EDITORIAL TEAM

18BYTE Editorial Team

18BYTE Editorial Team is the publication team at 18BYTE Product Studio. We report on AI releases, review AI products against the outcomes people actually need, and publish the sources behind every claim with the date each was last verified. Where a story touches One Conductor, a product of the same studio, we disclose it.

PUBLISHED WORK

A central AI program coordinated by separate strategic and operational leadership systems

Google split its AI leadership in two. The problem it is solving is a scheduling one.

Influence moving from a fading central AI model hub to a growing distributed open model network

Meta retired Llama. The centre of gravity for open AI moved without it.

Large AI compute systems compressed into smaller efficient units as cost per answer falls

The price of frontier AI fell 80 percent in a week. The race is now about cost per answer.

User documents moving through separate AI data-use paths for model training and private processing

Does your AI train on your work? The defaults, read plainly

Downloadable AI model weights contrasted with a fully open and modifiable source system

Open weights is not open source, and the difference is in the licence

A long AI context window holding extensive information while effective attention weakens across distance

Context windows explained: why a million tokens is not a million tokens

An advanced AI processor package on a circuit board, lit by multicoloured optical signal paths

Anthropic is designing its own AI chips. The model race is becoming a systems race.

AI research and execution pathways branching from a shared intelligence core

Google restructures its AI leadership around science and execution

AI evaluation signal crossing a layered cybersecurity containment boundary

OpenAI says cyber evaluations crossed intended testing boundaries

Regional AI data centers connected across the United States

NVIDIA joins a U.S. program to widen access to AI infrastructure