The archive
News · edition

Edition, Wed, 08 Jul 2026.

5 stories on the wire this day. Open any story for its sources. Summaries are written by software from the linked reporting and can be wrong. Spot an error? Email [email protected] and we will correct or remove it.

AI & Technology

The US Commerce Department lifted restrictions on Anthropic’s most advanced models, allowing Fable 5 and Mythos 5 to resume service worldwide on July 1 with updated safeguards.

2 sources
Lead
this edition

AI & Technology

AI & Technology
Anthropic introduced Claude Sonnet 5 for agentic automation and coding tasks, priced at $2 per million input tokens with a new API structure that drops the temperature parameter.
1 source
AI & Technology
Elon Musk’s xAI launched Grok 4.5 in private beta, built on the V9 foundation model with 1.5 trillion parameters and trained with supplemental Cursor data.
1 source
AI & Technology
Google has pushed back the general availability of Gemini 3.5 Pro with no confirmed date, citing excessive token consumption on agentic tasks and gaps in long-context reasoning.
1 source
AI & Technology
Effective July 15, China’s Interim Measures for AI Anthropomorphic Interactive Services require anti-addiction systems and session limits, prompting ByteDance to suspend Doubao’s agent features and Alibaba to pause Qwen’s.
1 source

Anthropic’s Fable 5 returns globally after US lifts export controls

The US Commerce Department lifted restrictions on Anthropic’s most advanced models, allowing Fable 5 and Mythos 5 to resume service worldwide on July 1 with updated safeguards.

Following a June 12 suspension citing national security concerns, the models are back online after Anthropic agreed to new industry-specific “jailbreak” safety frameworks and cybersecurity protocols. The temporary ban highlighted tensions between frontier AI capabilities and government oversight of autonomous weapons and domestic surveillance applications.

Anthropic
CNBC

Claude Sonnet 5 launches with agentic coding focus and removed temperature controls

Anthropic introduced Claude Sonnet 5 for agentic automation and coding tasks, priced at $2 per million input tokens with a new API structure that drops the temperature parameter.

The model delivers Opus-level reasoning at a lower cost but requires code updates as the API no longer supports temperature or top_p settings. Pricing is introductory through August 31, after which it rises to $3/$15 per million tokens.

AIapps

Grok 4.5 enters private beta at SpaceX and Tesla with V9 architecture

Elon Musk’s xAI launched Grok 4.5 in private beta, built on the V9 foundation model with 1.5 trillion parameters and trained with supplemental Cursor data.

The model is described internally as matching or exceeding Opus performance, with a planned monthly cadence for V9 variants leading to Grok 5, which targets 6–10 trillion parameters by year end.

Build Fast with AI

Google delays Gemini 3.5 Pro again as token efficiency and long-context reasoning lag

Google has pushed back the general availability of Gemini 3.5 Pro with no confirmed date, citing excessive token consumption on agentic tasks and gaps in long-context reasoning.

The delay comes as competitors like Claude Sonnet 5 and GPT-5.6 have advanced, and enterprise buyers increasingly weigh intelligence per dollar alongside capability.

Build Fast with AI

China’s new AI companion law forces Doubao and Qwen to pause agent features

Effective July 15, China’s Interim Measures for AI Anthropomorphic Interactive Services require anti-addiction systems and session limits, prompting ByteDance to suspend Doubao’s agent features and Alibaba to pause Qwen’s.

Users can view configurations in read-only mode until October 15, and ByteDance may relaunch as a separate compliance-first product; Alibaba has not announced a similar path.

Build Fast with AI
Every edition

The wire publishes daily and keeps an archive of every edition, listed by date, newest first. If a summary is wrong, we correct or remove it.