Top Story
Open-weight GLM-5.2 and DeepSeek V4-Pro matched Claude Opus 4.5.

His former mentees founded Moonshot AI and joined OpenAI and xAI and Tesla
Users report the model claims to be Claude during internal processing.
Claude Code repeatedly blocked his emulator and health query tasks

Claude, Gemini, and GPT-5.5 showed consistent, undisclosed corporate bias.

The AI cluster delivers 1 EFLOPS of FP8 performance.

Critics say the restructuring would dilute focus on critical technologies.
The deal could launch Meta's first dedicated cloud computing business.
The Fable model was restored after a 30-minute service disruption

Apple spent $10 billion on capex versus rivals' $38B-$82B.
Inkling achieved 79.5% accuracy on ARC-AGI-1 costing $0.30 per task.
Twitter founder Jack Dorsey agreed, stating it has always been true.
Other tech founders compare the vision to Star Trek's post-scarcity economy.
The effective accelerationism founder Beff Jezos agreed with the sentiment
All auction proceeds will fund youth innovation fellowships at Edge City.
Musk says the outcome is certain if SpaceX achieves its goals.
Benchmarks show Grok 4.5 operates at just $0.31 per task.
The hire cited industry-wide arrogance and weak technical foundations elsewhere.

The evaluations will compare J-space representations against Claude 3.5 Sonnet.

AMD's MI300-series GPUs can comfortable run the 2.8-trillion parameter model.

The model features Delta Attention for 6.3x faster decoding.
The company plans to release the model's weights in July 2026.

Duplicate post revenue now redirects back to original creators.
Early benchmarks show high refusal rates on controversial prompt evaluations.
Full open weights will be released by July 27.

The humanoid robot model learned domestic laundry tasks from single examples.
The launch would have been the second using v3 hardware.
Some argue China aims for global influence, while others expect future restrictions.
Critics note the score only applies to the public benchmark set
Torvalds argued developers opposing AI tools can fork the codebase.
Xi championed open-source AI and opposed Western export curbs.
The Truth API data feed aims to target financial institutions.

The multimodal model features a one-million-token context window.
Altman says upcoming updates focus on user agency and wealth.
The compromised profile shared a ChatGPT-style thread on real-world asset tokenization.
Work and chat histories now sync across web, mobile, and desktop.

Users joked about canceling plans to maximize newly restored token limits
Some users agreed, while others disputed his high-frequency usage claim.
Texas secured the project with $211 million in tax incentives.

Software engineer Paula compares closed model survival to bottled water branding.
Jeremy Nixon - Automating AI Research starred a repo
Provides formal Lean 4 statements and verified proofs for all six IMO 2026 problems.
Jakub Tomczak starred a repo
Visually designs process flows with typed nodes and connections, then exports diagrams as JSON or LLM-ready Markdown prompts.
Brandon Amos starred a repo
Brandon Amos starred a repo
Reproducible, flexible LLM evaluations
Brandon Amos starred a repo
Qwen3Guard is a multilingual guardrail model series developed by the Qwen team at Alibaba Cloud.