People often default to private conversations instead of public ones, despite advocating for transparency. This reluctance comes from the perceived pressure of performing in public, the need for intimacy, the burden of providing extensive context, and the vulnerability of admitting ignorance. Some solutions to this are the "No DM Before Public" rule, using "WIP" labels to encourage early feedback, summarizing private decisions publicly, and leading by example by openly asking questions and admitting mistakes.
2025.10.31·
leadthroughmistakes.substack.com·9 min read·TLDR Dev
An engineer interviewed at eight large tech companies before accepting a Principal Software Engineer role at Atlassian. He completed over 60 interviews and found that algorithmic coding skills and behavioral interviews were still very important for senior-level positions. Even in a challenging job market, strong candidates with relevant experience are still in demand.
Zig's new async I/O primitives, coming in version 0.16.0, have core synchronization APIs like `async`, `await`, and `cancel` that will help with asynchronous vs concurrent execution.
Harbor uses atomic state management with Jotai in its React-based clinical trial data capture UI to improve performance in deeply nested component trees by preventing unnecessary rerenders.
Etsy improved the performance of product listing pages on its search results by implementing the Speculation Rules API (SRA) to prefetch those pages when users hover over listing cards, resulting in large performance improvements (20-24% in some metrics).
Developers often prematurely adopt complex technologies like Kafka when Postgres would suffice. Benchmarks show Postgres' impressive scalability for both pub/sub messaging and queueing, often meeting the needs of most organizations. Despite concerns about scaling, the article shows Postgres' scalability and suggests that many projects have ample time to migrate if necessary.
A developer built the same Kanban app 10 times to evaluate mobile performance, finding that next-gen frameworks (Marko, SolidStart, and SvelteKit) delivered 35-39ms load times with bundles 4-6x smaller than Next.js, which shipped 176.3 kB compressed and loaded in 467ms. It was found that React has an unavoidable performance ceiling, since TanStack Start with React delivered 118.2 kB compressed versus 60.4 kB with Solid using identical infrastructure, proving React's runtime adds 2x overhead.
OneUptime migrated from AWS to bare-metal two years ago, and they are now saving $1.2M/year. The migration involved running MicroK8s + Ceph with high availability, improving latency, and reinvesting savings into AI servers. They found bare metal cheaper than AWS after factoring in all costs, including bandwidth and control plane fees, and despite initial concerns, reliability improved.
uv, a new open-source Python package manager, makes Python installation and dependency management much easier. uv can install Python versions, manage virtual environments, and resolve dependency conflicts faster and more efficiently than other tools.
Developers are leaving Next.js for TanStack Start due to frustration with Next.js's growing complexity, constant changes, and tight coupling with Vercel's platform. TanStack Start feels more appealing because it feels closer to vanilla React with less “magic,” offering type-safe routing, integrated server functions, and flexible deployment without vendor lock-in.
Data-driven organizations require an agile, cloud-native database to support modern cloud infrastructure and application architectures. DISTRIBUTED SQL DATABASES FOR DUMMIES [11] provides a detailed explanation of how to meet today's cloud-native, GenAI, and retrieval-augmented generation (RAG) application demands using a
Emoji-Regex provides a regular expression to match all emoji symbols and sequences based on the Unicode Standard. It was built using `emoji-test-regex-pattern` and can be easily updated with new Unicode emoji releases, guaranteeing deterministic results.
Paul Mockapetris invented DNS to map domain names to IP addresses, creating a centralization point in the Internet. There was a domain gold rush where early investors registered domains for free and later sold them for millions (like sex.com for $12 million). This article covers ICANN's 1998 creation to govern TLDs, how registries bid for TLD rights in auctions (Google paid $25 million for .app), and dramatic geopolitical domain stories, like one about Yugoslavia's .yu domain, which was literally stolen when scientists cut network lines with scissors during the country's breakup.
Recurring disasters in microservice architectures are usually due to four specific problems: having more services than engineers, poorly designed gateway layers, unconstrained technology sprawl, and architectures mirroring organizational charts rather than domains. Distributed systems, including emerging AI agents, still suffer from fundamental challenges like latency and observability.
Meta's Llama 3.2 1B model is a useful, lightweight coding assistant for development work. It's not as bad as critics claim. However, it can't handle agentic coding and produces functional but imperfect code, constantly requiring manual fixes. Llama is best for simple projects like todo apps.
A Microsoft Azure outage, caused by an inadvertent configuration change, disrupted numerous services, including Microsoft 365 and Xbox, and impacted other companies like Starbucks and Alaska Airlines.
Minecraft: Java Edition is removing code obfuscation to make modding easier for creators, starting with the first snapshot after the Mounts of Mayhem launch.
Google's plan to require mandatory registration for Android app developers, including fees, identification, and app information, downgrades the platform's openness and users' freedom to install software.
Software engineers who have survived strokes should prioritize self-care, workload management, and use legal protections and AI to accommodate cognitive limitations.
Patreon scaled its platform to support live events by addressing the challenges of handling a surge of users. Its team focused on understanding and quantifying load, creating a load test to measure capacity, and then optimizing both infrastructure and client performance. To improve infrastructure, they fine-tuned web requests and reduced requests made at cold app launch for iOS.
Agentic AI systems have a fundamental security flaw: LLMs cannot separate instructions from data, making them vulnerable to prompt injection attacks where untrusted content contains hidden commands. The worst security occurs when an AI has access to sensitive data, reads untrusted content, and can communicate externally, allowing attackers to steal information by embedding instructions in sources like Jira tickets or web pages. To prevent this, it's best to run LLMs in sandboxed containers, avoid production credentials in files, limit untrusted content sources, and split tasks to isolate risks.
