It’s a new day at Digg. Fresh stories are clustering now. 85% complete.
Top Tech Stories
1
Top Story
François Fleuret doubts eight H100s can run 3-trillion-parameter Kimi 3
Zephyr suggests Huawei SuperPoDs or NVL72 systems are required.
2
Chamath Palihapitiya warns restricting open-source AI harms US competitiveness
Proprietary tokens could cost US firms 50 times more.
3
Russ Salakhutdinov says Moonshot AI founder left US voluntarily
Yang turned down an Apple job to build Moonshot
4
Naval Ravikant calls Demis Hassabis's AI proposal regulatory capture
Pim de Witte argued the framework protects open competition.
5
Local open-weight models threaten cloud giants with Sun Microsystems-style collapse
Commodity hardware running Linux originally triggered Sun's rapid demise.
6
Moonshot AI restructures ahead of potential Hong Kong IPO
The listing could be finalized within six months.
7
Chris Paxton argues RL environments offset Big Tech's data advantage
Shital Shah responds that environment scaling now requires automation.
8
Prime Intellect's Will Brown warns OPSD may not work
Lisan al Gaib argues Composer may already use OPSD.
9
Russ Salakhutdinov argues student autonomy drives PhD success
Edinburgh's Pasquale Minervini agreed, urging advisors to avoid micromanagement.
10
LangChain's Harrison Chase argues AI agents are shifting to graphs
LangGraph was built to orchestrate stateful, multi-actor workflows.
11
TeortaxesTex speculates on unreleased DeepSeek V4 benchmark targets
The targets include scoring 72% on the DeepSWE benchmark.
12
Moonshot AI's 2.8T Kimi K3 tops WebDev coding leaderboard
Moonshot will release the full weights on July 27.
13
GPT Researcher integrates LangChain Deep Agents to double verified citations
Emergent planning and delegation significantly reduce the system's codebase
14
Google DeepMind reconstructs Pelé's unfilmed 1959 goal using AI
The reconstruction visualizes Pelé's famous triple-sombrero play.
15
Moonshot AI launches 2.8-trillion-parameter multimodal model Kimi K3
It supports a 1 million token context window.
16
Moonshot AI's Kimi team reportedly outpaces OpenAI on scaling commitment
The team rejects research not directly contributing to AGI.
17
DeepSeek V4 Pro requires 4.14 GiB KV cache at 1M tokens
Rivals GLM-5.2 and MiniMax M3 require up to 70.8 GiB.
