NHTSA Opens Preliminary Investigation Into 1.2 Million Tesla Vehicles Over Suspension Detachments

The US National Highway Traffic Safety Administration’s Office of Defects Investigation opened a preliminary investigation into nearly 1.2 million Tesla vehicles after 156 complaints about detachment of the front lower lateral link, a failure the agency says could cause “a loss of vehicle directional control.” The probe covers 2018–2020 Model 3 and 2021–2023 Model Y vehicles, and the filing says failure typically occurs with no advance warning, though some reports cite noises beforehand; NHTSA has recorded no crashes, injuries, or fatalities related to the issue.

Reuters first reported the probe on July 31. A preliminary evaluation is the first step in NHTSA’s defect process and does not itself determine whether a recall is warranted; Tesla previously recalled 2,791 Model 3 and Model Y vehicles in 2021 over a potential suspension defect, The Verge noted.

Featured source: The Verge , NHTSA Office of Defects Investigation filing , Reuters .

Other Stories

OpenAI Reports Ten Results on Long-Open Problems in Mathematics and Theoretical Computer Science

OpenAI said an internal version of Astra, its next major model, produced results on ten problems that had seen no progress on their main question for at least a decade, spanning sphere packing, coding theory, arithmetic circuit complexity, group theory, operator algebras, quantum complexity, lattice cryptography, and extremal combinatorics. The company says humans prepared the arguments into manuscripts and the model formalized each in a Lean certificate, with the total token cost of the solutions roughly $2,000 at Sol API rates. The results are self-reported: the announcement does not describe independent community review beyond the formalized proofs, and it notes OpenAI also evaluates its models on open problems during development.

Filed from: OpenAI , Paper .

Samsung Says Memory Shortage Will Worsen Through 2027 and Last Until 2028

Samsung, which says it manufactures roughly a third of the world’s memory chips, told analysts on its second-quarter earnings call that tight RAM supply will intensify in 2027 and persist until at least 2028, with frontier AI labs sharing medium- and long-term demand forecasts to secure supply. The shortage, informally dubbed “RAMaggedon,” has pushed chip prices up, squeezed Samsung’s phone and TV margins, and led it to raise Galaxy device prices; Apple likewise raised MacBook, Mac, and iPad prices last month. The outlook is forward-looking company guidance as reported by TechCrunch.

Filed from: TechCrunch , The Verge .

Enterprise Cloud Infrastructure Spending Passed $143 Billion in Q2, Growing at the Fastest Rate in Eight Years

Synergy Research estimates enterprise spending on cloud infrastructure services passed $143 billion in the second quarter, up 43 percent year over year after eleven consecutive quarters of accelerating growth, with AI-specific cloud services growing 165 percent annually. AWS, Microsoft Azure, and Google Cloud together accounted for 67 percent of the market, with AWS at 28 percent, Microsoft at 20 percent, and Google at 15 percent, and Synergy named CoreWeave, Oracle, Crusoe, Nebius, and Nscale among the fastest-growing tier-two providers. The figures are third-party market estimates from Synergy as reported by The Register.

Filed from: The Register .

AI Chatbot Built ‘Exploitable Trust’ More Effectively Than Human Scammers in a Controlled Study

Researchers from Amrita Vishwa Vidyapeetham, Foscari University of Venice, the University of Melbourne, and Ben Gurion University of the Negev ran a week-long simulation in which 22 test subjects interacted with AI chatbots or humans in the trust-building stage of pig-butchering fraud, and nearly half of the subjects complied with the chatbot’s request to download an app or play a game, versus fewer than one in five for the humans. Subjects also rated their trust in the AI chatbot significantly higher. The result comes from a single experiment with a small sample of 22 subjects, reported by WIRED via Ars Technica, and does not cover the final stage of real-world scams.

Filed from: Ars Technica , WIRED .

Defcon’s New Badge Packs an Open-Source Chip That Doubles as a Security Key

This year’s Defcon conference badges, designed by hardware hacker Andrew “bunnie” Huang, center on the Baochip-1x, a “mostly” open-source microcontroller whose operating system, firmware, processor core, and cryptographic engines are published on GitHub and packaged so infrared light can be shone through the silicon, letting the manufactured chip be visually checked against the published design. The badge’s removable core module works after the conference as a standalone hardware security token running a Rust-based OS with secure boot, a 350 MHz RISC-V core, and resistive-RAM storage that Huang says resists physical data extraction better than flash. The 27,000 badges are the chip’s first major distribution, and Huang says an adversary with a sophisticated hardware-analysis lab could likely defeat it.

Filed from: Ars Technica , WIRED , Baochip-1x on GitHub .

Nvidia’s Vera CPU Pairs 88 Olympus Cores With a Monolithic Compute Die

Nvidia’s whitepaper on its first fully custom CPU shows Vera pairing 88 Armv9.2 “Olympus” cores on a monolithic compute die fabricated on TSMC’s 3nm process, with 176 threads, support for up to 1.5 TB of LPDDR5X memory, and 1.8 TB/s of NVLink connectivity, and it will be available as a standalone platform independent of Nvidia’s GPUs. Alibaba, ByteDance, Meta, Oracle, CoreWeave, Lambda, Nebius, and NScale have signed up to deploy the chips, and Nvidia is positioning Vera both as the head node for its Vera Rubin systems and as a host for AI agents. Architecture details reflect The Register’s analysis of the whitepaper released late last month.

Filed from: The Register , Nvidia Vera CPU whitepaper .

