo_xd @jullerino any plans to add the ability to go from a codex or claude code session to a t3code session and vice versa. From what I can tell your data projection keeps T3code sessions isolated currently.

[Lee Moore @leegmoore]

@intellectronica opus 4.5/4.6 for coding but not quite as big feeling. I consider 4.6 and glm 5.2 generally better than opus 4.7 and 4.8