Search the archive

distributed SQL database architecture_._👉 DOWNLOAD THE EBOOK

INTRODUCING CLAUDE OPUS 4.5

Anthropic's Claude Opus 4.5 is the best model in the world for coding, agents, and computer use. It outperforms human candidates in technical skills assessments and shows creative problem-solving abilities, surpassing previous models in various benchmarks. Opus 4.5 is also the company's most robustly aligned model, with better security against malicious attacks.

INTRODUCING ADVANCED TOOL USE ON THE CLAUDE DEVELOPER PLATFORM

Anthropic has introduced three new beta features on the Claude Developer Platform to improve AI agent tool use: Tool Search Tool, Programmatic Tool Calling, and Tool Use Examples. The Tool Search Tool allows Claude to dynamically discover and load relevant tools, reducing context window consumption and improving accuracy by avoiding upfront loading of extensive tool definitions. Programmatic Tool Calling lets Claude orchestrate multiple tools through code execution. Tool Use Examples provides concrete examples of tool usage, clarifying ambiguous parameters and improving the reliability of tool invocations.

THREE YEARS FROM GPT-3 TO GEMINI 3

Gemini 3 can now code complex tasks, design interfaces, and conduct research, showing a shift from chatbots to agentic models capable of acting as digital coworkers. Gemini 3 is capable of PhD-level research, though it still requires human guidance.

THE BITTER LESSON OF LLM EXTENSIONS

LLM extensions over the past three years have shifted from complex protocols like ChatGPT Plugins and MCP to simpler, more accessible methods. Early attempts at extensive API integrations were hampered by model limitations, leading to the adoption of straightforward solutions like custom instructions and repo-level rules. The current trend, shown by Claude Code's Agent Skills, favors giving agents general-purpose tools with simple natural language instructions.

GOOGLE'S NEW 'ALUMINIUM OS' PROJECT BRINGS ANDROID TO PC: HERE'S WHAT WE KNOW

Google's "Aluminium OS" is a new Android-based operating system designed to replace ChromeOS on PCs.

I BUILT AN ACTUALLY FASTER NOTION IN RUST

Imed Adel left Stripe to build Outcrop, a faster, simpler knowledge base alternative to Confluence built in Rust.

LISAGUI RECREATES APPLE'S LISA INTERFACE IN YOUR BROWSER

LisaGUI is a JavaScript project that recreates the user interface of Apple's Lisa OS, the precursor to the Macintosh.

STAPPLY

Stapply is a visual, interactive map of AI jobs available around the world.

SHADERS

Shaders are small programs that run in parallel, using mathematical calculations and knowledge of x and y coordinates to create high-fidelity graphics. The graphics pipeline consists of vertex shading, rasterization, and fragment shading, each playing a specific role in determining the final color of each pixel. Various APIs like OpenGL, DirectX, Metal, and CUDA facilitate interaction with the GPU for both graphics rendering and general-purpose computing tasks.

TIL: `SATISFIES` IS MY FAVORITE TYPESCRIPT KEYWORD

The `satisfies` keyword in TypeScript acts as a type-safe cast, making sure an assignment is assignable to a provided type while allowing TypeScript to infer a more specific type based on the value. It's different from explicit variable type annotations, as it's best when specific type enforcement and value-based type inference are needed.

ABSOLUTELY EVERYTHING YOU NEED TO KNOW ABOUT ACT() IN REACT TESTS

The `act()` function in React tests makes sure that all state changes and side effects are fully processed before assertions run, preventing tests from checking against outdated state values. While React Testing Library functions like `userEvent` and `findBy...` automatically wrap their operations in `act()`, you'll need to manually use it when testing hooks with state changes, directly calling DOM methods like `button.click()`, or manually dispatching events.

ARE CONSUMERS JUST TECH DEBT TO MICROSOFT?

There's a potential shift in Windows market share among consumer computers. Microsoft's focus has shifted away from consumers towards AI and corporate clients, leading to a lack of excitement around Windows. The rumored affordable MacBook and Valve's Steam Machine are potential disruptors that could attract consumers and gamers away from Windows.

MARKDOWN IS HOLDING YOU BACK

Markdown has limitations for technical content due to its lack of structure and semantic markup. While Markdown is easy to use, its implicit typing and inconsistent flavors make content reuse, machine consumption, and scalability harder. Alternatives like reStructuredText, AsciiDoc, DocBook, and DITA have improved content structure and transformation capabilities.

WHEN EVERYONE'S A DEVELOPER, HOW DO WE PROMOTE THE WEB PLATFORM OVER REACT?

