Mask

staff ai-ml eng, production breaking caffeine driven life form, built kitedb, aci, minidex, heimdaal

[Mask @DeveloperMask]

experiments for "agentic swarms"

opencode uses more memory than I like. .. i need to figure out why...

pi is pretty good, but the per session addition is a bit high?

muzen is where I want memory to be, but its also less capable, ill make it better tho

[Mask @DeveloperMask]

@ibuildthecloud as a person that spent a whole year working with Lit for 20k devs to use, its not a terrible concept but boy are they annoying to work with

[Mask @DeveloperMask]

@arXmedes @cb_doge their data is worth 60B alone, how do you think modela are getting better at code

[Mask @DeveloperMask]

@wesbos i get inconsistent resulta with gpt images 2...

sometimes it does remove the bg

[Mask @DeveloperMask]

@theo as with anything with the gov, it takes too long

[Mask @DeveloperMask]

@IterIntellectus goodbye permanent underclass

[Mask @DeveloperMask]

@RhysSullivan oai did say they had better guardrails in place

[Mask @DeveloperMask]

@robinebers https://t.co/Xftp9ZI8h3

[Mask @DeveloperMask]

PRO TIP:

Tell codex to use chrome and consult gpt pro, for planning.

[Mask @DeveloperMask]

10 sessions

[Mask @DeveloperMask]

experiments for "agentic swarms"

opencode uses more memory than I like. .. i need to figure out why...

pi is pretty good, but the per session addition is a bit high?

muzen is where I want memory to be, but its also less capable, ill make it better tho

[Mask @DeveloperMask]

does opencode really need 300mb ram iddle?

[Mask @DeveloperMask]

@tibor_tee Bugbot is better, i have more things to improve

[Mask @DeveloperMask]

@luisllanes_ I used both 4.8 in claude code and 5.5 in codex, and gpt annoys me less than opus, not to mention gpt is faster and cuts less corners

[Mask @DeveloperMask]

@jxnlco did you watch wistoria?

[Mask @DeveloperMask]

idk, I've been sending same prompt to 4.8 and 5.5 and, opus is giving dogshit answers...

[Mask @DeveloperMask]

@robinebers now we know how the gpt-4o people felt lol

[Mask @DeveloperMask]

@kr0der Wasnt 5.5 an early checkpoint of the new base?

[Mask @DeveloperMask]

@LexnLin it was fable, falls back to 4.8 and 5.5

[Mask @DeveloperMask]

@yacineMTB sad but true, i have seen gpt 5.5 test a prompt by checking if the string contains the prompt text... these things are bad