Tech • AI • Robotics
Boston Dynamics Atlas battery swap, Grockbot agents, brain organoids
Briefing
Atlas gets autonomous battery swaps
Boston Dynamics has given its production Atlas humanoid autonomous battery swapping, a practical upgrade aimed at factory uptime rather than spectacle. The robot can detect low charge, walk to a station, replace its own battery, and return to work in under 3 minutes. That compares with roughly 90 minutes for a standard recharge, a major improvement for a platform with about 4 hours of typical runtime and closer to 2 hours under heavy loads. The move strengthens the industrial case for humanoids by attacking one of the biggest barriers to shift-length utilization.
Grockbot pushes multi-agent office automation
Grockbot launched as a workspace for building and managing specialized AI agents across desktop and mobile. Its pitch is persistent, collaborative automation: bots for inbox handling, health tracking, and content operations can share context and workflows inside one interface. The system connects to tools including Gmail, ClickUp, and Slack, and can create invoices, generate task records, draft emails, and track payment status. By bundling these actions into reusable routines, the platform targets a step beyond chat toward operational autonomy.
Virtual computers expand agent reach
A notable Grockbot feature is the use of per-agent virtual computers, allowing automations to operate software even when no native integration exists. That broadens the scope from API-first tools to ordinary desktop workflows, a key limitation for many current agent products. The platform also supports scheduled and event-driven actions, such as daily inbox reviews at 8 a.m. or triggers from new Slack messages. With Zapier access via MCP, coverage extends to a much larger app ecosystem.
Brain organoids edge toward computing
Human brain organoids are moving beyond disease modeling into experimental computing research, opening both scientific promise and ethical controversy. Researchers can reprogram adult cells from skin, blood, hair, or teeth into induced pluripotent stem cells and then grow millimeter-scale brain-like tissue. These organoids can contain around 5 million cells, including roughly 2.5 million neurons, enough to study neural development directly. The same biological complexity that makes them useful also sharpens questions about sentience, moral status, and research limits.
Organoids show infant-like oscillations
When maintained at about 98.6°F for roughly eight months, some organoids begin producing repetitive electrical oscillations resembling patterns seen in premature infants. Researchers do not treat that as evidence of consciousness, but the similarity has intensified debate over what these systems represent. The signals suggest a level of organized activity beyond simple cell cultures while remaining far from a functioning human brain. That gray zone is rapidly becoming the central policy and bioethics challenge in the field.
UC San Diego scales organoid research
At UC San Diego, developmental biologist Alysson Muotri is producing brain organoids by the tens of thousands, turning the technology into a scalable research platform. The lab has used them to study autism, including tissue derived from autistic donors, as well as more speculative lines such as Neanderthal-like organoids and exposure to space radiation. The appeal is twofold: closer modeling of human neurodevelopment and a potentially faster path for screening therapies. If validated, organoids could become an important bridge between animal studies and human clinical research.
Hyperscalers pour $700 billion into compute
Amazon, Google, Microsoft, and Meta are projected to spend more than $700 billion on AI infrastructure this year, underscoring how model serving has become a capital-intensive arms race. That total is roughly four times the level of four years ago and nearly 80% above the prior year. McKinsey expects almost $7 trillion to flow into data centers by 2030, spanning compute, memory, networking, and operations. The spending surge is remaking demand for engineers skilled in GPUs, model servers, and Kubernetes-based orchestration.
Antimatter bets on energy and routing
Antimatter, launched in April 2026, is arguing that future AI value will shift away from proprietary models and toward infrastructure, energy access, and compute routing. The company traces its roots to Hivnet and a series of pivots from distributed storage to distributed computing, then further toward physical capacity and power. Its thesis hardened in summer 2024, when the team proposed containerized infrastructure packed with GPUs, CPUs, and memory. The strategy is explicitly contrarian: win on lower-cost, globally distributed capacity rather than on frontier-model branding.
Videos covered
They Just Made a Chip With Living Human Brain Cells
- •From skin cells to brain tissue
- •Brain waves that resemble early development
- •UC San Diego’s organoid work
ChatGPT Now Watches You... Claude Adds Watermarks (and more)
- •Grockbot launches as a multi-agent workspace
- •Agents can complete business tasks across apps
- •Virtual computers and event-based routines expand automation
Boston Dynamics New Atlas Upgrade Makes It Fully Autonomous
- •Autonomous battery swaps cut downtime
- •Designed for factory handling
- •Runtime remains a core economics issue
Cloud, IA, Énergie : des Frenchies contre l’empire américain !
- •Investor trust over market validation
- •From distributed storage to AI infrastructure
- •A contrarian thesis on timing
AI Infrastructure Explained (GPUs, vLLM, and LLM-D)
- •Spending surge reshapes infrastructure demand
- •Models are software structures plus learned weights
- •Why laptops struggle and GPUs dominate
Un pays qui s'embrase, une caste qui s'embrasse... Avec Alexis Poulin
- •A village buried in minutes
- •Heavy human toll
- •A preventable industrial failure