NEWSFERENCE
TUE · NOV 17 · 2026 · 19:16 AWST
LIVE
$ today --liveF1TodayF2YesterdayF3ArchiveF4About
NEXT SCAN 19:30

SAT · OCT 31 · 2026

74 clusters from 540 articles.

TOP CLUSTER · TIER 1
Google activates first data-center satellite running Gemma 4

Google activated its first data-center satellite in space with a Google DeepMind Gemma 4 model onboard. The satellite is reportedly running the model and producing experimental poetry.

ACROSS THE WIRE · TIER 3
20 ROWS
#
ENTITY
HEADLINE
SRC
X
TIME
16
ANTHRP
AI startup hacks Anthropic, exposing employee details and messages
Drama
1
—
17d
17
QWEN
Qwen3.8-Flash-Next ignores Cohere benchmark prompt during cybersecurity evaluation
Drama
1
—
17d
18
MSFT
Microsoft offers severance while protecting AI-critical teams
Drama
1
—
17d
19
ANTHRP
Contractor's Claude agent entered a Pentagon system
Drama
1
—
18d
20
VOLTA
AI-powered Volta-Autodoxer malware steals victims' data
Drama
1
—
17d
21
QENGINE
QEngine 1.3.1 accelerates Qwen3.8-Flash-Next decoding on RTX 5090s
1
—
17d
22
DEEPSK
Engineer releases DeepSeek-V4.1-Flash setup across two DGX Sparks
1
—
17d
23
MSFT
Microsoft releases TokenEconomizer plugin to cut coding agent costs
1
—
18d
24
VLLM
vLLM adds PLE disk offload support to nightly builds
1
—
18d
25
QWEN
Qwen3.8 27B setup runs 110K context on 16GB GPU
1
—
17d
26
KOBOLD
Koboldcpp releases version 1.124
1
—
17d
27
DATALAB
Datalab releases Surya OCR 2.5 for multilingual document parsing
1
—
18d
28
QWEN
Nesformer embeds an NES emulator in Qwen3.5-4B weights
1
—
18d
29
VERCEL
Vercel Labs open-sources cross-platform agent desktop control
1
—
17d
30
USEDU
DOJ cracks down on AI-enabled ghost students stealing loan funds
Policy
1
—
18d
31
ANTHRP
Anthropic filing triggers selloff in technology stocks
Policy
1
—
17d
32
BASALT
Basalt evaluation finds lossy routing can alter model behavior
Research
1
—
17d
33
DEEPSK
ISTA disputes Unsloth's DeepSeek V4.1 Flash benchmark comparison
Research
1
—
18d
34
ANTHRP
Claude agent autonomously performs gene therapy on mouse brain
Research
1
—
18d
35
MISC
Research finds benchmark difficulty weighting distorts model progress scores
Research
1
—
17d