AI code generation tools are undermining web platform adoption by defaulting to React/Next.js solutions, especially as "vibe coders" (non-programmers using AI to create apps) increasingly rely on these tools without understanding native web capabilities. This trend is creating worse web experiences, and some solutions include teaching people to prompt for vanilla web solutions, creating open datasets of framework-free code, and spotlighting projects that use modern web features without heavy frameworks.

SMARTCOMMIT

Smartcommit is an AI-powered CLI tool for creating semantic, Conventional Commits messages. It analyzes staged changes, asks clarifying questions about the code's intent, and then generates a structured commit message.

TRAVELS

Travels is a fast, framework-agnostic undo/redo library that uses JSON patches to store only the differences between states. It is more memory-efficient than traditional snapshot-based systems. Travels uses Mutative for fast immutable updates and integrates with popular frameworks like React, Vue, and Zustand. It has a mutable mode for reactive states, an archive mode for controlling when changes are saved, and a maxHistory option to limit memory usage.

GITLOGUE

Gitlogue is a command-line tool that visually replays Git commit history with animated typing, syntax highlighting, and file tree transitions. It can be used as a screensaver, for educational purposes, presentations, or to create coding demos.

HOW WE PREPARE SHOPIFY FOR BFCM (2025)

Shopify prepared for Black Friday Cyber Monday (BFCM) from March to October by running bimonthly "fire drills" that simulated 150% of the previous year's peak traffic, conducting chaos engineering exercises called "Game Days," and systematically identifying and fixing infrastructure bottlenecks. Its preparation included capacity planning across multiple cloud regions, load testing with its tool, "Genghis," and creating a "Resiliency Matrix" to document vulnerabilities and response procedures.

IS AI CREATING A NEW CODE REVIEW BOTTLENECK FOR SENIOR ENGINEERS?

The increasing use of AI for code generation is creating a bottleneck in code review. Senior engineers are struggling to keep up. While AI can quickly produce code, integrating it with existing systems, ensuring security, and handling edge cases is still challenging.

MUX

Mux is a desktop application for parallel agentic development.

SUNSETTING SUPERMAVEN

Supermaven, acquired a year ago, is being sunsetted with refunds issued and free autocomplete inference provided for existing Neovim and JetBrains users.

MCP APPS: EXTENDING SERVERS WITH INTERACTIVE USER INTERFACES

The MCP Apps Extension is being proposed to standardize interactive user interfaces for MCP servers, allowing them to deliver rich UI experiences to hosts and enabling bidirectional communication.

"GOOD ENGINEERING MANAGEMENT" IS A FAD

Engineering management trends are fleeting fads driven by business realities rather than moral imperatives.

Schedule a demo to learn more

BRUTE-FORCEABLE AIRLINE RESERVATION API LEFT MILLIONS OF PASSENGER RECORDS VULNERABLE

A security researcher discovered a vulnerability in Avelo Airline's reservation API that allowed unauthorized access to passenger data through a brute-force attack due to the lack of last name verification and rate limiting. By exploiting this flaw, an attacker could have potentially accessed millions of passenger records, including PII, government ID numbers, and partial payment data, within a few hours. The researcher responsibly disclosed the vulnerability to Avelo on October 15, who quickly responded and patched the issues by November 13.

OPTIMIZING RUBY PERFORMANCE: OBSERVATIONS FROM THOUSANDS OF REAL-WORLD SERVICES

Based on analysis of over 3,000 real-world Ruby services, Datadog found that Ruby applications spend 82% of their CPU time in library code, making careful library selection necessary for performance optimization. Upgrading from Ruby 2 to Ruby 3 (with YJIT enabled) reduces CPU usage, while some popular libraries like mysql2 can be replaced with more efficient alternatives like trilogy for performance gains.

CODE EXECUTION WITH MCP: BUILDING MORE EFFICIENT AI AGENTS

MCP allows AI agents to connect to external systems, but as agents scale to use hundreds or thousands of tools, loading all tool definitions upfront and passing intermediate results through the context window creates inefficiencies and increased costs. By using code execution environments where agents write code to interact with MCP servers as APIs rather than making direct tool calls, agents can load only needed tools on-demand, process data locally before returning results to the model, and reduce token usage by up to 98.7%.

WE ARE REPLACING OOP WITH SOMETHING WORSE

OOP is not disappearing but shifting into new domains like microservices and distributed systems. The core principle of OOP, information attenuation and separation of concerns, is now being applied at the process and network level. This shift replaces OOP with a more cumbersome and complex system characterized by microservices, Docker, and Kubernetes.

WHY TOP FIRMS PARADOXICALLY FIRE GOOD WORKERS

Elite firms often hire top talent, train them extensively, and then let many go after a few years. This "up-or-out" system isn't a flaw but a strategic way for firms to signal talent and boost profits by creating a reputation for excellence. By periodically churning employees, firms maintain a perceived higher quality, allowing them to underpay those who remain while benefiting from their association with the prestigious firm.

