infrastructure

Discusses why Linux is an ideal operating system for running AI agents.

Discusses the ongoing fragmentation of GPU access in 2026 and questions whether a better market structure for compute could emerge.

a16z published an in-depth article arguing that SpaceX is turning space into a key part of AI infrastructure, because the bottleneck for AI is energy, and orbit, the Moon, and Mars will become future data centers.

The author describes a practical AI agent setup where the agent (Hermes) accesses infrastructure only through an access layer (Teleport) that enforces the human user's identity and permissions, arguing that useful agents must operate within existing permission models rather than having their own god-like access.

A reporter visits the Rogfast tunnel project in Norway, which will be the world's longest and deepest subsea road tunnel, describing the engineering challenges and the work environment.

A speculative analysis argues that if AI capabilities plateau and become a commoditized utility, China's ability to rapidly scale energy infrastructure and produce cheaper tokens could allow it to dominate the global AI market, paralleling the offshoring of manufacturing.

Programmers in the community see Hermes as the standard base for next-generation AI Agents, introducing multiple related projects such as Honcho (memory engine), Hermes Web Search Plus (intelligent retrieval), NemoClaw (enterprise-grade expansion), etc., aiming to provide persistent memory and structured capabilities for Agents.

Proposes using desalination integrated with data center water cooling systems to address water usage concerns, with the resulting fresh water added to municipal supplies.

A personal reflection from a UCLA dropout startup founder on how building AI infrastructure changed his perspective on the world's imperfect foundations.

Ivan Burazin argues that the name 'Sandbox' is misleading for their product, which is actually a composable computer that allows AI agents to define their own hardware configuration on the fly—CPU, RAM, disk, GPU, and OS—assembled instantly.

SemiAnalysis refutes the widely circulated claim that half of 2026 US datacenter capacity is delayed or canceled, attributing it to inaccurate models and showing their own data indicates only minor changes in forecasts.

Baseten, a $13 billion AI startup, provides software and computing capacity to companies using lower-cost AI models as alternatives to OpenAI and Anthropic.

An analysis arguing that companies fail at AI because they focus on the model rather than the foundational layers—process design, governance, knowledge architecture, human judgment, and feedback loops—which are the true sources of value. The article cites Nadella's 'token capital' concept, Apple's model-swappable Siri, and survey data showing a wide gap between strategy and execution.

Marc Andreessen redefines SpaceX as a civilization infrastructure company that integrates AI, energy, and space, rather than a pure rocket launch company.

mirrors.wikimedia.org will be shut down on May 15, primarily affecting users who manually configured it for Debian/Ubuntu APT; most users can switch to alternative mirrors.

A massive breach exposed credentials for thousands of sensitive networks, including a NATO defense contractor, with attackers using a 45-GPU cluster to crack VPN authentication hashes and compromise Active Directory environments.

The article discusses the concept of adding a network layer to AI agents, building on existing tools and vector stores to enable better coordination and communication among agents.

Garry Tan argues that using a database for AI agents is akin to building a Foxconn factory, implying it's insufficient for proper agent functionality.

a16z podcast invites former partner Benedict Evans to explore the analogy between AI and the internet in 1997, pointing out that current AI infrastructure investment is massive but ROI is unclear. Historical experience suggests value will shift upward, and models themselves will find it difficult to achieve differentiated profits.

A discussion on the missing infrastructure required to run AI agents in production, including monitoring, permissions, recovery, and audit trails, questioning whether this will become a new infrastructure category.