Sakhanda Wire
NVDA $150.96 -2.89% MSFT $508.29 +0.40% GOOGL $282.44 +0.80% META $616.38 -0.40% AMZN $267.74 +3.82%

Latest news

Older stories

TechCrunch

Builders Stage to return at TechCrunch Disrupt 2026

The Builders Stage will return at TechCrunch Disrupt 2026, offering discussions for founders, startup operators and investors on building and scaling companies. The event is scheduled for October 13–15 at Moscone Center in San Francisco and is expected to bring together more than 10,000 founders, investors, technology leaders and operators.

TechCrunch

Palo Alto Networks acquires AI help desk startup Console for $500 million

Palo Alto Networks paid $500 million in cash and stock for Console, a two-year-old startup developing AI agents for routine IT help desk work, according to two people familiar with the transaction. The companies announced the acquisition on Tuesday but did not disclose its financial terms publicly.

TechCrunch

TechCrunch Disrupt 2026 announces Real World AI Stage lineup

TechCrunch Disrupt 2026 will introduce a Real World AI Stage alongside its existing AI Stage. The new program will examine how artificial intelligence is moving from digital environments into physical settings, including public spaces, homes, battlefields and autonomous machines. The event is scheduled for October 13–15 at Moscone West in San Francisco.

TechCrunch

OpenAI’s Astra raises concerns over opaque reasoning

OpenAI’s upcoming Astra model is reportedly set to use “recurrent depth,” a reasoning method also known as “opaque recurrence.” The technique processes a query repeatedly in a loop rather than following a strictly sequential path. According to The Information, Astra’s use of it is limited, but the approach could leave fewer readable traces of the model’s reasoning.

The Verge

Google launches Gemini 3.8 Flash with potentially higher token use

Google has launched Gemini 3.8 Flash, only a few weeks after the previous Gemini 3.7 Flash model. The company says the newer system “works harder” on complex tasks by taking more reasoning steps and calling tools iteratively. These changes are presented as ways to maximize performance, particularly when users select higher effort levels.

The Decoder

Justice Department backs fair use for AI training

The US Department of Justice has sided with AI companies in the consolidated copyright lawsuit involving The New York Times and other rights holders. In a court filing, the department argues that using copyrighted material to train large language models can qualify as fair use rather than infringement.

Ars Technica

Google announces Gemini 3.8 Flash with reasoning and cybersecurity variants

Google has announced Gemini 3.8 Flash, its third Gemini Flash model release in six weeks. The company says the new model is its strongest Flash system so far for reasoning and coding. The announcement comes after Google went without a frontier-level Gemini Pro release since early 2026, and the continued focus on Flash models has raised questions about whether the previously promised Gemini 3.5 Pro will appear.

MarkTechPost

NVIDIA releases Switchyard proxy for routing and translating LLM traffic

NVIDIA has released Switchyard, an Apache-2.0 Rust proxy and library designed to connect coding agents with different model providers. Claude Code can use the Anthropic Messages API while Codex CLI uses OpenAI APIs, even when the selected backend runs through vLLM, NVIDIA NIM, or Ollama. Switchyard is intended to provide the translation layer without changing the client.

Ars Technica

Protect Democracy seeks secret rules for Trump-era AI safety reviews

Protect Democracy has filed a complaint seeking records about the GOLD EAGLE program, an administration initiative that reportedly supports reviews of advanced AI models. The organization says the White House has announced the program but has not explained how it operates, who participates, or what legal authority it relies on.

MarkTechPost

Google releases Gemini 3.8 Flash variants with different safety access

Google released Gemini 3.8 Flash and Gemini 3.8 Flash Cyber on September 2, 2026. Both use the same underlying model, while their differences concern safety mitigations and access conditions. Gemini 3.8 Flash is generally available through the Gemini API, Google AI Studio, Antigravity, Android Studio, and Gemini Enterprise. Its weights remain closed, ruling out self-hosted or on-premises deployment.

TechCrunch

AI detection expands beyond simple “real or fake” judgments

The spread of AI-generated text and images is creating new trust challenges online. Such material is appearing in job applications, product reviews and insurance claims, making it harder for platforms and users to determine whether content is authentic. Pangram is among the startups developing tools intended to serve as a trust layer for the internet.