Search the archive

US INTERCEDES FOR ELON MUSK'S X OVER EUROPEAN FINE

The European Commission fined X $140 million in December for failing to prevent deceptive behavior on its site.

Start earning on your Al spend →

META LAUNCHES LIGHTWEIGHT $1,299 VR HEADSET THAT LOOKS LIKE GLASSES

Meta has launched the Meta VR Glasses at $1,299. The device has many of the same features as Meta's existing Quest headsets. Its processor, cooling fan, battery, and other computing components are housed in an external pack. At about 100 grams, the device is roughly one-sixth the weight of Apple's Vision Pro.

CLAUDE DISCOVERS A NOVEL ENZYME SYSTEM WITH CRISPR-LIKE REPEATS

Claude autonomously discovered a novel enzyme system that is associated with an array of DNA repeats. The system's function is still unknown, but it is programmable and can perform operations like cutting, copying, and pasting DNA. It is based on a reverse transcriptase previously known to science, but Claude appears to be the first to notice the system's defining features - an associated array of non-coding DNA sequences and an additional accessory protein of unknown function.

GRAIL'S MULTI-CANCER BLOOD TEST GETS FDA COMMITTEE VOTE FOR PREMARKET APPROVAL

Grail has received a favorable vote from an FDA panel regarding premarket approval for its multi-cancer early detection test. The test is designed to detect cancer-specific methylation patterns and, if detected, predict the cancer signal origin with high accuracy to help guide diagnostic evaluation. While the FDA isn't bound to the committee's decision, it takes them into consideration when making the final regulatory decision. It is expected to make its final ruling on the test's premarket approval in the coming months.

CHINA PUTS AI COMPUTE INTO ORBIT WITH SUPERCOMPUTING-1 SATELLITE — ONBOARD PROCESSING AIMS TO CUT EARTH-OBSERVATION DATA PROCESSING FROM HOURS TO MINUTES

China has launched nine satellites into orbit as part of a push to put AI computing in space. The Supercomputing-1 satellite that launched on the mission is designed to capture and process Earth-observation data in orbit instead of sending data back to terrestrial data centers. It features both a high-res optical payload and an image-processing AI computer. Local processing will cut cross-regional data processing times from hours to minutes.

A REDISTRIBUTION OF CODE OWNERSHIP

Product managers and designers used to have to think through a great deal ahead of time because the build was a larger investment, and the people paying for it wanted to know what they were getting before they committed. Agent coding changes the process so that nobody has to hold the whole vision in their head and transfer it losslessly to someone else. Engineers have historically owned code out of necessity, but AI enables a model of ownership and collaboration that is distributed a little more evenly.

HOW WE MADE CLAUDE.AI 3X FASTER IN TWO WEEKS

Anthropic made the core user experience of claude.ai and the Claude desktop app about 3x faster during a two-week sprint in August. Time to a typeable page on a fresh load of claude.ai went from 3.1 seconds to 0.55 seconds, starting a new Claude Code session went from 0.8 seconds to 0.3 seconds, and loading a Claude Cowork cloud session went from 2.6 seconds to 0.73 seconds. The optimizations will likely save an estimated tens of thousands of user-hours of waiting every day. This post covers what the Claude team shipped, how they measured it, and the loop they built with Claude to do it safely.

JENSEN HUANG THINKS AI ALARMISM HAS GONE TOO FAR

Nvidia CEO Jensen Huang controls one of the central resources for training new AI models and using them to answer questions. He has become very influential in the Trump administration. While Huang is worried about AI safety, he sees it as a very solvable engineering problem. He does not want to see new regulation to change the direction things are going in.

META MUSE IS PEOPLE

Meta has been testing a human concierge for its Muse AI personal assistant. Human contractors are quietly handling some of the phone calls placed via the digital agent. The Muse agent is designed to autonomously perform tasks like sending emails, shopping, and booking travel on a person's behalf. The phone-calling feature allows users to tell the agent to dial phone numbers for businesses in the US and talk with people to handle errands like booking haircuts, checking whether a store has an item in stock, or getting quotes from different contractors.

tester, and audit-ready for your SOC 2. Book a demo.

