Posts

A hardware-specific fork runs Qwen3.8-Flash-Next across two POWER9 processors and four V100 GPUs, showing what model-aware optimisation can recover from a 2018 system.

A plan to run September engineering on GLM 5.3 Flash consumed two billion tokens, but only half reached the target model. Prototypes, provider capacity and evaluation explain the gap.

The cryptographer argues that containment remains necessary but cannot solve the hardest part of agent security: deciding which information and instructions are authorised.

Speaker diarisation, model-cognition papers, local projects, prompting practices, community tests and the day's safety and policy developments.

The German-English reasoning model ships with Apache-2.0 weights and serving instructions. Its small active parameter count still leaves a large memory requirement.

A controlled preprint separates detecting an activation change from reporting its location. The experiment uses fixed input text and scores answers without an AI judge.

A small trained intervention sharply improves Qwen3-8B on synthetic reference chains. The preprint traces the change to how middle layers pass information between lines.

The Ollama-based project adds voice input and everyday document tools. Its published benchmark argues that small-model results depend heavily on the surrounding software.

The developer's workflow keeps specifications and decisions in editable Markdown. His argument comes with an open-source implementation, but no comparative benchmark.

His proposal favours automatic shutdown at a budget limit. Removing the limit would require an explicit choice.

Research, projects and discussions beyond today's six articles, with source links and evidence labels.

Spotlight Memory addresses a small region of expandable storage at each step. Percepta reports strong retrieval beyond the models’ training length.