Level Up Coding

𝗥𝗔𝗚 𝘃𝘀 𝗔𝗴𝗲𝗻𝘁𝗶𝗰 𝗥𝗔𝗚. With 𝗥𝗔𝗚, retrieval is a step, not a decision. The query goes in, the top documents come back, and the model answers with whatever it found. If that first retrieval misses, the answer can miss with it. That's where 𝗔𝗴𝗲𝗻𝘁𝗶𝗰 𝗥𝗔𝗚 changes the flow. An agent wraps retrieval in a loop. It can rewrite a weak query, search another source, or decide the context isn't good enough and try again before committing to an answer. But more flexible retrieval doesn't necessarily mean better answers. Whether retrieval is predefined or agent-driven, the system still needs to find the right evidence. Keyword search works well for exact matches, vector search for semantic similarity, and SQL for structured data. Hybrid search can combine approaches when needed. The challenge is knowing which approach actually retrieves the most relevant evidence. Oracle's guide to Production RAG Evaluation explores how to compare these retrieval strategies, measure answer quality, and test for common production failures like stale data and missing evidence. 𝗥𝗲𝗮𝗱 𝘁𝗵𝗲 𝗳𝘂𝗹𝗹 𝗴𝘂𝗶𝗱𝗲 → lucode.co/guide-to-productio… What else would you add? —— ♻️ Repost to help others learn RAG. 🙏 Thanks to @OracleDevs for sponsoring this post. ➕ Follow me ( Nikki Siapno ) to improve at system design and AI.

𝗧𝗼𝗽 𝟭𝟲 𝗔𝗪𝗦 𝗦𝗲𝗿𝘃𝗶𝗰𝗲𝘀 𝗘𝗻𝗴𝗶𝗻𝗲𝗲𝗿𝘀 𝗦𝗵𝗼𝘂𝗹𝗱 𝗞𝗻𝗼𝘄. Understanding AWS isn't about memorizing service names. It's about knowing the role each one plays in your architecture. 𝗖𝗼𝗺𝗽𝘂𝘁𝗲 𝗿𝘂𝗻𝘀 𝗰𝗼𝗱𝗲: ↳ Lambda runs event-driven functions with automatic scaling. ↳ ECS orchestrates containers using AWS-native tooling. ↳ EKS provides Kubernetes with a managed control plane. 𝗗𝗮𝘁𝗮 𝘀𝘁𝗼𝗿𝗲𝘀 𝘀𝘁𝗮𝘁𝗲: ↳ RDS manages relational databases. ↳ DynamoDB provides serverless key-value and document storage. ↳ ElastiCache provides managed in-memory caching for low-latency access. 𝗜𝗻𝘁𝗲𝗴𝗿𝗮𝘁𝗶𝗼𝗻 𝗺𝗼𝘃𝗲𝘀 𝘄𝗼𝗿𝗸: ↳ SQS provides durable queues for asynchronous tasks. ↳ SNS handles pub/sub messaging and fanout. ↳ EventBridge routes events between applications and services. ↳ Step Functions orchestrates workflows with retries, waits, and branching. 𝗧𝗿𝗮𝗳𝗳𝗶𝗰 𝗰𝗼𝗻𝗻𝗲𝗰𝘁𝘀 𝘀𝗲𝗿𝘃𝗶𝗰𝗲𝘀: ↳ API Gateway manages APIs with authorization and throttling. ↳ CloudFront caches content closer to users. ↳ VPC provides isolated virtual networks for cloud resources. 𝗦𝗲𝗰𝘂𝗿𝗶𝘁𝘆 𝗮𝗻𝗱 𝗼𝗽𝗲𝗿𝗮𝘁𝗶𝗼𝗻𝘀 𝗸𝗲𝗲𝗽 𝘀𝘆𝘀𝘁𝗲𝗺𝘀 𝗰𝗼𝗻𝘁𝗿𝗼𝗹𝗹𝗲𝗱: ↳ IAM manages permissions through policies and roles. ↳ CloudWatch collects metrics, logs, and operational telemetry. ↳ Secrets Manager stores and helps rotate application credentials. Building scalable infrastructure is only part of the challenge. As software companies integrate AI into their products, they also need to connect models to existing data, manage AI workloads, and keep them secure and reliable in production. AWS supports this through services like Amazon Bedrock for building generative AI applications and Amazon Bedrock AgentCore for deploying and operating AI agents at scale. AWS's AI resources help software companies explore how to integrate AI into their products, move from prototypes to production, and scale AI-powered features across customers. 𝗘𝘅𝗽𝗹𝗼𝗿𝗲 𝗔𝗜 𝗳𝗼𝗿 𝘀𝗼𝗳𝘁𝘄𝗮𝗿𝗲 𝗰𝗼𝗺𝗽𝗮𝗻𝗶𝗲𝘀 → lucode.co/ai-for-software-co… What else would you add? —— ♻️ Repost to help others learn AWS. 🙏 Thanks to @awsdevelopers for sponsoring this post. ➕ Follow me ( Nikki Siapno ) to improve at AI and system design.

