X algorithm explained 📱, Chrome AI auto-browse 🌐, the 80% problem in AI 🌗

X's feed algorithm, which determines what 550 million users see through a component-based pipeline, was recently open-sourced ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌  ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ 

Sign Up |Advertise|View Online TLDR Together With  TLDR Dev 2026-01-29 See real PRs where Greptile caught the bug (Sponsor) AI is writing more code, but code review is still a bottleneck for quality and velocity. Greptile reviews each PR with full repo context and learns your team's conventions from comments, reactions, and what actually gets merged. ✅ Trusted by engineering teams at ScaleAI, Zapier, and Whoop. 🧩 Now integrated with Claude Code: install via /plugin. 🎯 Free for open source. Try Greptile free for 14 days 🧑‍💻 Articles & Tutorials How X Decides What 550 Million Users See (9 minute read) X's feed algorithm, which determines what 550 million users see through a component-based pipeline, was recently open-sourced. The pipeline has sources like Thunder (for in-network posts) and Phoenix (for discovery), which feed candidates into a system of hydrators, filters, and multi-stage scorers. Posts are evaluated based on eighteen predicted engagement probabilities, weighted, and adjusted for author diversity and whether they are from followed accounts. We Fixed A 6-Year-Old JavaScript Memory Leak (4 minute read) DebugBear fixed a 6-year-old memory leak in its Google Cloud Functions caused by lodash's `memoize` caching all data indefinitely without clearing it. The leak only showed up with thousands of unique URLs in production, not in local testing with repeated URLs, so the company just lived with occasional crashes and auto-retries. 🧠 Opinions & Advice The 80% Problem in Agentic Coding (19 minute read) Many engineers now use AI agents to write 80-100% of their code, transforming their role from direct implementers to orchestrators. This adoption has introduced new problems, such as conceptual AI errors, comprehension debt within codebases, and a productivity paradox where review times balloon. Engineers need to shift to a declarative approach, focusing on problem definition and architectural oversight. Don't solve problems that don't exist (8 minute read) People frequently waste time solving problems that don't exist, driven by procrastination, unclear next steps, and a subconscious avoidance of discomfort. This tendency is made worse by Parkinson's Law and an aversion to high-uncertainty tasks. 🚀 Launches & Tools The new era of browsing: Putting Gemini to work in Chrome (7 minute read) Google is integrating major Gemini AI updates into Chrome, including a new side panel for multitasking across tabs and deeper integrations with other Google apps like Gmail and Calendar. For AI Pro and Ultra subscribers, Chrome now has "auto browse," an agentic feature that automatically browses the web to handle complex multi-step tasks such as vacation planning, online forms, and smart shopping. Introducing LibPDF: The PDF Library TypeScript Deserves (6 minute read) LibPDF is a new TypeScript-first document parsing library that supports lenient parsing, native digital signature support (eliminating complex workarounds like Rust bindings), and incremental saves vital for multi-party signing workflows. Introducing ReliCSS: A Tool for Front-End Archaeology (4 minute read) ReliCSS is a browser-based tool designed to scan CSS and identify "browser hacks" and historical artifacts. It parses CSS, highlights these findings with severity levels, and suggests ways to modernize or remove them. 🎁 Miscellaneous Suspiciously precise floats, or, how I got Claude's real limits (11 minute read) This dev reverse-engineered Claude's hidden usage limits and internal credit system by analyzing unrounded float values from the API. Findings show the 5x subscription plan has amazing value, providing much more usage (up to 8 times weekly) than advertised, while the 20x plan is less efficient for weekly tasks. Subscription plans are always superior to API pricing, mainly due to free cache reads and discounted cache writes. Adding dynamic features to an aggressively cached website (6 minute read) This dev implemented dynamic features on his aggressively cached website, which uses Cloudflare with a 15-minute cache header. He used localStorage and client-side JavaScript for features that bypass full-page caching. Examples include "Edit" links visible only to the administrator based on a `localStorage` key, and a "Random" button for navigating posts within a specific tag. ⚡ Quick Links Amazon axes 16,000 jobs as it pushes AI and efficiency (7 minute read) Amazon has confirmed 16,000 corporate job cuts, completing a plan for approximately 30,000 total reductions, as it streamlines operations, shuts down underperforming businesses, and embraces AI. Introducing LM Studio 0.4.0 (15 minute read) LM Studio 0.4, used to run AI locally on your computer, introduces a server-native daemon for headless deployment, parallel inference for high throughput, a new stateful REST API, and a refreshed user interface. A verification layer for browser agents: Amazon case study (9 minute read) Small local AI models are viable for executing browser tasks by using structured snapshots and explicit assertions as a verification layer. From Microsoft to Microslop to Linux: Why I Made the Switch (14 minute read) A long-time Windows user made a reluctant but ultimately successful switch to Linux due to Windows adding more ads and bugs into the user experience. 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.