CHATGPT IN SIRI 'PERSISTENTLY UNDERPERFORMING,' SAYS OPENAI

Elon Musk alleged that OpenAI and Apple's deal harmed xAI's growth and customer acquisition, but xAI can't prove harm cause the ChatGPT Siri integration just didn't draw in many new ChatGPT users.

BYD'S NEW ULTRA-LUXE EV HAS COACH DOORS AND MAY BE ITS FIRST WITH SOLID-STATE BATTERIES

BYD's upcoming Yangwang model features Rolls-Royce-like coach doors, 5-minute Flash Charging, and possibly solid-state batteries.

META DEBUTS DEDICATED ‘CHARM' DEVICE FOR USING MUSE AI

Muse Charm is a palm-sized, dedicated gadget for using Muse, Meta's new AI assistant.

I DON'T WANT THE DETAILS

When something goes wrong, don't talk about the details - focus on what will change so it doesn't happen again.

OPENAI'S AI TRIED BREACHING 4 OTHER TARGETS, WITHOUT PROMPTING

Three out of the four attempts were unsuccessful, but the Australian Medicare Statistics Reporting Service reported that OpenAI's AI acquired health data from its website.

YOUTUBE PROMISES CUSTOM FEEDS AND A LOT MORE AI LATER THIS YEAR

The annual Made on YouTube event introduced several AI features coming to the online streaming platform in the coming months, as well as a handful of non-AI updates.

Explore Mercury Books

_*Mercury is a fintech company, not an FDIC-insured bank. Banking services provided through Choice Financial Group and Column N.A., Members FDIC._

CLAUDE OPUS 5.5, GPT-6 SOL, GPT-6 LUNA, AND A NEW PRICE WAR

Anthropic and OpenAI both released new models within an hour of each other. OpenAI's GPT-6 Sol and Luna are half the price of the promotional pricing for GPT-5.6 equivalents. Opus 5.5 is cheaper per token than Opus 5.0 and has the intelligence of Fable 5.1. It is very token efficient and works across every effort level.

APPLE DEVELOPING NEW FITNESS TRACKER AIMED AT RIVALING WHOOP

Apple is developing a screenless health and fitness tracker as part of an effort to rethink its smartwatch lineup and add new wearable technologies. The project is still in a preliminary phase and likely won't debut before 2028. Apple is working on a major update to its smartwatch lineup for release as early as late 2027.

ALZHEIMER'S IS NO LONGER AN UNTREATABLE DISEASE, MAJOR REPORT CONCLUDES. HERE'S WHY

There are currently hundreds of clinical trials worldwide investigating possible medicines, new diagnostic options, and lifestyle interventions to identify, treat, or delay cognitive decline. The possibility of a cure is still far away, but it now seems more possible than ever before. Alzheimer's disease is being increasingly viewed as a condition for which earlier diagnosis, better understanding of disease biology, and a wider range of interventions may collectively improve outcomes. There is now a genuine case for optimism.

INTRINSIC OPEN SOURCES KEY PARTS OF ITS PLATFORM FOR EASIER DEVELOPMENT

Intrinsic Core is an open source set of Robot Operating System-compatible capabilities for building robotic applications. The capabilities act like the building blocks of robot behavior. The release allows users to quickly and easily combine capabilities to achieve adaptive AI systems. The capabilities are ROS-compatible.

LITEPARSE

LiteParse is a standalone OSS PDF parsing tool. Focused exclusively on fast and light parsing, it provides high-quality spatial text parsing with bounding boxes, without proprietary LLM features or cloud dependencies. It runs everything locally. A cloud-based version of the tool is available.

BUG BLINDNESS

Some bugs give normal users such a bad experience that they tell their friends about it. Developers and advanced users can be blind to these bugs because they're used to performing the non-intuitive workarounds required to get around them. Humans have a high ability to ignore the negatives in things they like. Being open to criticism can help developers notice more flaws in their work.

DEFENSIVE DRIVING FOR YOUR CAREER

To have a positive career, you need to be appropriately recognized and rewarded for your work. Take public credit in a way that isn't obnoxious or wasteful. Don't let anyone lie about you, and beware of people who think you got lucky. Maintain leverage at all times - leverage mainly comes from your ability to walk a valuable asset (your labor) out the door.

