AI Daily Brief — 12 April 2025

A quiet Saturday wedged between Saturday’s Llama 4 launch and Monday’s GPT-4.1 family. ChatGPT memory rollout kept advancing, Thinking Machines Lab’s $2B round narrative dominated weekend commentary, and the Llama 4 benchmark scandal kept burning at low heat.

Top stories

• ChatGPT memory rollout continues into weekend. OpenAI’s enhanced memory feature (announced Apr 10) actively rolling out to Pro and Plus subscribers globally, excluding the EU, UK, Switzerland, EEA, Norway, Iceland and Liechtenstein. ChatGPT now references all past conversations beyond saved memories — ‘AI systems that get to know you over your life.’ via OpenAI
• Mira Murati’s Thinking Machines Lab $2B reporting dominates weekend. Reports of a $2B seed round at $10B+ valuation led by Andreessen Horowitz continued to drive AI commentary into the Saturday cycle. The round (with Albanian government participation) ultimately closed later at $12B post-money. via SiliconANGLE
• Llama 4 community scrutiny weekend. Independent benchmarks and developer testing through the week culminated in widespread Saturday discussion about whether Llama 4 Maverick’s LMArena performance matched real-world coding ability, after LMSYS confirmed Meta used a ‘Llama-4-Maverick-03-26-Experimental’ variant tuned for human preference rather than the open-weights model.

Quiet corners

No frontier-lab releases (OpenAI, Anthropic, Google DeepMind, xAI). No China-lab releases (DeepSeek, Qwen, Mistral, Tencent). No NVIDIA / AMD chip announcements. No notable arxiv landmark. No EU AI Act, tariff or regulatory action dated Apr 12.

By the numbers

• $2B at $10B+ — Thinking Machines weekend reporting
• 2 days — until OpenAI’s GPT-4.1 family launches Monday
• Most-mentioned company: OpenAI

Compiled by AI Feed’s editor from verified web sources for 12 April 2025.