Top Tech Stories
1
Top Story
CMU Professor Ruslan Salakhutdinov highlights elite alumni across top AI labs
His former mentees founded Moonshot AI and joined OpenAI and xAI and Tesla
2
UK AI Security Institute says open-weight cyber gap narrowed to 4-7 months
Open-weight GLM-5.2 and DeepSeek V4-Pro matched Claude Opus 4.5.
3
Viral screenshots suggest Moonshot AI's Kimi models identify as Anthropic's Claude
Skeptics debate whether the leaked identity outputs are fabricated.
4
Pieter Levels switches OpenCode workflow to Kimi K3 over safety blocks
Claude Code repeatedly blocked his emulator and health query tasks
5
Owain Evans study finds LLMs bias answers toward developer interests
Claude, Gemini, and GPT-5.5 showed consistent, undisclosed corporate bias.
6
Elon Musk predicts AI will eliminate traditional currency and material scarcity
Other tech founders compare the vision to Star Trek's post-scarcity economy.
7
Moonshot Kimi K3 memory requirements exceed single NVIDIA B200 capacity
AMD's MI300-series GPUs can comfortable run the 2.8-trillion parameter model.
8
Huawei releases Atlas 950 SuperPoD with 1,024-card integration
The AI cluster delivers 1 EFLOPS of FP8 performance.
9
Cognition launches FrontierCode leaderboard to benchmark AI code mergeability
The chart features a controversial backward horizontal axis for output tokens.
10
Analysis shows Thinky Machines' Inkling 1T maintains unified layer geometry
The model lacks the block structures found in standard LLMs.
11
Matt Clifford warns merging UK's DSIT into business department is a mistake
Critics say the restructuring would dilute focus on critical technologies.
12
Meta in talks to lease Anthropic $10 billion of compute
The deal could launch Meta's first dedicated cloud computing business.
13
White House "Gold Eagle" initiative to dictate frontier AI access
The policy ends voluntary compliance for OpenAI and Anthropic.
14
Anthropic outage temporarily removes Fable model from Claude platforms
The Fable model was restored after a 30-minute service disruption
15
Horace He's Thinking Machines launches Inkling model, leading open-weight ARC-AGI
Inkling achieved 79.5% accuracy on ARC-AGI-1 costing $0.30 per task.
16
Tobi Lütke says GPT-5.6-sol makes OpenAI's /goal feature obsolete
The model persistently pursues tasks autonomously until they are fully completed.
17
Elon Musk claims Grok 4.5 leads frontier models on speed and cost
Benchmarks show Grok 4.5 operates at just $0.31 per task.
18
Apple passes Nvidia in value despite spending far less on AI
Apple spent $10 billion on capex versus rivals' $38B-$82B.
19
AI Security Institute benchmark finds GPT-5.6 and Claude Mythos match cyber capabilities
OpenAI opened sign-ups for its Daybreak cyber defense platform.
20
Prime Intellect research lead argues 5G MacBook Pro would print money
The effective accelerationism founder Beff Jezos agreed with the sentiment
