Six research systems from 2026 treat the agent harness as a separately optimizable artifact alongside model weights, cutting inference cost by up to 98.6% and raising agent success rates.

Six systems that decide what long-horizon agents are allowed to forget: ACON, Self-GC, StateComp, Forget Reasoning, PaMER, and DRSR — with exact numbers and design tradeoffs.

From separate GPU pools to different model checkpoints and custom silicon, five systems show prefill and decode splitting into distinct computational products.

Five systems let clients verify that a remote LLM ran the claimed model on their prompt, without re-running inference or taking the provider at their word.

Memory Attention replaces the value projection with token-indexed lookup plus key addition, improving LM perplexity while allowing model tables to live in CPU memory.

Google DeepMind added server-side memory to Private AI Compute; keys stay on user devices and inference runs inside hardware-isolated cloud enclaves.

AlloyDB routes agent queries through ephemeral microVM nodes on dedicated Colossus storage, scaling to 1,000 nodes with zero measured primary cluster impact.

Lovable separates agent history from model context using an immutable event log, enabling free forking, async compaction, and cross-deploy continuity.

Researchers at Tel Aviv and Stanford show two randomly initialized models can self-improve across text, images, and audio with no human training data.