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

Hugging Face

Sentence Transformers v6.0 adds ColBERT-style multi-vector retrieval

Sentence Transformers v6.0 introduces MultiVectorEncoder, a fourth model type for ColBERT-style late-interaction retrieval. The Python library already supports dense, sparse and reranker models for uses including semantic search and retrieval-augmented generation. MultiVectorEncoder can load PyLate and Stanford-NLP ColBERT checkpoints, while ColPali-family models support visual document retrieval through the same API.

TechCrunch

Anthropic’s annualized revenue exceeds $65 billion

Anthropic’s annualized revenue run rate exceeded $65 billion at the end of July, according to Bloomberg. The figure, which projects a full year of revenue from a shorter recent period, rose from $47 billion in May and $9 billion at the end of last year. The company did not immediately comment on the report.

TechCrunch

Relay shuts down as founder joins Google’s Chrome team

Relay, an AI-powered workflow automation startup launched in 2021, is closing its service. The company initially sought to become a new Zapier by helping businesses automate repetitive work, including document drafting, copyediting and project-management tasks. Free users lost access on August 15, while access for paying customers is scheduled to end on September 14. The closure was first announced in July.

Ars Technica

Former SpaceX engineers are building a robotic steel-parts factory

Three former SpaceX engineers are turning from rocket-engine production to steel manufacturing, using AI-based software and robots to automate much of the work. Their startup, 1872, officially began operations on July 22 with a ribbon-cutting ceremony at its Factory One site in Cincinnati, Ohio.

Hugging Face

GPU Allocator Raises Utilization by Up to 33 Percentage Points

A team developed a constraint-aware GPU allocator and compared it with a FIFO scheduler across seven benchmark scenarios. The tests used identical hardware and workloads. Utilization increased by as much as 33 percentage points, while priority-weighted output improved in every scenario, reaching a maximum gain of 105.1%. In the training-heavy test, utilization rose from 53.6% to 87.0% and weighted value more than doubled.

MIT Tech Review

Flock changes leave broader surveillance concerns unresolved

Flock, which operates about 120,000 automatic license plate readers across the United States, has introduced measures intended to deter improper police searches. The changes include software that flags unusual activity and a requirement that officers enter a criminal case number. The Washington Post identified 50 cases involving misuse of Flock and competing systems, including alleged stalking by officers. Flock told MIT Technology Review that it does not verify the case numbers, leaving open the possibility of fabricated entries.

MarkTechPost

MiniMax releases open-weights model for five-minute song generation

MiniMax has released MiniMax-Music3, an open-weights text-to-music model that creates complete songs of up to five minutes in one generation. It accepts lyrics containing section tags such as [Verse] and [Chorus], together with a structured caption describing global metadata, vocals and arrangement. The output is a 32 kHz, 16-bit stereo WAV file.

Ars Technica

Hidden tracker links rare book order to Amazon AI facility

Booksellers have suspected for roughly a year that artificial intelligence companies were purchasing large lots of rare books, scanning them and then destroying them. Evidence reported by 404 Media now links at least one such bulk order to Amazon, which is developing advanced AI models.

MarkTechPost

docTR Tutorial Builds a Full Document Intelligence Pipeline

A new tutorial presents an end-to-end document intelligence workflow built with docTR. The pipeline combines text detection and recognition with geometric information, layout analysis, structured extraction and document export. It creates realistic synthetic invoices, loads images and PDFs through DocumentFile, and configures OCR predictors for GPU-aware processing.

TechCrunch

Amazon reportedly cuts rare books to scan them for AI training

Amazon is buying rare books, removing their spines and scanning the pages for artificial-intelligence training, according to 404 Media. The report followed a tracking device placed inside a rare book, which later reached an Amazon facility in Las Vegas. The site, identified as VGT3, uses a logo showing a dinosaur holding a book in its claws.

TechCrunch

Groq raises $350 million as it shifts toward Nvidia-powered neocloud services

Groq has raised $350 million in funding as it moves away from making AI chips and develops a neocloud business focused on GPU access and AI infrastructure. The round was led by Disruptive, with planned participation from Nvidia, and values Groq at $3.5 billion. That is below the company’s $6.9 billion valuation last September.

TechCrunch

Nvidia to invest $1.5 billion in developer of OpenAI data center

Nvidia said Monday that it will invest $1.5 billion in SB Energy, a data center developer connected to SoftBank and OpenAI. The deal would make Nvidia the exclusive provider of compute infrastructure for OpenAI’s Ports-Pike data center near Cincinnati, Ohio. Nvidia also plans to offer up to $105 billion in credit for construction, according to documents filed with the SEC. The facility could grow from an initial 4.25 gigawatts to 8 gigawatts.