REST vs GraphQL vs gRPC vs WebSockets. 𝗥𝗘𝗦𝗧 exposes resources through HTTP endpoints, making it practical for public APIs and standard CRUD operations. With 𝗚𝗿𝗮𝗽𝗵𝗤𝗟, clients can request specific fields and related data, making it useful when different clients need different data shapes. 𝗴𝗥𝗣𝗖 uses strongly typed contracts and efficient serialization, making it well suited for backend service communication. For real-time communication, 𝗪𝗲𝗯𝗦𝗼𝗰𝗸𝗲𝘁𝘀 enable persistent, two-way connections, allowing applications to exchange data as events happen. But exchanging data in real time doesn't necessarily mean an application can respond naturally. Take AI voice agents used in customer support. They need to do more than stream audio. They also need to handle interruptions, understand conversational context, and adapt how they respond, while meeting enterprise security and deployment requirements. Deepgram Flux TTS addresses the speech generation side of this by using conversational context to adjust tone, pacing, and emotion in real time. For example, a support agent can respond empathetically to a frustrated customer, then switch to clearly reading out a tracking number. Combined with streaming and interruption handling, this helps developers build more natural voice experiences without scripting every interaction. 𝗛𝗲𝗮𝗿 𝗙𝗹𝘂𝘅 𝗧𝗧𝗦 𝗶𝗻 𝗮𝗰𝘁𝗶𝗼𝗻 → lucode.co/deepgram-flux-tts-… What else would you add? —— ♻️ Repost to help others learn and grow. 🙏 Thanks to @DeepgramAI for sponsoring this post. ➕ Follow me ( Nikki Siapno ) to improve at system design and AI.

Workflow vs AI Agent. A 𝘄𝗼𝗿𝗸𝗳𝗹𝗼𝘄 follows a path you define in advance. Step A → Step B → Step C. That works well when the process is predictable and the next step is already known. But an 𝗔𝗜 𝗮𝗴𝗲𝗻𝘁 works differently. You give it a goal, and it can decide which steps, tools, or actions to use based on the context it encounters. That makes agents useful when the path cannot be fully scripted upfront. But an agent is only as powerful as its execution layer. In enterprise environments, that means connecting it to the applications, data, and business processes where the work actually happens. Workday is starting to open up more of that layer to developers. 𝗙𝗼𝗿 𝘁𝗵𝗲 𝗳𝗶𝗿𝘀𝘁 𝘁𝗶𝗺𝗲, 𝗪𝗼𝗿𝗸𝗱𝗮𝘆 𝗵𝗮𝘀 𝗼𝗽𝗲𝗻𝗲𝗱 𝗶𝘁𝘀 𝗗𝗲𝘃𝗲𝗹𝗼𝗽𝗲𝗿 𝗣𝗹𝗮𝘁𝗳𝗼𝗿𝗺 𝘁𝗼 𝗶𝗻𝗱𝗲𝗽𝗲𝗻𝗱𝗲𝗻𝘁 𝗱𝗲𝘃𝗲𝗹𝗼𝗽𝗲𝗿𝘀. That means builders can now deploy AI-powered solutions around Workday applications, data, and business processes, with access to an ecosystem powering 75 million+ active users. For developers building skills around enterprise AI, this opens up an ecosystem that was previously out of reach. 𝗝𝗼𝗶𝗻 𝘁𝗵𝗲 𝗪𝗼𝗿𝗸𝗱𝗮𝘆 𝗗𝗲𝘃𝗲𝗹𝗼𝗽𝗲𝗿 𝗣𝗹𝗮𝘁𝗳𝗼𝗿𝗺 → lucode.co/workday-developer-… What else would you add? —— ♻️ Repost to help others learn AI agents. 🙏 Thanks to @Workday for sponsoring this post. ➕ Follow me ( Nikki Siapno ) to improve at AI and system design.

