103K dots in 82 KB 🎨, JavaScript that resumes tomorrow 💾, more ambitious software 🚀
TLDR
Together With 
TLDR Dev 2026-10-26
Survey: Developers are concerned about giving AI agents cloud access. They do it anyway. (Sponsor)
Developers are feeling pressure to implement AI across the SDLC, despite serious concerns about security and quality.
Developers are feeling pressure to implement AI across the SDLC, despite serious concerns about security and quality.
A new developer survey found:
- Nearly 80% give AI agents access to the cloud, while 97% cited concerns about doing so
- Nearly 40% of AI-generated code is never released to production
- 58% feel pressure to control token and cloud costs, but lack the information to do so
Serious problems in the tech stack identified in the report may explain these challenges. Get the report
🧑💻
Articles & Tutorials
Shipping 103,802 Dots in 82 KB (11 minute read)
Layers compressed 13 hand-drawn dot illustrations into an 82 KB payload by quantizing coordinates and separating them into streams that expose repeated rows to Brotli. The result is 63% smaller than compressing the original floats, while typed arrays and an animation loop that sleeps when idle reduce rendering work.
Layers compressed 13 hand-drawn dot illustrations into an 82 KB payload by quantizing coordinates and separating them into streams that expose repeated rows to Brotli. The result is 63% smaller than compressing the original floats, while typed arrays and an animation loop that sleeps when idle reduce rendering work.
Reclaim goes AI-native without starting over (7 minute read)
Dropbox routes user edits, automated scheduling, and agent actions through one shared Schedule Action type so they use the same validation logic. Preview Mode makes proposed changes reviewable, backed by a scheduler refactored into a pure function and accelerated with Redis.
Dropbox routes user edits, automated scheduling, and agent actions through one shared Schedule Action type so they use the same validation logic. Preview Mode makes proposed changes reviewable, backed by a scheduler refactored into a pure function and accelerated with Redis.
How Databricks uses AI Runtime for research (6 minute read)
Databricks uses GPU health checks, fast container launches, and automatic MLflow tracking to support experiments that can span hundreds of GPUs. Agent-led research has increased training jobs eightfold, making startup time and reliable infrastructure central to the research loop.
Databricks uses GPU health checks, fast container launches, and automatic MLflow tracking to support experiments that can span hundreds of GPUs. Agent-led research has increased training jobs eightfold, making startup time and reliable infrastructure central to the research loop.
🧠
Opinions & Advice
Dead code shouldn't exist (8 minute read)
A conference talk shows how Sentry removed 28,000 lines of code with Knip. The page collects the slides, transcript, and recordings, with a practical look at finding unused code and making deletion part of maintenance.
A conference talk shows how Sentry removed 28,000 lines of code with Knip. The page collects the slides, transcript, and recordings, with a practical look at finding unused code and making deletion part of maintenance.
Pure UI in the Agentic Era (4 minute read)
A UI can pass an agent's spot check while still showing viewers a Delete button because role, plan, and billing state are scattered across components. Making display-relevant state explicit at a page boundary lets teams render combinations in a visual grid and turn them into stories or tests.
A UI can pass an agent's spot check while still showing viewers a Delete button because role, plan, and billing state are scattered across components. Making display-relevant state explicit at a page boundary lets teams render combinations in a visual grid and turn them into stories or tests.
🚀
Launches & Tools
What AI Agents Are Missing (Sponsor)
AI agents need context across systems and memory to reason effectively. Join Neo4j NODES 2026, a free, 24-hour online developer conference on November 12, featuring 100+ live technical sessions and a fireside chat with Andrew Ng.
AI agents need context across systems and memory to reason effectively. Join Neo4j NODES 2026, a free, 24-hour online developer conference on November 12, featuring 100+ live technical sessions and a fireside chat with Andrew Ng.
Tamagui 3 (25 minute read)
Tamagui 3 adds a C++ native style runtime, more native compilation, flat style props, React Strict DOM, and a Tailwind frontend. Eligible theme updates can reach native views without re-rendering their React components, but the runtime is an opt-in beta and the breaking release requires migration work.
Tamagui 3 adds a C++ native style runtime, more native compilation, flat style props, React Strict DOM, and a Tailwind frontend. Eligible theme updates can reach native views without re-rendering their React components, but the runtime is an opt-in beta and the breaking release requires migration work.
