10 appearances on the backlist front page in the last 30 days.
i think this is the reason why the claude desktop app is behind codex, and it's totally fine, i think slack is actually a great way for "teams of humans" to interact with "teams of agents"
to be clear, this is a closed source orchestrator on top of closed source models. if before you didn't control the models, now you don't even control which ones are used or how much. this is not "AI sovereignty" i've also read the tech rep
Editor’s note: imported_from_x_likes
EssentialAI seems to have been acquihired also crazy to find this out from linkedin, i checked and there is 0 information about it (they were valued around ~$1B in August 2025)
this part is actually very interesting, for the mtp head at t+2 they don't include the kv of the indexer of the predicted value at mtp t+1 for efficiency (indexer sharing) AND found that it leads to better results because it avoids training
deepseek v1 -> v3 (no details in v4 about this) and k2 don't use muP and instead use naive N(0,0.006) initialization. so how do they do hyperparam selection? they basically fit scaling laws to get optimal batch size and learning rate. ther
this is the exact same lr schedule as deepseek v2 in terms of when the lr changes the difference is that the second phase does seem to be a slight decay instead of constant, and of course cooldown is also a decay instead of constant
also weird behavior where codex compacted after only ~3min, i don't think this necessarily bad (might be the opposite if it's smart!), just surprised
with claude code and codex shipping similar features, i was curious which one is actually leading so i made a timeline of features they both have like /goal, side conversations, subagents and the new "dreaming" feature https:// x.com/MTSl
they also put dreaming in gray because different product and also date flip if you count "memories" from oai (i'm not sure codex catch the last part btw)
nemotron 3 is significantly less sparse than other models (~10% active vs ~3% for kimi K2/deepseek v4)