Our team of engineers at Parloa stress-tested GPT‑5.1. We re-ran thousands of simulated calls from a real enterprise deployment in the insurance sector, benchmarking our existing setup against the new model. parloa.com/blog/gpt-5-1-rele…

🎙️ Full episode here: open.spotify.com/episode/0Ij…

Just dropped by the AIKetchup podcast! Talked about the Parloa journey and the many lessons learned while building in the AI era. 🎧 Check out the full episode!

Small Language Models: @NVIDIAAI achieved a 20% boost in inference speed using pruning and distillation, with minimal performance drop (MMLU 70.1% → 69.5%). Scaling this to larger models could be a real game changer. Kudos to the team @PavloMolchanov! arxiv.org/pdf/2408.11796

#Apple opens up #Siri for developers trying to catch up with #amazonecho theverge.com/2016/6/13/11916…

This is how self-driving cars would work at an intersection. x.com/yugaishibuchi/status/6…

What is your structure to #prioritize #AB-Test ? Thanks for sharing!

Die netten kleinen Weisheiten von dem #es11 "a fool with a tool is still a fool"

Faltin: "Es ist nicht der Haben-Modus der glücklich macht, sondern der Sein-Modus" bei #es11