PropelAuth. Read it now

INTRODUCING NANO BANANA PRO

Google DeepMind has launched Nano Banana Pro (Gemini 3 Pro Image), a state-of-the-art image generation and editing model built on Gemini 3 Pro. This new model improves image creation through improved reasoning, world knowledge, and real-time information integration. Nano Banana Pro allows users to generate accurate visuals with legible text, maintain consistency across multiple elements in complex compositions, and apply studio-quality creative controls like adjusting lighting and focus.

CC SWITCH

CC-Switch is a cross-platform desktop application for managing and switching between Claude Code, Codex, and Gemini API configurations. It supports MCP integration, API speed testing, config import/export, and multiple languages.

AI SLOP SECURITY ENGINEERING: OKTA'S NEXTJS-0AUTH TROUBLES

Joshua Rogers reported a security vulnerability in Okta's auth0/nextjs-auth0 project and submitted a patch. However, the maintainer used AI to rebase the commit, resulting in incorrect attribution and the removal of Rogers' contribution history. The maintainer then used AI to apologize for the error and refused to correct the commit, despite Rogers' request.

READONLY CHARACTERS ARE A BIG DEAL

Read-only characters in text editors help create interactive applications. Emacs allows marking specific substrings as read-only, enabling bidirectional interaction where both the user and the editor can modify different parts of the text. This contrasts with VS Code, where the lack of this feature hurts the creation of similar interactive tools.

ANDROID AND IPHONE USERS CAN NOW SHARE FILES, STARTING WITH THE PIXEL 10 FAMILY.

Quick Share on Android will now work with AirDrop, making file sharing easier between Android (starting with the Pixel 10 family) and iPhone devices.

SIX THINGS I BET YOU DIDN'T KNOW YOU COULD DO WITH CHROME'S DEVTOOLS

Some lesser-known Chrome Devtools features include `console.time()` functions, DOM element change monitoring, and being able to attach listeners to any function with `monitor`.

PARETO PRINCIPLE: THE SIGNIFICANT 20%

20% of bugs often result in 80% of the total failures.

PRESERVING CODE THAT SHAPED GENERATIONS: ZORK I, II, AND III GO OPEN SOURCE

Microsoft, in collaboration with Team Xbox and Activision, is open-sourcing the code for Zork I, II, and III under the MIT License to preserve gaming history and provide a valuable resource for education and exploration.

THE $1,000 AWS MISTAKE

Geocodio, a geocoding service, experienced a $1,000 AWS bill due to unexpected data transfer charges through a NAT Gateway when syncing data to S3. The team learned that traffic between EC2 and S3, even within the same region, can route through the NAT Gateway and incur costs. The solution was to implement VPC Gateway Endpoints for S3, which provide a free, direct route, bypassing the NAT Gateway.

PARTIAL PRERENDERING

Next.js's Partial Prerendering (PPR) is a new rendering paradigm that optimizes performance by serving static parts of a page immediately while allowing dynamic components that need request data to render on the server. The first implementation in Next.js 14 used error throwing to detect request data access, but this caused issues with existing applications that had try/catch blocks and retry logic. The updated approach in Next.js 15 switches to a Promise-based system where request APIs like `cookies()` and `headers()` become asynchronous and return never-resolving promises during prerendering.

JUST JAVASCRIPT

"Just JavaScript" is code that works without custom transforms, allowing devs to understand and refactor their code easily. Svelte and React are not "just JavaScript" due to their unique languages/semantics and custom transforms, respectively. However, Remix 3 adheres to this principle.

WHY TRADITIONAL SAAS PLAYBOOKS FAIL FOR AI-NATIVE PRODUCTS

Traditional SaaS go-to-market strategies don't work for AI-native products. Unlike traditional SaaS, which focuses on perfect MVPs, narrow customer profiles, and monetization-first approaches, AI-native products require breaking these rules by launching imperfect products, building viral distribution channels first, and targeting broader early customer profiles who can provide feedback and social sharing. AI-native companies must prioritize building "omnipresent" distribution channels and momentum over traditional metrics like product-market fit and willingness to pay.

WHAT AI IS REALLY FOR

The AI boom is overhyped, potentially a bubble, and may not deliver on its grand promises, especially in design. While AI can be useful in small-scale applications, its implementation doesn't actually result in as many savings as expected. Instead, the true motive behind the AI frenzy is the acquisition of resources like land, energy, and water needed to power massive data centers.

In the meantime, check out another can't-miss event from Atlassian, focused on strategies to future-proof your engineering org: The modern

tech leader's playbook for AI-powered teamwork