5 things slowing developers down in 2026 (and why faster coding doesn't solve them): Atlassian's latest State of DevEx report reveals the biggest sources of developer friction. 𝟱. 𝗨𝗻𝗰𝗹𝗲𝗮𝗿 𝗿𝗲𝗾𝘂𝗶𝗿𝗲𝗺𝗲𝗻𝘁𝘀 (𝟰𝟬%) ↳ Clarifying what actually needs to be built. 𝟰. 𝗧𝗲𝗰𝗵𝗻𝗶𝗰𝗮𝗹 𝗱𝗲𝗯𝘁 (𝟰𝟰%) ↳ Working around existing code and architectural limitations. 𝟯. 𝗖𝗼𝗻𝘁𝗲𝘅𝘁 𝘀𝘄𝗶𝘁𝗰𝗵𝗶𝗻𝗴 (𝟰𝟱%) ↳ Moving between tasks, tools, and priorities. 𝟮. 𝗠𝗲𝗲𝘁𝗶𝗻𝗴𝘀 (𝟱𝟬%) ↳ Time spent coordinating instead of executing. 𝟭. 𝗙𝗶𝗻𝗱𝗶𝗻𝗴 𝗶𝗻𝗳𝗼𝗿𝗺𝗮𝘁𝗶𝗼𝗻 (𝟱𝟭%) ↳ Searching across fragmented documentation and tools. But why do these bottlenecks persist even as AI makes coding faster? The video explains what's happening across the SDLC. For a deeper look at what engineering teams can do about it, join Atlassian's expert panel discussing the findings. 𝗥𝗲𝗮𝗱 𝘁𝗵𝗲 𝗳𝘂𝗹𝗹 𝗿𝗲𝗽𝗼𝗿𝘁 → lucode.co/2026-state-of-deve… 𝗪𝗮𝘁𝗰𝗵 𝘁𝗵𝗲 𝘄𝗲𝗯𝗶𝗻𝗮𝗿 → lucode.co/2026-state-of-deve… What else would you add? Thanks to @Atlassian for sponsoring this post.

Git branching strategies: Do you know the differences? 𝗙𝗲𝗮𝘁𝘂𝗿𝗲 𝗯𝗿𝗮𝗻𝗰𝗵𝗶𝗻𝗴 keeps each feature in its own branch, isolated from the main branch, making pull requests easier to review. Once complete, the feature is merged back into main. 𝗚𝗶𝘁𝗳𝗹𝗼𝘄 uses two long-lived branches: dev for development and main for production. Features are built in separate branches, then merged into dev. Releases are prepared in dedicated branches before merging into production. 𝗚𝗶𝘁𝗟𝗮𝗯 𝗳𝗹𝗼𝘄 combines feature branching with environment-based deployment workflows. Changes move through environments like staging before reaching production, making it well-suited for CI/CD and staged releases. 𝗚𝗶𝘁𝗛𝘂𝗯 𝗳𝗹𝗼𝘄 simplifies things. The main branch is always deployable. Developers create short-lived branches, open pull requests, and merge once approved, often triggering deployment. 𝗧𝗿𝘂𝗻𝗸-𝗯𝗮𝘀𝗲𝗱 𝗱𝗲𝘃𝗲𝗹𝗼𝗽𝗺𝗲𝗻𝘁 minimizes branching. Changes are merged into main frequently, supported by strong automated testing, CI pipelines, and feature flags to maintain stability. There's no one-size-fits-all branching strategy. The right approach depends on your team's size, release process, and how frequently you deploy. What else would you add? —— Thanks to our partner Kickresume who keeps our content free. 𝗔𝗿𝗲 𝘆𝗼𝘂 𝘀𝘂𝗿𝗲 𝘆𝗼𝘂'𝗿𝗲 𝗳𝗼𝗹𝗹𝗼𝘄𝗶𝗻𝗴 𝘂𝗽 𝗼𝗻 𝗲𝘃𝗲𝗿𝘆 𝗷𝗼𝗯 𝘆𝗼𝘂 𝗮𝗽𝗽𝗹𝘆 𝘁𝗼? Keep track of every application with Kickresume's free Job Application Tracker. Track your applications here → lucode.co/jobs-tracker-kickr… —— ♻️ Repost to help others Git. ➕ Follow me ( @NikkiSiapno ) to improve at system design.