THE PLUNGING PRICE OF THOUGHT

The cost of a given level of AI performance has fallen an average of around 47% per quarter over the past three years. AI spending can still remain high despite falling prices. Lower prices allow for more agents to be run. Also, AI's ability keeps rising, and the price of harder tasks is higher.

See how it works

SHUFFLING THE DECK CHAIRS: XBOX CONTINUES ITS “RESET” WITH DRAMATIC RESTRUCTURING

Xbox has announced a radical restructuring of its internal studios designed to reduce the independence of some of its teams.

IT'S RAINING BOXES: AMAZON DRONES OVERWHELM A TEXAS SUBURB

Those living closest to the Amazon flight paths are hearing constant noise from the drones.

MYSTERIES OF AI GENERALIZATION

We still don't understand what's going on with AI generalization.

THE MUNDANITY OF EXCELLENCE, SMALL WINS, AND WHY YOU SHOULD FIX BUGS BEFORE WRITING NEW FEATURES

If you don't fix your bugs, your code will be built on an unstable foundation.

AI WILL WORSEN OUR ESTIMATES OF GDP

AI will make GDP estimation difficult because it is hard to measure quality.

ANALYZING JEV, A NEW AI MODEL

Jev is a close relative of LLMs, but it is actually more similar to the BERT series of models, which are used for classification, not generation.

GOOGLEBOOKS LAUNCH OCTOBER 4 STARTING AT $899—HERE ARE THE FIVE MODELS YOU CAN PREORDER TODAY

Google's new Android-powered Googlebooks laptops are intended to be 'real' computers with premium hardware, on-device apps, and powerful development tools. The first Googlebooks are now available from several OEMs. While Chromebooks were mostly budget machines, Googlebooks are priced much higher, with only one model under a grand. AI is a core part of the Googlebooks experience - Gemini models are going to be deeply integrated with the software.

SPACEXAI RELEASES GROK 4.7 FOR CODING AND KNOWLEDGE WORK

SpaceXAI has released Grok 4.7, a model for coding and knowledge work. Access is now open in Cursor, Grok Build, the Grok API, third-party coding harnesses, model routers, and cloud platforms. Pricing starts at $2 per million input tokens and $6 per million output tokens. There is a 'fast' variant that offers twice the output speed at twice the price.

A ROCKET SUPPLY CRUNCH IS MAKING IT HARDER TO HITCH A RIDE TO SPACE

Many rockets under development are behind schedule. Prices for many missions are rising. SpaceX is winding down the Falcon 9 to focus on its new Starship vehicle. The company has stopped selling missions where several companies share space on a Falcon 9. Launch demand is set to outstrip supply at least through the end of the decade.

AGILITY ROBOTICS: INSIDE THE FIRST US HUMANOID COMPANY TO GO PUBLIC

Agility Robotics is going public through a merger with Michael Klein's Churchill Capital XI at a $2.5 billion pre-money valuation. The company's robots are deployed in a few places today and can do tasks such as moving totes, feeding lines, and tending machines. Its Digit v5 robot will be out later this year. The new robot will be able to work alongside humans rather than in a cage/workcell. It will also support fast charging, a higher payload, and swappable end effectors.

JEV INTRODUCES A NEW SHAPE OF LLM—SYSTEM ONE, AKA DECISION MODELS

Jev is a new category of model called 'System One models' that accepts text inputs, but only returns floating point numbers corresponding to categories, yes/no questions, ratings, and associated confidence scores. It is fast and really cheap - output is free, and input is $0.042 per million tokens. The model was released just under a week ago, but the amount of activity around it has been extremely impressive.

A SUMMER OF AI OPTIMIZATION

Several open source projects recently started running faster due to being optimized with the help of AI. The techniques used to optimize these projects were well known. However, it is likely developers never tried them because it would be too risky. AI has made it cheap to try new ideas.

PRODUCT MANAGER, APPLIED AI AT TLDR ($200K BASE + $60K BONUS, FULLY REMOTE)

TLDR is hiring its first PM to help build the agent-first operating layer used across the company. We're looking for a builder who has shipped real products/systems with LLMs. Click here to learn more [13].