HN 500

My $500M Mars rover mistake

The author describes a mistake he made while testing motors for NASA's Spirit Mars rover at JPL. During a test of the rover's rock abrasion tool motor, he accidentally sent an electrical surge into the spacecraft instead of the motor. Fearing he destroyed the $500 million rover with only 2 weeks until launch, he informed his colleagues and wrote a report. After examining what happened, they discovered the surge likely glitched the motor driver circuit without permanent damage. Attempts to restart the spacecraft failed. The next day, they realized he had inadvertently disconnected the multimeter that powered telemetry during motor testing. Restoring that connection brought the spacecraft back to life, revealing that Spirit had not been destroyed after all due to his mistake. It was a frightening failure turned recovery story for the Mars rover mission.

Charlie Munger has died

The news release announces that Charlie Munger, vice chairman of Berkshire Hathaway, has passed away at age 98. Warren Buffett, CEO of Berkshire Hathaway, expresses that the company would not have achieved its current success without Munger's inspiration, wisdom and involvement. Munger played a pivotal role in building Berkshire Hathaway over many decades. The family will handle affairs according to Munger's instructions. Marc Hamburg is listed as the contact for Berkshire Hathaway. In summary, the document announces the passing of Charlie Munger, longtime vice chairman of Berkshire Hathaway, and credits his significant contributions to the company's growth and success over the years.

MeshGPT: Generating triangle meshes with decoder-only transformers

MeshGPT introduces a new method for generating triangle meshes using decoder-only transformers. It learns a vocabulary of latent embeddings to represent local mesh geometry and topology, which are then sequenced and decoded into triangles. A transformer is trained to predict the next embedding based on previous ones, allowing for autoregressive generation of new meshes. Generated meshes are compact with sharp edges, resembling human-crafted designs. MeshGPT demonstrates improved shape coverage and FID scores over baselines. It can complete partial meshes by inferring multiple possibilities and generate 3D assets to populate scenes. The method produces high-fidelity meshes while avoiding over-triangulation seen in other approaches.

My toddler loves planes, so I built her a radar

The author built a radar app for his plane-obsessed toddler daughter to track nearby flights. He created prototypes using location data and OpenSky Network's API. Through iterations adding orientation, accurate plane positioning, and a retro scanner interface, the MVP was a success. Further improvements like scaling by altitude and true flight direction enhanced the experience. Publishing to App Store required screenshots on older devices and licensing clarification. User testing informed additional customization options and sensory elements. Now available, the app aims to expand tracking capabilities and add fanciful modes for continued enjoyment. The project highlights an inspiring example of designing with a child's interests and rapid feedback in mind.

$20k bounty was claimed

Prettier, a JavaScript code formatter, offered a $20k bounty for a Rust project that could pass 95% of its test suite. The Biome project claimed the bounty after only 3 weeks of work, providing a faster alternative implementation. This competition helped uncover bugs and inefficiencies in Prettier's own code. The bounty was made possible by donations from companies like Indeed and Frontend Masters as well as individuals, allowing Prettier's maintainers to dedicate more time to the project. Going forward, the improved competition from Biome should push Prettier to further optimize performance. The author calls for additional donations to continue funding development for at least 8 more months.

Why Bother with uBlock Being Blocked in Chrome? Time to Switch to Firefox

This article discusses various privacy-oriented web browsers that provide alternatives to mainstream browsers like Chrome and Safari. It divides browsers into categories of mainstream privacy, alternative privacy, and top anonymity. Firefox is recommended for general users due to its extensive privacy extensions and controls. DuckDuckGo is also recommended for its built-in ad blocking and anti-tracking. The Tor Browser is highlighted as the best option for anonymity since it routes traffic through volunteers to hide a user's identity and IP address. Other browsers mentioned include Brave, Mullvad, and Puffin Secure, with each having their own unique privacy approaches. Overall, the article evaluates different browser options based on their privacy features, open-source policies, and ability to provide security and anonymity for users.

