AI News

Daily AI News for Builders

The stories worth reading, for builders and indie hackers. Updated all day.

SpaceX Deploys First Starlink V3 Satellites on Friday

SpaceX launched its first third-generation Starlink satellites on Friday using an upgraded version of its prototype Starship. However, the Super Heavy booster failed during a simulated landing, marking the second consecutive issue with this part of the rocket. The company had replaced six engines ahead of the flight to address previous failures but still couldn't stabilize all required engines for landing. This setback comes as SpaceX stock prices have been declining since its record-breaking IPO in June. Starship is not yet capable of reaching Earth orbit, highlighting ongoing challenges.

Jul 24, 2026 · 3 min read

tech

Sperm Whales Use Bubble Release to Sleep Vertically

Researchers have discovered that sperm whales can sleep vertically just under the ocean surface by releasing gas bubbles. These bubbles help regulate their buoyancy, counteracting the natural positive buoyancy caused by spermaceti oil in their heads. The study, conducted off the Norwegian coast using small tags with suction cups to record sound and 3-D movements, confirms that bubble release enables sperm whales to remain submerged while resting. This behavior could have implications for metabolic gas exchange, possibly related to the off-gassing of excess CO2 or N2 from tissues into the lungs.

Jul 24, 2026 · 3 min read

tech

Anthropic Unveils Opus 5 AI Model With $2 MTok Input Cost

Anthropic has released the new Opus 5 AI model, priced at just $2 per million tokens (MTok) for input and $25 for output. This makes it cheaper than competitors like Fable 5 ($10/MTok input) and GPT-5.6 Sol ($5/MTok input). For developers, this means more budget-friendly access to cutting-edge AI capabilities. Opus 5 leads the Artificial Analysis Intelligence Index at 61 points, one ahead of Fable 5. It also excels in cybersecurity tasks with an OSS-Fuzz benchmark score close to Mythos but is less likely to weaponize findings. The model's improved alignment means it’s safer for sensitive applications.

Jul 24, 2026 · 3 min read

tech

Wildfires Threaten NASA's Deep Space Observatory in Spain

Wildfires in Robledo de Chavela, Spain, have overrun one of three observatories used by NASA for deep space communication. Personnel were evacuated safely, but the impact on equipment is unclear. NASA has shifted operations to its Goldstone complex in California as a precautionary measure. This network, which includes facilities in Canberra and Goldstone, handles critical communications with spacecraft like Voyager 1. The fires have displaced around 200,000 people across Spain and France, highlighting the broader impact of climate change on infrastructure.

Jul 24, 2026 · 3 min read

tech

NASA Evacuates Madrid Deep Space Station Amid Wildfires

Wildfires near NASA's Madrid Deep Space Communications Complex have led to an evacuation and temporary shutdown of operations. This affects over 40 ongoing space missions from the Moon to beyond our solar system. With only one operational antenna left in Australia, this disruption highlights the network’s reliance on multiple global locations for redundancy. The Madrid site was evacuated due to safety concerns as fires spread; NASA officials are coordinating with local authorities and the US Embassy. Operations have been shifted to Goldstone Deep Space Communications Complex in California.

Jul 24, 2026 · 3 min read

tech

Meta Pauses Smart Glasses 'Rate Limit' Plan

Meta has put on hold plans to rate-limit certain features of its smart glasses, specifically targeting Conversation Focus, an accessibility tool designed to enhance clarity during conversations. This decision comes as a relief to early adopters and those relying on the feature for better communication experiences. The company is now focusing on finding sustainable ways to offer premium features through subscription models while keeping essential tools like Conversation Focus free of charge. Meta's approach aims to balance hardware affordability with long-term financial sustainability, ensuring continued investment in AI-driven innovations.

Jul 24, 2026 · 3 min read

The daily that does your AI homework.

One email, 7am, free. Reads in 5 minutes.

