W31 · 2026London · New York

Markets Companies Videos Exchange Community Jobs Learn Notebook Lab Research

The Desk Is Live879 companies850 dossiers7/14 topics fresh14 live callsTransformer Successors: State-Space Models and Linear Attention Are Worth WatchingWATCHINGAVGOWATCHING000660.KSBULLISHMUBULLISHNVDABULLISHBTC Miners Pivoting to AI Infrastructure: The Power ArbitrageBULLISHNuclear Is the Only Power Source That Scales for AI Data CentersBULLISH000660.KSBULLISHTSMBULLISHParameter Scaling Is Dead. Multi-Dimensional Scaling Is the New Paradigm.BULLISHInference Costs Will Fall 90% by 2028 — The Memory Stack Is the MechanismBULLISHEnergy Ownership Is the New AI MoatBULLISHLITEBULLISHKV Cache Compression Will Cut Inference Costs 50% Before HBM4 Ships at ScaleBULLISHTransformer Successors: State-Space Models and Linear Attention Are Worth WatchingWATCHINGAVGOWATCHING000660.KSBULLISHMUBULLISHNVDABULLISHBTC Miners Pivoting to AI Infrastructure: The Power ArbitrageBULLISHNuclear Is the Only Power Source That Scales for AI Data CentersBULLISH000660.KSBULLISHTSMBULLISHParameter Scaling Is Dead. Multi-Dimensional Scaling Is the New Paradigm.BULLISHInference Costs Will Fall 90% by 2028 — The Memory Stack Is the MechanismBULLISHEnergy Ownership Is the New AI MoatBULLISHLITEBULLISHKV Cache Compression Will Cut Inference Costs 50% Before HBM4 Ships at ScaleBULLISHStack Watch — free →

Lead StoryMarketsTransforme · WATCHING

The Rung Never Moves

OpenAI shipped GPT-5.6 this morning, and the flagship costs exactly what last generation's flagship cost — $5 in, $30 out, to the cent. The named rungs did not move; they just added a cheaper one below. Read off our own Token Price Index, here is what a frontier that never discounts is telling you.

MenFem Editorial Team·5 min read·Tracked daily, in public

Live CallsTracked daily

TransformeWATCHINGlive

Transformer Successors: State-Space Models and Linear Attention Are Worth Watching

AVGOWATCHINGlive

Broadcom: the right custom-silicon thesis at the wrong price

000660.KSBULLISHlive

SK Hynix: the largest HBM rent, valued like a Korean cyclical

Every call we publish is tracked to its outcome — validated or invalidated, in public. See the record →

Why one master session runs the week

From the NotebookAll entries →

July 15, 2026

Why one master session runs the week

Standard advice says spin up a new AI conversation for every job. I run the whole company from a single chat that opens on Sunday and closes the next Sunday. Here is the model, and the constraint that forces it.

From the LabAll projects →

Building