Rapier JXL (GitHub Repo)
Rapier JXL encodes JPEG XL in JavaScript with a roughly 30 KB gzipped core and WebGPU-accelerated quality search. It runs in browsers, workers, Node, and Deno, with no dependencies and support for lossless encoding, HDR, and JPEG transcoding.
Rapier JXL encodes JPEG XL in JavaScript with a roughly 30 KB gzipped core and WebGPU-accelerated quality search. It runs in browsers, workers, Node, and Deno, with no dependencies and support for lossless encoding, HDR, and JPEG transcoding.
dquickts (GitHub Repo)
This QuickJS fork can pause a JavaScript program mid-loop, serialize its heap, and resume execution in another process. It also runs TypeScript by erasing types during parsing and includes Rust bindings plus a REPL whose state survives restarts.
This QuickJS fork can pause a JavaScript program mid-loop, serialize its heap, and resume execution in another process. It also runs TypeScript by erasing types during parsing and includes Rust bindings plus a REPL whose state survives restarts.
🎁
Miscellaneous
GrokBot is amazing. Here's why I stopped using it (11 minute read)
A week with 16 always-on bots produced 78 merged pull requests, followed by $100 spent over a weekend largely on wakeups that yielded no useful messages. Missed notifications and unreliable follow-ups lead to an architectural argument for keeping routing, queues, and durable state in ordinary code, alongside a pitch for Builder's own framework.
A week with 16 always-on bots produced 78 merged pull requests, followed by $100 spent over a weekend largely on wakeups that yielded no useful messages. Missed notifications and unreliable follow-ups lead to an architectural argument for keeping routing, queues, and durable state in ordinary code, alongside a pitch for Builder's own framework.
Reverse engineering a Beyblade launch tracker (10 minute read)
A Bluetooth teardown works out how a Beyblade accessory reports launches without relying on its official app. It follows the process from inspecting the hardware to capturing payloads with LightBlue and PacketLogger, then identifying patterns in the transmitted data.
A Bluetooth teardown works out how a Beyblade accessory reports launches without relying on its official app. It follows the process from inspecting the hardware to capturing payloads with LightBlue and PacketLogger, then identifying patterns in the transmitted data.
⚡
Quick Links
Foxit API: Turn PDF Contracts into Structured Data (Sponsor)
Contracts hide the fields you actually need: who signed, when it renews, what the payment terms are. This tutorial shows how to extract them into structured JSON with Foxit's API, including messy layouts and scans, with working code you can drop into your own pipeline.
Extract your first contract
Contracts hide the fields you actually need: who signed, when it renews, what the payment terms are. This tutorial shows how to extract them into structured JSON with Foxit's API, including messy layouts and scans, with working code you can drop into your own pipeline.
Extract your first contract
We should be more ambitious with the software we build (25 minute read)
Better coding tools make projects such as database rewrites, regex engines, and compiler improvements feasible for smaller teams.
Better coding tools make projects such as database rewrites, regex engines, and compiler improvements feasible for smaller teams.
Fully customise your select elements (4 minute read)
Native select elements can use appearance: base-select and picker pseudo-elements for custom styling while retaining a usable default in unsupported browsers.
Native select elements can use appearance: base-select and picker pseudo-elements for custom styling while retaining a usable default in unsupported browsers.
Discourse adds Markdown endpoints (3 minute read)
Discourse can serve topics and topic lists as processed Markdown through a .md URL suffix or an Accept: text/markdown header.
Discourse can serve topics and topic lists as processed Markdown through a .md URL suffix or an Accept: text/markdown header.
Unicode Security Mechanisms (Website)
Unicode's security standard documents identifier restrictions, confusable detection, and mixed-script checks for software that accepts international text.
Unicode's security standard documents identifier restrictions, confusable detection, and mixed-script checks for software that accepts international text.
ADRs4AI (Website)
ADRs4AI keeps human and AI decisions in Git with templates that record options, rationale, open questions, and the final decision.
ADRs4AI keeps human and AI decisions in Git with templates that record options, rationale, open questions, and the final decision.
Love TLDR? Tell your friends and get rewards!
Share your referral link below with friends to get free TLDR swag!
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 [email protected] 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
Thanks for reading,
Priyam Mohanty, Jenny Xu