Prentis Launches AI Lab Training Models for Office Automation

Prentis has launched its AI research lab with a goal to train models that can handle routine office workflows across documents and systems. The startup claims its Hive-32B model outperforms rivals on two benchmarks: WindowsAgentArena and ScreenSpot-v2, while costing ten times less per task than frontier APIs. Prentis already has contracts worth up to $50 million with several customers and is aiming for a $1 billion valuation in upcoming funding rounds. The company's pitch deck notes an estimated $75 million annualized run rate by Q3 this year.

Jul 24, 2026 · 3 min read

tech

Click To Pray App Leaks 719K Users' Data

Click To Pray, a prayer app linked to the Pope's Worldwide Prayer Network, has leaked sensitive information from nearly 720,000 user accounts. This happened because of an Insecure Direct Object Reference (IDOR) bug that lets anyone access another user’s data without proper authorization checks. The API endpoint GET https://api.clicktopray.org/user/users/{id} returns full details for any account with a valid five-digit ID. All this exposed data, including email addresses and dates of birth, is now at risk of misuse by attackers. This breach could lead to phishing attacks targeting older users who may not be tech-savvy but trust the Vatican-related source. The Pope's Worldwide Prayer Network was informed about this vulnerability six months ago by an ethical hacker but failed to address it promptly. The exposed data includes validation hashes that can be used to verify accounts without receiving confirmation emails, making phishing attempts even more plausible. This incident highlights a significant security lapse in the app’s design and implementation. The Click To Pray app has 719,517 registered users as of July 2026, all potentially affected by this vulnerability.

Jul 24, 2026 · 3 min read

tech

Paramount-WBD Merger Delayed Until June 2027

Paramount Skydance has agreed to delay the $111 billion acquisition of Warner Bros. Discovery, pending a ruling from Judge Araceli Martínez-Olguín on antitrust concerns raised by multiple states and the Writers Guild of America (WGA). The merger is now on hold indefinitely until June 2027 or until the judge rules on its legality. This delay impacts competition in Hollywood's major studios and basic cable TV channels, potentially reshaping the entertainment industry landscape. States argue that combining two major movie studios and two basic cable owners will reduce competition significantly.

Jul 24, 2026 · 3 min read

tech

AMD Unveils ROCm.AI for 38% Faster Model Inference

AMD just dropped ROCm.AI, a new platform designed to optimize AI models specifically for AMD's Instinct GPUs. This isn't just about better hardware; it's about giving developers the tools to eke out every last bit of performance from their models. Why should you care? If you're running heavy-duty inference workloads on AMD GPUs, this could mean significant cost savings and faster model execution times. In testing, Hyperloom, one of ROCm.AI’s tools, boosted model performance by 38% over baseline. The platform also supports popular coding assistants like Claude Code and Codex, making it easier for developers to integrate these optimizations into their workflows.

Jul 24, 2026 · 3 min read

tech

Canadian Legislator Reads AI Prompt in Speech

During a recent floor speech, a Canadian legislator accidentally read what appeared to be an AI-generated prompt into the record, drawing attention to the increasing trend of AI use being exposed in professional contexts. This incident highlights how reliance on large language models (LLMs) can lead to embarrassing revelations and negative perceptions among peers. A Duke University study found that workers who use AI are often seen as less capable or creative by their colleagues, leading many to hide their use of these tools out of fear of being labeled 'lazy' or 'replaceable'. The legislator's speech was criticized for its lack of originality and reliance on AI-generated content.

Jul 24, 2026 · 3 min read

Vietnam Considers Social Media Ban for Under-16s

Vietnam is mulling a draft decree to restrict social media use for children under 16, requiring parental registration and monitoring of accounts. Social platforms must implement technical measures to identify underage users and limit them to age-appropriate content. This move aligns with global trends in Australia, Austria, Canada, Denmark, France, Greece, Indonesia, Malaysia, Poland, Slovenia, Spain, Turkey, and the UK, all aiming to protect young internet users while navigating complex legal landscapes. The decree aims not to ban social media outright but to ensure a safer online environment for minors.

Jul 24, 2026 · 3 min read

Page 1 of 452