Rust at scale 📈, how Cursor shipped Composer 🌱, Karpathy on Claude Code 🧠

WhatsApp added a layer of security for three billion users by replacing 160,000 lines of C++ with 90,000 lines of Rust. ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌  ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ 

Sign Up |Advertise|View Online TLDR Together With  TLDR Dev 2026-01-28 Report: 96% of devs don't fully trust AI code (Sponsor) AI is accelerating code generation, but it's creating a bottleneck in the verification phase. Based on a survey of 1,100+ developers, Sonar's newest State of Code report analyzes the impact of generative AI on software engineering workflows and how developers are adapting to address it. Findings include:  96% of developers don't fully trust that AI-generated code is functionally correct yet only 48% always check it before committing 61% agree that AI often produces code that looks correct but isn't reliable 24% of a developer's work week is spent on toil work Download now 🧑‍💻 Articles & Tutorials Rust at Scale: An Added Layer of Security for WhatsApp (5 minute read) WhatsApp added a layer of security for three billion users by replacing 160,000 lines of C++ with 90,000 lines of Rust. This global deployment, called Kaleidoscope, mitigates memory safety vulnerabilities in media processing. The cross-platform migration improved performance and runtime memory usage across Android, iOS, and web clients. How Cursor Shipped its Coding Agent to Production (12 minute read) Cursor's team shipped Composer, a coding agent that completes most tasks in under 30 seconds, by solving three gnarly engineering problems: the "diff problem" (models aren't great at making clean code edits, so they trained specifically on edit trajectories and search-replace operations), compounded latency (each step in the agent loop adds delay, so they used MoE architecture, speculative decoding with draft models, and aggressive context compaction to stay fast), and sandboxing at scale (they built custom VM infrastructure to spin up thousands of isolated execution environments without the startup time becoming a bottleneck). 🧠 Opinions & Advice A few random notes from Claude coding quite a bit last few weeks (5 minute read) Andrej Karpathy has had a rapid shift in their coding workflow, moving from 80% manual to 80% LLM agent-driven programming within weeks. While LLMs boost productivity, stamina, and project scope by handling code generation, they still produce subtle conceptual errors, overcomplicate code, and necessitate close human oversight within an IDE. Despite these flaws, the LLM-assisted workflow makes coding more enjoyable by removing drudgery and expanding engineering capabilities. We're All Beginners Again (2 minute read) Early AI adoption mirrors the early Python ecosystem, restoring play to programming. Since no definitive advanced standards exist, developers should embrace experimentation over legacy maintenance. 🚀 Launches & Tools 💣 v0 by Vercel is vibe coding without the security time bomb (Sponsor) Most AI coding tools optimize for speed and novelty, which risks leaked credentials and company data. Need something that can actually go to prod? v0 is built on the same infrastructure that's handled hundreds of millions of deployments and blocked hundreds of billions of threats in the last year. Start here Open Coding Agents: Fast, accessible coding agents that adapt to any repo (11 minute read) Ai2 has launched Open Coding Agents, an initiative to address the limitations of existing closed and expensive coding agent systems. They are releasing SERA (Soft-verified Efficient Repository Agents), a family of strong open coding models and an accessible training method. This method significantly reduces the cost and complexity of fine-tuning powerful coding agents on private codebases, allowing small teams to achieve state-of-the-art performance. Ultimately, Ai2 aims to democratize access to agentic coding, making it more reproducible and adaptable for researchers and developers worldwide. Prism (Website) OpenAI has launched Prism, a free, AI-native workspace to accelerate science writing and collaboration for researchers. Powered by GPT-5.2, this cloud-based, LaTeX-native platform integrates drafting, revision, collaboration, and preparation for publication into a single workflow. 🎁 Miscellaneous From Prompts to Clicks: Shipping MCP Apps in Production (17 minute read) Monday.com has successfully implemented MCP Apps, an extension to the Model Context Protocol, allowing AI tools like ChatGPT and Claude to return interactive UI components directly within chat conversations. This approach enables users to "ask, see, and act" on information without leaving the chat, but it's important to only show UI selectively where seeing beats reading. AGENTS.md outperforms skills in our agent evals (9 minute read) AI coding agents struggle with outdated framework knowledge, leading to incorrect code generation. Researchers from Vercel evaluated two approaches for providing current Next.js API documentation: "skills" (on-demand knowledge packages) and "AGENTS.md" (a persistent, embedded markdown file). Surprisingly, a compressed 8KB docs index embedded directly in AGENTS.md achieved a perfect 100% pass rate in their evals, outperforming skills, which only reached 79% even with explicit triggering instructions. ⚡ Quick Links Doing the thing is doing the thing (2 minute read) Only actual execution, even if imperfect, constitutes "doing the thing," as opposed to contemplation, planning, or discussion. Clawdbot renames to Moltbot (GitHub Repo) Clawdbot has been renamed to Moltbot due to the original name being too close to "Claude," which is trademarked. The Age of Pump and Dump Software (5 minute read) A worrying new trend in 2026 is where easily generated, unmaintainable vibe-coded software is used by crypto scammers and tech FOMO for pump and dump schemes involving associated cryptocurrencies. One Human + One Agent = One Browser From Scratch (7 minute read) Working with a single LLM agent, the author successfully built a basic, cross-platform HTML/CSS browser from scratch in approximately 70 hours and without third-party Rust dependencies. Love TLDR? Tell your friends and get rewards! Share your referral link below with friends to get free TLDR swag! https://refer.tldr.tech/d0af4569/3 Track your referrals here. Want to advertise in TLDR? 📰 If your company is interested in reaching an audience of web developers and engineering decision makers, you may want to advertise with us. Want to work at TLDR? 💼 Apply here, create your own role or send a friend's resume to jobs@tldr.tech and get $1k if we hire them! TLDR is one of Inc.'s Best Bootstrapped businesses of 2025. If you have any comments or feedback, just respond to this email! Thanks for reading, Priyam Mohanty, Jenny Xu & Ceora Ford Manage your subscriptions to our other newsletters on tech, startups, and programming. Or if TLDR Dev isn't for you, please unsubscribe.