Home
Welcome
About
Profile & credentials
The Casebook
Real case studies with measured outcomes
Product Playground
Try a concept. Play with a product.
The Radar
Insights on AI in logistics
AIHot
Real-time AI news feed
The Manifest
Daily AI × supply chain signal
August 9, 2026 · 16 stories

The Manifest

Daily · AI × Supply Chain

Google folds DeepMind into its core org as Hassabis eyes an exit, AI's power demands push Nvidia and Amazon into multibillion-dollar energy infrastructure bets, and shippers keep shifting to intermodal even as UK courts drown in AI-drafted lawsuits.

Get The Manifest in your inbox

One sharp digest of AI × supply chain. Free. No noise.

Today's Top

  1. 01Google dismantles Deepmind and bets on a fresh start as Hassabis heads for the exitThe Decoder
  2. 02AI's energy appetite drives Nvidia and Amazon to pour billions into massive power infrastructureThe Decoder
  3. 03AI is flooding Britain's employment courts with lawsuitsThe Decoder
  4. 04DeepMind's hurricane breakthrough has surprised weather scientistsArs Technica AI
  5. 05Modal shift dampens trucking marketFreightWaves
01

Models & Releases

5 stories

Google's DiffusionGemma proves you don't need to train from scratch to build a text diffusion model

Retrofitting an existing model into a diffusion architecture at a fraction of the training budget points to cheaper paths for faster inference. It still trails on reasoning benchmarks, so treat this as a speed play rather than a quality upgrade for now.

The Decoder↗ source

Meet Shepherd: An Open-Source Python Substrate That Lets Meta-Agents Fork, Replay, and Revert Any Agent Run

Long agent runs accumulate state no transcript captures, and Shepherd lets teams fork and revert without re-paying the full token cost of a restart. Useful for any team running agents against real codebases or live systems.

MarkTechPost↗ source

Pokee AI Releases Pokee-Isaac 28B: A 10M-Token Context Agentic Model Built to Run Inside the Customer Boundary

A 10-million-token context window that holds accuracy where every rival flatlines past 2M matters for document-heavy work like contract review or long procurement histories. Running inside the customer boundary is the part buyers with data governance requirements will care about most.

MarkTechPost↗ source

Anthropic sets Claude Code to Auto Mode by default to protect developers from bad approvals

Anthropic's own tests show a classifier catches far more dangerous commands than human reviewers do, which is a quiet admission that manual approval was mostly theater. Expect this shift from writing code to supervising agents to keep spreading beyond coding tools.

The Decoder↗ source

Claude Code sessions can now talk to each other and share context across terminals

A small feature with big implications for how teams parallelize agent work on macOS and Linux. Worth testing before scaling any multi-agent setup, since coordination bugs are easy to miss until several sessions are running at once.

The Decoder↗ source
02

Supply Chain & Ops

3 stories
03

Deals & Market

2 stories
04

Research & Frontier

3 stories
05

Org & AI Architecture

3 stories
Past issues
The Manifest · 2026-08-09Follow on LinkedIn