Brother have gotten to where they are now, by not innovating

The document discusses printers and the brand Brother. It notes that "Brother" has become the brand that many tech experts recommend for printers simply because other printer brands have overly innovated, making their products worse over time with each new feature. However, Brother printers have remained basic, only providing the necessary printing functionality without unnecessary additions. While once seen as inferior for their lack of new features, Brother printers are now seen as the best option since other printer manufacturers have degraded their products through non-stop innovations. The document finds it fascinating that Brother has risen to the top of recommendations in the printer space not by innovating more but rather by innovating less. It argues the company has succeeded by avoiding the "Who Can Get Crappiest Fastest race" that other printer brands engage in.

The document highlights how Brother printers have become the recommended brand not through superior features or technology, but rather by opting out of the constant race to add unnecessary new functions that tend to degrade overall printer quality and experience over time at other manufacturers.

Google Drive files suddenly disappeared

Many users reported that their Google Drive files suddenly disappeared after a Drive update, reverting to the state from May 2023. Thousands of files went missing, in some cases dating back months of work. Google support is investigating but has not yet determined the root cause or provided a restoration solution. Comments indicate the issue may be widespread, with engineers now involved to understand what happened. For impacted users, the offline cache stored in the DriveFS folder contains their only remaining copies of the missing files, though support has been unable to leverage this for restoration so far. This unexplained data loss represents a serious issue that Google must resolve urgently to recover users' work.

The Dunning-Kruger effect is autocorrelation

The Dunning-Kruger effect, which suggests less competent people overestimate their abilities, is not actually a real psychological phenomenon. It is a statistical artifact known as autocorrelation that occurs when a variable is correlated with itself. The original 1999 study that discovered the effect mixed test scores into both axes of their analysis chart, essentially correlating variables with themselves. Several later studies from 2016-2020 exhaustively debunked the effect, showing it disappears when measures are taken independently. However, most scientists are still unaware the Dunning-Kruger effect has been discredited as it is a very compelling idea. The paper argues the irony is that in their seminal work, Dunning and Kruger were the ones demonstrating statistical incompetence by confusing autocorrelation for a psychological bias. In reality, there is no evidence people are systematically biased in their self-assessments based on their actual ability level.

Sqids – Generate short unique IDs from numbers

Sqids (formerly Hashids) is an open-source library that can generate short, unique IDs from numbers in the style of YouTube video IDs. These IDs are compact but still avoid collisions even when generated from a custom alphabet. The library allows web applications to use IDs in place of numbers for a more visual and compact representation. For example, instead of showing a numeric user ID, a website could display a shortened string ID. Sqids is useful for this purpose of translating numeric values into brief, readable identifiers. In addition to brevity, the randomized strings ensure each ID remains distinct even when the source numbers are close in value. The documentation suggests Sqids as a potential choice for applications wanting to represent numbers visually with compact, unique identification codes.

The Bond villain compliance strategy

Binance operated as a global cryptocurrency exchange for years while actively skirting regulations through lies and shifting locations. As the exchange grew enormously in scale, regulators increasingly scrutinized its illegal activities like money laundering. The document describes how Binance misled authorities in countries like Japan, China, Malta and the U.S. regarding its real operations. Eventually the U.S. was able to build a massive criminal case against Binance through monitoring and informants within the organization. In late 2023, Binance pleaded guilty to operating the world's largest criminal money laundering scheme and agreed to pay over $4 billion in fines. As part of the settlement, Binance must now open its records to U.S. authorities and help in further investigations and prosecutions across the crypto industry. However, many see this as just the beginning of Binance's regulatory troubles rather than an end to its legal issues.

Autonomous excavator constructs a six-metre-high dry-stone wall

Researchers at ETH Zurich deployed an autonomous excavator called HEAP to construct a six-meter-high and sixty-five-meter-long dry-stone wall. The wall was built as part of a digitally planned and autonomously excavated landscape and park. HEAP uses sensors to map the construction site and locate existing stones, and has specifically designed tools and machine vision to scan and grab large stones. An algorithm determines where each stone should be placed, and the excavator then autonomously performs the task of placement. Remarkably, the autonomous machine can place 20 to 30 stones in a single operation, which is comparable to the amount that could be supplied in a single delivery. The project was a collaboration between Gramazio Kohler Research, the Robotics Systems Lab, Vision for Robotics Lab, and the Chair of Landscape Architecture as part of the NCCR dfab.