TLDR 2025-07-31

TikTok's Footnotes is officially rolling out in the US for users in the pilot program. There are around 80,000 users contributing to TikTok's version of Community Notes - all of them have had active TikTok accounts for at least six months, live in the US, haven't had community guidelines violations in the past six months, and are at least 18 years old. Footnotes will be moderated with a combination of automated moderation and human moderation. Users can report footnotes they think are untrue or break TikTok's rules.

Tesla could be sunsetting the Model S and Model X as it has ended international orders. The two oldest cars in the company's lineup, the vehicles account for a very small portion of overall sales. Tesla doesn't make enough of the Model S or Model X units to justify the expensive logistics process of shipping custom orders overseas. It is possible the company may build a bunch of random configurations, send them overseas every few months, and let them sell before replenishing inventory.

Clinical trials for a drug that could eventually help people regrow missing or damaged teeth have begun in Japan. While humans typically grow two sets of teeth in their lifetime, the buds necessary to grow a third set still exist. The drug blocks a protein that limits tooth growth. The researchers hope to roll the drug out to the wider public by 2030.

Helion Energy has started site work on its commercial fusion power plant. The company's first facility is scheduled to start operating in 2028. It is building a 50 megawatt reactor to power Microsoft's data centers in the region. The entire tech sector is scrambling for new clean energy sources to power the servers that support artificial intelligence. Helion has spent 12 years on research and development and building prototypes in preparation for this deployment.

Delve ran just 4 ads in TLDR newsletters. The payoff?

• 66 high-quality leads (including enterprise companies)
• $1M in pipeline
• 52x ROI

We break down Delve's exact strategy, detailed results, and share the actual ads that crushed it.

Read the Delve case study.

Choose novel technology when trying to learn something new and use boring technology when trying to solve a problem. Boring technologies have known failure modes, well-understood capabilities, and proven operational reliability. The advent of large language models and agent AI coding tools has made this principle even more critical. Developers who use technologies unknown to them have no way of verifying whether AI is leading them down the wrong path.

Fast software changes behavior. Fast eliminates cognitive friction and signals simplicity, which is rare in the world where code and content are commodities. Making software fast often requires stripping away non-essential features - the ultimate expression of respect in a world obsessed with adding rather than refining. It means the developer has thought deeply about what matters and eliminated everything else.

ChatGPT now seems to understand what it reads. The use of reasoning models shows that ChatGPT can be creative. It can summarize texts extremely well, a robust measure of intelligence. AI products now pay for AGI research, which means that AI has reached early stages of the long-awaited self-improvement loop.

Mark Zuckerberg says that the development of superintelligence is now in sight. He is extremely optimistic that superintelligence will help humans accelerate their pace of progress and begin a new era of personal empowerment where people will have greater agency to improve the world in the directions they choose. Meta's vision is to bring personal superintelligence to everyone to help them achieve goals, create what they want to see in the world, experience any adventure, and grow to become the person they aspire to be. It believes that the benefits of superintelligence should be shared with the world as broadly as possible.

A look at the current ecosystem of chat protocols, the problems that exist, and possible solutions to make a better open chat protocol.

Vibe coding is perfect for prototypes and throwaway projects because the code doesn't have to be maintained - anything more serious needs a more careful approach.

The single biggest cost for any metro system is the stations - a system that uses smaller trains would result in smaller stations and huge cost savings.

The big hidden risk with automation through large language models is that it encourages mental disengagement, resulting in engineers overestimating agents' capabilities.

Meta and Stanford recently released a paper describing a prototype pair of VR glasses with a total optical stack thickness of less than 3 millimeters.

Lean is a programming language that lets mathematicians treat mathematics as code.