GLM-5 is open to inspect—not proven yet
GLM-5 has arrived on Hugging Face with bilingual and mixture-of-experts tags, but independent evidence of its real performance is still missing.
THE LOG — TRANSMISSIONS FROM ALTITUDE
What we learn running AI systems every day — workflows, growth loops, tools, and the occasional hard lesson. Written down so you can reuse it.
GLM-5 has arrived on Hugging Face with bilingual and mixture-of-experts tags, but independent evidence of its real performance is still missing.
Grok Bot pitches an early-beta shift from chat answers to delegated computer work, while reliability, security and results remain unproven.
ElevenLabs says Deutsche Telekom is moving AI voice assistance, translation, transcription and summaries into the telecom network.
MiniMax is pitching a connected creative workspace that divides a brief among specialist media agents and returns control to the creator at review checkpoints.
OpenAI says GPT-5.6-Cyber reached 95% on one internal task test and helped identify a V8 exploit chain, but wider results are mixed and access is restricted.
OpenRouter says Seedance 2.5 can combine image, video and audio references, generate up to 30 seconds, and extend or revise the result—with operational and privacy limits.
Anthropic’s Managed Agents now expose controls for session spend, inference geography, repository skills and stronger-model advice—with important limits still unstated.
Muse Glimmer brings a 30B agent model onto high-memory consumer hardware, but local does not mean lightweight, universally compatible or independently proven.
Grok’s new image model promises precise local edits, reusable visual worlds and more practical creative workflows—but its supplier-selected examples still need real-world testing.
Anthropic says its rewritten biology safeguard lets more everyday health and learning questions reach Fable 5 while dual-use research still falls back to Opus 5.
Claude Code sessions can exchange bounded text updates while remaining separate in history, files, permissions and control.
OpenAI says Astra may meet its highest cyber-risk threshold, prompting stronger controls and paused internal work while evaluation continues.
Muse Code coordinates parallel workers, background reviewers and traceable actions—but remains a supplier-claimed beta with a notable pricing and data-use trade-off.
MiniMax H3 promises mixed-media understanding and stereo-audio video up to 2K, but its public release contains the central generation engine rather than the complete three-module system.
Qwen is pitching a multi-day AI coworker, but duration, control and promised open weights still need verification.
DeepSeek says V4-Flash improves agent performance and lowers repeated-context costs, but the beta remains API-only and the benchmark evidence is supplier-reported.
OpenAI says GPT-5.6 combines lower prices with a more direct path through API work, shifting the buying question from token price to cost per useful result.
OpenAI says Luna and Terra are cheaper, Sol has a premium fast lane, and moving Auto-review to Luna could cut its cost by about tenfold.
Google Gemma presents Gemma 4 26B as the engine for a local OpenClaw assistant, but its headline 90% workload claim remains unverified.
Google DeepMind’s three-part robotics system aims to connect whole-body control, planning and local adaptation—but its announcement does not establish deployment readiness.