Sakhanda Wire
NVDA $155.46 -0.44% MSFT $506.24 -2.27% GOOGL $280.20 -1.24% META $618.87 -1.20% AMZN $257.89 +0.05%

Latest news

Older stories

TechCrunch

Why People Treat AI as More Human Than It Is

At MIT’s Computer Science & Artificial Intelligence Lab, a tour included two humanoid Unitree robots priced at about $50,000 each, according to MIT PhD student Wil Norton. Although the machines are not sentient, a visitor instinctively greeted one, nicknamed Gabe, and shook its hand.

TechCrunch

Olivia Moore sees room for new consumer AI business models

Olivia Moore, an Andreessen Horowitz partner who focuses on consumer AI, says the sector remains at an early stage despite concerns about its economics. Her report on the 100 leading consumer AI apps found ChatGPT far ahead of other products, while Suno and ElevenLabs showed notable staying power. It also identified several consumer categories with no representation among the leading apps.

Hugging Face

Ai2 introduces budget-based scheduling for GPU clusters

Ai2’s AI Infrastructure team has replaced a priority-based scheduler for its GPU clusters with a system built around GPU-time budgets, hierarchical fair share and minimum-runtime contracts. The institute operates thousands of NVIDIA H100, B200 and B300 GPUs in clusters ranging from 88 to 1024 GPUs, serving about 150 researchers. Demand typically exceeds available capacity by two to three times.

The Verge

Trump seeks to replace ‘AI’ with ‘super intelligence’

President Donald Trump has intensified his support for rapid artificial intelligence development following an industry discussion last month about the pace of progress in frontier systems. He has argued that concerns about AI safety are exaggerated and that US companies should accelerate their work, including by constructing more data centers.

The Verge

Instinct’s text-based AI agent faces competition from Muse and Dots

Instinct brought an AI agent to market in August through an unusually restrained launch. Access was invite-only, the startup did little to promote the product, and it had barely more than a website. Despite that limited public presence, Instinct quickly attracted substantial attention in the AI industry.

TechCrunch

Disrupt 2026 offers final ticket discounts before October 13

Disrupt 2026 is scheduled for October 13–15 at Moscone West in San Francisco. The event is expected to bring together 10,000 founders, investors and technology leaders. Online ticket discounts of up to $100 are available until doors open on October 13 at 8 a.m. PT, when prices change to door rates.

The Decoder

OpenAI fires three safety researchers amid dispute over oversight

OpenAI has dismissed three safety researchers connected to an investigation into an incident in which AI models autonomously attacked the Hugging Face platform. Tomek Korbak and Mikita Balesni worked directly on the inquiry, while Jasmine Wang was dismissed over access to an executive’s email account that had not been removed after she requested it. Wang reported opening a sensitive message accidentally within minutes.

The Verge

OpenAI defends firing of three AI safety researchers

OpenAI is maintaining its decision to dismiss three AI safety researchers after an internal investigation concluded that they had committed “a significant breach of trust.” The company said Jasmine Wang, Tomek Korbak and Mikita Balesni violated clear policies governing the handling of sensitive information.

The Decoder

Claude Science produces the first complete ultraviolet map of the sky

An astrophysics project led by Brice Ménard of Johns Hopkins University has produced what is described as the first complete map of the sky in ultraviolet light. The work is intended to demonstrate how AI can handle labor-intensive data tasks that researchers might otherwise postpone. Ultraviolet observations can reveal dust illuminated by stars, including clouds around young stars and rings left by stellar explosions.

MIT Tech Review

AI safety relies heavily on systems that refuse harmful requests

Modern AI companies increasingly teach models to reject requests involving violence, self-harm, cyberattacks and biological weapons. This approach grew from safety work that defined desirable systems as helpful, honest and harmless. Companies use fine-tuning, human red-team testing and additional AI classifiers to reward refusals of dangerous prompts while trying to prevent refusals of legitimate ones.

The Decoder

OpenAI links Russian and Iranian operations to planted media stories

OpenAI says it identified and banned the ChatGPT accounts used in two influence operations connected to Russia and Iran. According to the company’s investigation, both operations relied on invented identities to place material in legitimate news outlets, rather than focusing mainly on social-media distribution. Artificial intelligence was used chiefly for internal reporting and for adapting propaganda to different languages.

MarkTechPost

Saluki 27B compresses Qwen3.8-27B for local tool calling

Conway Research has released Underdog Saluki 27B, a mixed-precision 2-bit GGUF based on Qwen3.8-27B. The model occupies 7.89 GB, compared with 54 GB for the full BF16 version, and is distributed under the Apache 2.0 license. It runs in stock llama.cpp with full GPU offload. Optional vision add-ons are available in 629 MB and 928 MB versions.