Best practices for writing great Go code include keeping main functions minimal, using named constants instead of magic values, preventing security holes with tools like os.Root, embedding static data with go:embed for single binaries, and making sure goroutines terminate properly using contexts and waitgroups. Code should work first with a “walking skeleton,” and then can be refactored after.
Personal agency, the capacity to take initiative and act decisively, is more necessary for success than raw intelligence. Intelligence can lead to analysis paralysis and fear of failure, while agency drives action, learning, and the creation of opportunities. In reality, agency is rare due to societal conditioning and psychological factors, but can be cultivated through a bias towards action and surrounding oneself with proactive individuals.
Amazon's private open house for senior engineering leaders in Seattle exposed three key themes defining Amazon's engineering culture: mission-driven purpose, clear alignment between titles and roles, and a focus on craft over perks. The event showed Amazon's commitment to customer obsession and solving problems at scale, with reversible decisions encouraged to move fast and one-way decisions approached with careful consideration.
Team underperformance often comes from lack of self-confidence rather than lack of skill, and managers inadvertently destroy confidence through micromanaging, public criticism, jumping in too quickly to help, and inconsistent expectations. To build confidence, managers should give autonomy with clear guardrails, criticize privately while praising publicly, ask questions before solving problems, and provide consistent encouragement and reassurance.
Code reviews should be less style nits and more "this is going to break prod". Sentry's AI Code Review blends context and issue history with the code you just touched [10] - function calls, class or objects dependencies, database connections - to provide specific and actionable feedback rather than generic linting advice. Read the blog
Dexto is a toolkit for building AI agents that can turn natural language into actions by orchestrating LLMs, tools, and data. It has a configuration-driven framework, runtime, and smooth developer experience, supporting various interfaces like CLI, Web UI, and APIs. Dexto helps the creation of autonomous agents, digital companions, and multi-agent systems, with features like session management, multimodal support, and production-ready observability.
Butter is a cache that saves money by identifying and serving repeat LLM responses, making AI systems deterministic. It's compatible with Chat Completions APIs like OpenAI and integrates with tools like LangChain and Crew AI. Butter is ideal for autonomous agents performing repetitive tasks. It is priced at 5% of the savings it generates, though currently free.
Microsoft and OpenAI have announced a new agreement extending their partnership. The deal involves Microsoft holding a large investment in OpenAI Group PBC following recapitalization, along with refined IP rights and continued exclusivity for Microsoft until Artificial General Intelligence (AGI) is verified by an independent expert panel. This new phase allows both companies to independently innovate, with Microsoft extending its IP rights and pursuing AGI independently, while OpenAI gains flexibility in developing products and providing API access.
Google's new developer verification policies basically end users' freedom to choose their software sources. This policy requires developers to register with Google, which goes against the open nature of Android.
The challenge of continual learning in AI models is updating model parameters without compromising existing knowledge. Memory layers, a sparse attention mechanism, is a potential solution for continual learning due to its targeted updates, high capacity, and adaptive integration. A recent paper shows that finetuning memory layers leads to much less forgetting compared to full finetuning or LoRA when learning new information.
Amazon has laid off about 14,000 corporate employees to reduce bureaucracy and shift resources towards investments in artificial intelligence following disappointing AI growth and pressure from investors to tighten finances.
The predictable, incremental improvements in Apple silicon Macs, now at the M5 generation, are a positive outcome and a fulfillment of what Mac users desired after years of inconsistent hardware updates during the PowerPC and Intel eras.
Using `await` in JavaScript loops requires careful implementation to avoid performance issues and unexpected behavior - it's usually better to use Promise.all or Promise.allSettled for reliability.
A critical security bug found in this dev's Next.js application was caused by a seemingly perfect one-line equality check. Due to Next.js's server function implementation, the synchronous function was unexpectedly converted to an asynchronous function returning a Promise. In JavaScript, a Promise evaluates to true in an `if` statement, effectively bypassing the security check and granting access to everyone.
Using fetch() over the newly available JSON module imports in browsers is preferred as it has better error handling with response.status and the ability to access source text even when JSON parsing fails. The main concern with imported modules is that they are cached for the page's lifetime, creating memory leaks when importing dynamic data like search results, while fetch() allows garbage collection of unused data.
This team used AI agents to write code, resulting in a 10x increase in coding throughput. This agentic coding requires a shift in traditional software development practices, especially in testing, deployment, and team coordination, to avoid increased bugs and bottlenecks. AI agents can also be used to reduce the costs of implementing testing approaches and improving the infrastructure needed to sustain high velocity.
An increase in requests for technical cofounders suggests a gap in what AI can currently accomplish. While AI is great at coding specific, isolated tasks, it struggles with the complexities of building and maintaining production-ready software. Software engineering is different from mere coding, as it involves integration, expansion, and long-term maintainability, which AI currently doesn't have.
On November 18-19, 2025, Tessl's AI Native DevCon is coming to Brooklyn, NYC! Speakers from Snyk, Thoughtworks, Netlify, and AWS will cover what's possible today - as well as what's still emerging: MCP security, open source tool, high-voltage developer workflows, and more. GET 50% OFF WITH CODE TLDR [10] or watch the action online (free) [10].
GPUI Component is a Rust library that offers a set of over 60 cross-platform UI components for building desktop applications with a modern design inspired by macOS, Windows, and shadcn/ui. It provides features like customizable themes, flexible layouts, high-performance virtualized lists and tables, a code editor with syntax highlighting and LSP support, and native Markdown and HTML rendering.
The Expo MCP Server connects AI coding agents to your Expo projects so they can access docs, automate workflows, and verify your app visually. It integrates with tools like Claude Code, Cursor, and VS Code, teaching them about the Expo SDK and enabling interaction with mobile simulators and React Native DevTools. Some local capabilities include taking screenshots, opening DevTools, and automation features like tapping views and finding elements.