TLDR 2025-05-20
iPhone 17 Air leaks 📱, Microsoft's natural language web 🌎, Claude Code SDK 👨💻
OpenAI Admits You.com's Deep Research is More Accurate Than Its Own (Sponsor)
According to benchmarks,
ARI beats OpenAI's Deep Researchwith a 76% Win Rate.Put it to the test! You.com is giving TLDR readers their Team plan ($375 value), including:
📊 5 ARI reports – deep research agent that analyzes 400+ sources across private, public, and premium data to produce fully-customizable, McKinsey-style reports in < 5 mins
🧠 Access to 31+ latest AI models in one place – includes ChatGPT, Claude, Gemini, Llama, Grok, DeepSeek, and more
🤖 AI Agents – automate any work-related sequence as an agentic workflow that can be shared across teams for huge productivity gains.
↗️ Claim your FREE gift & try now. Offer expires 5/31.↗️
iPhone 17 Air Battery Capacity and Weight Allegedly Revealed (2 minute read)
Apple's iPhone 17 Air, which is expected to launch later this year, will weigh approximately 145 grams. The 6.6-inch, 5.5 mm-thin device will have a battery capacity of 2,800mAh and may use Apple's power-efficient C1 modem. There have been mixed reports regarding the iPhone 17 Air's potential battery performance, but Apple apparently determined that 60% to 70% of users will be able to use the device for a full day without needing to recharge throughout the day. The company is reportedly planning to release a battery case as an optional accessory to mitigate this problem.
NLWeb is Microsoft's project to bring more chatbots to web pages (2 minute read)
Microsoft's NLWeb project enables websites to provide a conversational interface for users with a few lines of code, the AI model of their choice, and their own data. Sites using NLWeb can make their content discoverable and accessible to platforms that support MCP. NLWeb could play a similar role to HTML for the agentic web and allow users to interact directly with web content in a rich, semantic manner. The NLWeb project started with OpenAI, which was working on an early version last November.
🚀
Science & Futuristic Technology
Nvidia opens up speedy NVLink interconnect to custom CPUs, ASICs (4 minute read)
Nvidia's NVLink Fusion, introduced at Computex this week, opens the NVLink interconnect tech to non-Nvidia-designed accelerators. NVLink enables multiple GPUs in a system or rack to behave like a single accelerator with shared compute and memory resources. The technology will be offered in two configurations: the first will be for connecting custom CPUs to Nvidia GPUs, and the second involves using NVLink to connect Grace and Vera CPUs to non-Nvidia accelerators. This should theoretically open up the possibility of superchip-style compute assemblies that feature any combination of CPUs and GPUs as long as Nvidia technology is involved.
23andMe and its user data will soon belong to a pharmaceutical giant (2 minute read)
23andMe will be sold to Regeneron Pharmaceuticals for $256 million. The sale includes 23andMe's Total Health and Research Services business and biobank of customer data and genetic samples. 23andMe says that customer data is anonymized and stored genetic samples are destroyed when users delete their accounts, but it is unclear how much information is retained. Regeneron's co-founder said that the purchase will further the company's large-scale genetics research into future drugs and treatments. 23andMe has collected genetic samples and data from more than 15 million customers.
💻
Programming, Design & Data Science
GitHub Copilot coding agent in public preview (2 minute read)
Copilot coding agent lets developers delegate issues so they can focus on the creative, complex, and high-impact work that matters most. Developers simply assign issues to Copilot as they would another developer, and Copilot will work in the background using a secure cloud-based development environment. The agent will explore the repository, make changes, and validate its work before it pushes. Developers can ask Copilot to make changes to pull requests by leaving comments. The agent excels in low-to-medium complex tasks in well-tested codebases.
Claude Code SDK (7 minute read)
Anthropic's Claude Code SDK allows developers to programmatically integrate Claude Code into their applications. It enables running Claude Code as a subprocess. The SDK currently supports command-line usage. TypeScript and Python SDKs are coming soon. This page provides a guide on basic SDK usage.
Join Fern and help us build the future of delightful developer experiences! (Job Posts, $170k-$230k + equity, Brooklyn)
There's No Free Lunch at Apple (8 minute read)
Apple is known for not giving its employees free lunches - even Steve Jobs paid for his own lunch. However, recently, the team working on Apple Intelligence were given vouchers to eat for free, breeding resentment among other teams. The team still shipped late. This article looks at the multiple failures within Apple's AI efforts. Apple may need to make some big changes to catch up, such as a big acquisition. While it doesn't usually make such moves, the current situation isn't typical - the company needs to let go of the past and operate in the future.
Vibe coding is rewriting the rules of technology (20 minute read)
Vibe coding represents a fundamental reimagining of the software development process. Instead of crafting lines of code, developers get to focus on vision and creative direction. It has the potential to transform technology development for the better - anyone with an idea for how technology could solve a problem can create that solution without specialized technical training. Vibe coding isn't just a technical shift, it's a cultural one that challenges assumptions about who gets to create technology and how that creation happens.
AI Engineer World's Fair (Sponsor)
Huawei is looking to surpass the billion-user mark with its HarmonyOS Next operating system, as it can no longer use Microsoft Windows (3 minute read)
What began as a survival strategy against US sanctions has evolved into an initiative to position HarmonyOS as a dominant operating system across smartphones, tablets, TVs, and PCs.
The principles of database design, or, the Truth is out there (3 minute read)
Poor design leads to semantic confusion and technical instability - this post presents a list of design principles to follow for formal database design.
Microsoft CTO Kevin Scott on how AI can save the web, not destroy it (56 minute read)
A transcript of an interview with Microsoft's Chief Technology Officer, Kevin Scott, where he talks about the future of search, Microsoft's relationship with OpenAI, Scott's feelings about AI's capabilities scaling up over time, and his thoughts on AI as a creative tool.
Turning Feedback Into Features: Building My npx Business Card (5 minute read)
This developer built custom terminal business cards to give others a way to connect to them that felt a little different.
LLM Memory (19 minute read)
This post looks at different ways to store information for large language models to access later in the interaction.
Judge pressures Apple to approve Fortnite or return to court (3 minute read)
Apple has been asked to approve Epic's submission for Fortnite on the US App Store or return to court to explain the legal basis as to why it has not done so.
Get the most interesting stories in startups, tech, and programming delivered in a free daily email.
Join 1,600,000 readers for
one daily email