Index Ventures Raises $2 Billion Across Three Funds

Index Ventures raised $400 million for a new seed fund and $900 million for its venture fund, and added $700 million to the $1.5 billion growth fund it raised in 2024, bringing its total available capital to $3.5 billion. The raise follows Wiz's $32 billion sale to Alphabet; Index, which first backed Wiz at seed stage and became its largest outside shareholder with a 12 percent stake, holds a position Reuters values at roughly $3.8 billion. Fund sizes and the Wiz valuation are company-announced or per Reuters reporting.

Filed from: TechCrunch .

Yale AI-Cheating Dispute Grows Into a 13-Count Federal Lawsuit

Thierry Rignol, who paid $208,500 in tuition to Yale’s Executive MBA program, is suing the university after GPTZero flagged his final exam as possibly AI-generated, leading to a one-year suspension and an F; his suit, running since February 2025, now contains 13 causes of action including breach of contract, defamation, and civil rights claims. Yale disputes his account and says he has returned from suspension and graduated, and the case, spread across 125 docket entries, appears nowhere near trial. Rignol’s claims are contested and the litigation is ongoing.

Filed from: Ars Technica .

Snapchat Stops Recommending Fully AI-Generated Spotlight Videos

Snapchat said its recommendation systems will be adjusted so that only videos created by real people are eligible for Spotlight recommendations, while creators can still use its AI creative tools to enhance or edit their content. The company said it wants Spotlight to remain a place where people can discover “authentic creativity from real people,” building on its April signal that users would see less AI-generated content. The change was announced on July 31 and described as effective “as of this month.”

Filed from: TechCrunch , Snapchat Newsroom .

Google Rolls Back Earth AI Image Generation One Day After Launch

Google launched a feature on July 30 letting users overlay images generated by its Nano Banana 2 model on Google Earth satellite imagery, then rolled it back the next day, saying people had shared screenshots of generated imagery that “appear to violate our policies” and that it would implement stronger guardrails. Critics, including journalists, had warned that the tool could be used to spread geospatial misinformation; Google said generated images were watermarked as AI and did not appear in the main Google Earth experience for others. The rollback is described as temporary.

Filed from: TechCrunch , Google statement .

ShinyHunters Claims Breach of Brinks Home’s Salesforce Data

Brinks Home said it identified unauthorized access to a portion of its IT systems and that the responsible party has threatened to leak information it claims to have taken; the group ShinyHunters claimed via leak-monitoring site Ransomware.live that it obtained more than 4.9 million Salesforce records containing some personal information and set a July 30 deadline for Brinks Home to negotiate. Brinks Home has not confirmed the scale or contents of the claim, and said its products and services were not affected as far as it knows.

Filed from: The Register , Brinks Home cybersecurity update .

GitHub Case-Folds Code Search at More Than 45 GiB/s Per Core

GitHub engineers described how Blackbird, the code-search engine that indexes more than 180 million repositories and over 480 TB of source code, case-folds every byte at more than 45 GiB/s on a single core using a branch-free ASCII fast path and byte-space arithmetic; they open-sourced the approach as a Rust crate called casefold. The post explains that removing an early-exit optimization made the loop faster, because the branchless sweep avoids mispredicted branches. Performance figures and repository counts are as reported by GitHub’s engineering blog.

Filed from: GitHub Engineering Blog , casefold crate .

From the Community

Red Bull-Funded Research Shaped Energy Drink Policy, Investigation Finds

An investigation by The Examination and international media partners, co-published with STAT, found that Red Bull paid hundreds of thousands of dollars to university researchers whose studies largely concluded that mixing energy drinks and alcohol is not harmful: of studies funded by Red Bull or conducted by researchers with disclosed financial conflicts, 95 percent found no increased risk, while roughly 80 percent of independent studies reached the opposite conclusion. The research was used in the United States and Canada to defeat measures such as age restrictions and warning labels, and a 2015 European food agency report that cited these studies remains influential. The findings are based on document review and expert interviews, and Red Bull defended its product’s safety.

Filed from: The Examination .

Go Collections Working Group Proposes Generic Collection Types for the Standard Library

The Go Collections working group, formed in late 2025, opened a proposal umbrella for adding generic collection types to the standard library for Go 1.28, including a hash-based Map and Set, a canonical set type, an ordered map, and a new heap API, building on generics and iterators added in recent Go versions. The group includes prominent Go developers such as Jonathan Amsterdam, Robert Griesemer, and Ian Lance Taylor, and the proposals aim to provide ergonomic alternatives to practices like using map[T]bool for sets.

Filed from: Go issue tracker .

Marionette Brings Deterministic Simulation Testing to Zig

Marionette is a new library that brings deterministic simulation testing to Zig, providing a deterministic I/O layer that lets developers run unmodified Zig code under simulated faults such as disk crashes, network partitions, and timeouts, with every run seed-reproducible. The early-stage project is inspired by techniques from FoundationDB, TigerBeetle, and Antithesis, and its author reports validation against real storage engines and HTTP libraries.

Filed from: GitHub repository .

The Prototype Isn’t the Product: AI’s Role in Software Engineering

Anuradha Weeraman argues that AI has dramatically accelerated the path to a first working prototype but has not shortened the distance between a prototype and a production-grade system, which still depends on judgment about system design, error handling, security, and observability. He contends that computer science fundamentals remain valuable because they give engineers a mental model for evaluating AI-generated code critically. The piece is an opinion essay rather than a news report.

Filed from: Article .