unreal that grok is only 2 years old and did this from a standing start. i really want to tell this story on @latentspacepod . short writeup below
@swyx
-
Grok’s Voice Mode: Why SuperGrok Subscription is Worth Keeping
By
–
lowkey the best thing about @xai is the Voice Mode. I got SuperGrok just to test out Grok 4 and do KaraokeBench on Grok 4’s Eve Voice (the one that does opera) but actually I’m going to keep SuperGrok around because its actually just a much much better voice mode than
-
Prompt Engineering Evolution: Context Engineering and Intent Encoding in 2025
By
–
i think Prompt Engineering is "growing up" in two different ways in 2025. the first is "Context Engineering", which is well covered by @dexhorthy
's talk (now >100k views!) but Prompts = Intent + Context and as @sgrove points out the task of encoding intents/goals/principles -
Grok 4 Voice Fails All Karaoke Benchmark Tests
By
–
Grok 4 voice scores 0/7 on karaokebench pic.twitter.com/hZkDgC9fIj
— swyx 🐣 (@swyx) 11 juillet 2025Grok 4 voice scores 0/7 on karaokebench
-
Metr’s AI Interview Technique Emerges from Attempted Criticism
By
–
metr tried to dunk on ai but accidentally found a great way to interview ai engineers
-
Instructor vs Inference Provider Layer Support for Structured Output
By
–
does this problem "simply go away" if you use something like instructor? why must be supported at the inference provider layer
-
Training LLMs on AI Paper Tweet Summaries
By
–
train LLM on all AK tweet summaries of papers @_akhaliq @arankomatsuzaki and other AKs
-
Merged Experts in MoE: Coordination Without Joint Training
By
–
all experts are merged into an MoE. The anchor model ensures that all experts learn to coordinate, even though they are never trained on the joint dataset. this is rocket science wtf
-
Updated dev cycle with chart on Moloch hyperstitions
By
–
updated the dev cycle with chart https://
latent.space/p/hyperstition
s-of-moloch
… -
Community Notes Needed for LLM Benchmark Misleading Claims
By
–
we need community notes for llm benchmark porn