TLDR 2023-11-08

OpenAI's Keynote 🤖, Neuralink's vision chip 👁️, can language models predict events 🧠

📱
Big Tech & Startups

Reddit is starting to test its own add-ons for the platform (2 minute read)

Reddit is starting to test add-ons built through its Developer Platform. Its first add-ons include tools for mods to more easily monitor and remove comment threads, ban spammers and remove their content, and more. They will only be available to a small number of subreddit mods. The Developer Platform is still in beta - it is unknown when it will be generally available.

The OpenAI Keynote (24 minute read)

OpenAI recently held its first-ever developer conference. This article presents an analysis of the keynote and its main announcements. The keynote was well-presented and entertaining. The live demos went off without a hitch and leveraged the fact that they were live. All of the new products and features announced were immediately available rather than being available in the coming weeks or months, as is increasingly the case for events like I/O or WWDC.
🚀
Science & Futuristic Technology

Elon Musk’s Neuralink is working on a chip that will solve blindness (2 minute read)

Neuralink is planning to release a vision chip that will solve eye conditions that cause blindness within the next few years. The company has received approval to start recruiting humans for its first clinical trials. Its v0.9 chip aims to fix various ailments, including quadriplegia due to cervical spinal cord injury or amyotrophic lateral sclerosis (ALS). Neuralink is still waiting for regulatory approval for its first human tests.

Cruise confirms robotaxis rely on human assistance every four to five miles (4 minute read)

Cruise uses a remote assistance team to help its autonomous fleet in certain situations. The company claims that remote assistance is being used about 2-4% of the time on average in complex urban environments. It says that the number is low enough that there isn't a huge cost benefit to optimizing much further. The remote assistants mostly provide wayfinding intel to the vehicle instead of controlling it remotely and more than 98% of sessions are answered within 3 seconds. It is an industry standard to have humans on call.
💻
Programming, Design & Data Science

inshellisense (GitHub Repo)

inshellisense provides IDE-style autocomplete for shells. It supports more than 600 command-line tools. inshellisense can run on Windows, Linux, and macOS. It works on bash, zsh, fish, pwsh, and Windows Powershell.

Cami.js (GitHub Repo)

Cami.js is a toolkit for interactive islands and state management in hypermedia-driven web applications. It features reactive web components, tagged templates, store/state management, easy immutable updates, and more. Cami.js doesn't require any build steps and has no client-side router, JSX, or shadow DOM. It was designed for lean teams or solo devs working on multi-page applications.
🎁
Miscellaneous

Are language models good at making predictions? (5 minute read)

This post presents the findings of a study that looks at whether GPT-4 can make good predictions. GPT-4's predictions are not well calibrated - its predicted probabilities are often over-confident. Its prediction ability varies depending on the category - for unclear reasons, GPT-4 is well-calibrated for sports questions but horrendously calibrated for personal questions. The model is better at making predictions for politics than for science or technology.

Seeing like a Bank (30 minute read)

Banks often look like notably disorganized institutions to their customers, which is alarming and surprising for an entity that is supposed to accurately keep track of all of the money. Banks are extremely good at tracking ledgers, but bad at tracking other forms of truth for structural reasons. In extreme cases, banks can feel like they lack object permanence - every interaction can feel like you're starting from the beginning. This article explains how it got to be this way.
Quick Links

Almost an Agent: What GPTs can do (10 minute read)

A look at the current capabilities of large language models and what they might be able to do in the near future.

A four year plan for async Rust (16 minute read)

This post looks at features that async Rust needs to continue to improve its user experience and a potential change to the language that could take years to plan and prepare for.

YouTube’s ad blocking crackdown is facing a new challenge: privacy laws (4 minute read)

YouTube's ad blocker detection system may be a violation of privacy in the EU as it may be classified as spyware.

Rivian and Amazon are no longer exclusive (2 minute read)

Rivian is now letting other companies buy its commercial electric vans.

44% of US Job Postings Have Salary Information in Them (4 minute read)

Louisiana has the lowest level of salary transparency, while South Dakota has the highest.

Beej's Guide to Network Concepts (Book)

This book is a guide to concepts seen in networking written to help readers make sense of the terminology and do a bit of network programming in Python.
Get the most important tech, science, & coding news in a free daily email. Read by +1,250,000 software engineers and tech workers.
Join 1,250,000 readers for