cmpgervais

Co-Founder @ Enertel AI

[Kevin @linguinelabs]

If Canada made https://t.co/NwyRWAS8eb it would be called https://t.co/i6DK1zVxDF

[cmpgervais @cmpgervais]

@internetvin well, what did they buy?

[cmpgervais @cmpgervais]

@Thom_Wolf What happens if you add Ontario though? 🤭

[cmpgervais @cmpgervais]

@yacineMTB this but it's astra + fable, the simulators are all grid problems, and the architecture is a GNN

[cmpgervais @cmpgervais]

“it’s microduck but for zambonis and i call them microbonis”

[cmpgervais @cmpgervais]

@yacineMTB if I build a zamboni robot this winter will you sim2real it

[cmpgervais @cmpgervais]

@internetvin I test drove 14.3.8 in Kingston today during Queen’s move in - blew me away. Ordered the Model Y as soon as I got home.

[cmpgervais @cmpgervais]

@martianwyrdlord @yacineMTB co(ugh)ere

[cmpgervais @cmpgervais]

@yacineMTB kids are smart, this policy is not

[cmpgervais @cmpgervais]

@yacineMTB counter point: the worst alcoholics in first year university had parents who never introduced them to alcohol

[cmpgervais @cmpgervais]

@yacineMTB cold weather drones, make it a thing

[cmpgervais @cmpgervais]

@buildcanada @yacineMTB electricity imports have nothing to do with oil exports or west-east pipelines and everything to do with marginal costs

[cmpgervais @cmpgervais]

@Matt__Miller :)

[cmpgervais @cmpgervais]

@internetvin VinDictator.

[cmpgervais @cmpgervais]

@kalomaze @yacineMTB the codex vs claude debate is a skills issue, ie if you give them the same skills files and force them to use TDD both are more capable than their users

[cmpgervais @cmpgervais]

@jm_alexia Congratulations! Hope it's still in the Montreal region!

[cmpgervais @cmpgervais]

@omarsar0 @yacineMTB another for the scratch pad pile

[cmpgervais @cmpgervais]

@_arohan_ @samlakig cc seems to take this approach without being explicitly told, but i've noticed on the initial plans / correctness reviews it typically spawns one agent per specialized task instead of having many agents tackle the full plan / correctness review

[cmpgervais @cmpgervais]

@yacineMTB this is true regardless of how the grid simulator is obtained, fwiw

[cmpgervais @cmpgervais]

@yacineMTB a note from the power sector, if the physics is well-understood and the objective is clear, it's easier (read: faster) to stuff them both into a loss function and train directly via SSL than it is to build a full grid simulator and try to use RL to recover the desired behaviour