𝗔𝗴𝗲𝗻𝘁𝗶𝗰 𝗦𝗗𝗟𝗖 𝘃𝘀 𝗠𝗼𝗱𝗲𝗿𝗻 𝗦𝗗𝗟𝗖 𝘃𝘀 𝗧𝗿𝗮𝗱𝗶𝘁𝗶𝗼𝗻𝗮𝗹 𝗦𝗗𝗟𝗖. With 𝘁𝗿𝗮𝗱𝗶𝘁𝗶𝗼𝗻𝗮𝗹 𝗦𝗗𝗟𝗖 (𝗪𝗮𝘁𝗲𝗿𝗳𝗮𝗹𝗹), work moves through sequential phases, with planning done upfront and testing before release. This provides structure, but makes adapting to changing requirements difficult. Whereas with 𝗺𝗼𝗱𝗲𝗿𝗻 𝗦𝗗𝗟𝗖 (𝗔𝗴𝗶𝗹𝗲 + 𝗗𝗲𝘃𝗢𝗽𝘀), teams work in shorter iterations, continuously deliver updates, and respond to feedback faster. It's more flexible, but humans still handle most of the execution. Now, with 𝗮𝗴𝗲𝗻𝘁𝗶𝗰 𝗦𝗗𝗟𝗖 (𝗔𝗜 𝗔𝗴𝗲𝗻𝘁𝘀), agents can take on planning, coding, testing, and debugging, while humans shift toward setting goals, providing context, and reviewing outcomes. The key difference is how work gets done, and how much of it can be automated. But faster execution doesn't automatically mean faster delivery. According to Atlassian's 2026 State of DevEx report, 𝟴𝟳% of developers say AI helps them ship faster, yet 𝗼𝗻𝗹𝘆 𝟮𝟴% report less organizational friction. Developers still spend time finding information, context switching, and reviewing AI-generated work. This is why an AI-native SDLC requires more than AI coding tools. Teams need to rethink how requirements, context, collaboration, and validation work together. Atlassian's research explores this gap between individual productivity and team-wide efficiency, and what it takes to make AI useful across the entire development lifecycle. 𝗥𝗲𝗮𝗱 𝘁𝗵𝗲 𝗳𝘂𝗹𝗹 𝗿𝗲𝗽𝗼𝗿𝘁 → lucode.co/state-of-devex-rep… What else would you add? —— ♻️ Repost to help others learn SDLC. 🙏 Thanks to @Atlassian for sponsoring this post. ➕ Follow me ( Nikki Siapno ) to improve at AI and system design.

Things Every Developer Should Know: CI/CD Pipelines. CI/CD pipelines automate the path from code change to production. A typical pipeline connects: → Source control → Automated testing → Artifact creation → Security scanning → Deployment The goal is simple: catch problems earlier and make releases repeatable. But automating deployment creates another problem: the pipeline still needs permission to access your infrastructure. It may need to push a container image, deploy to Kubernetes, access a database, or modify cloud resources. One option is to give the pipeline a long-lived credential. But that means another secret to store, rotate, and protect. Workload identity takes a different approach. Instead of storing a permanent secret, the workload proves its identity and receives short-lived access when it needs it. Teleport’s Machine & Workload Identity applies this model to CI/CD jobs, AI agents, and other automated workloads accessing infrastructure. CI/CD controls how code moves through the deployment process. Workload identity controls what that automation is allowed to access. 𝗟𝗲𝗮𝗿𝗻 𝗺𝗼𝗿𝗲 𝗵𝗲𝗿𝗲 → lucode.co/machine-workload-i… What else would you add? —— ♻️ Repost to help others learn system design. 🙏 Thanks to @goteleport for sponsoring this post. ➕ Follow me ( Nikki Siapno ) to improve at system design.

𝗛𝗧𝗧𝗣, 𝗛𝗧𝗧𝗣 𝟮, 𝗛𝗧𝗧𝗣 𝟯: 𝗗𝗼 𝘆𝗼𝘂 𝗸𝗻𝗼𝘄 𝘁𝗵𝗲 𝗱𝗶𝗳𝗳𝗲𝗿𝗲𝗻𝗰𝗲𝘀? HTTP now powers complex modern applications, but it started as a simple protocol for retrieving documents. 𝗛𝗧𝗧𝗣/𝟬.𝟵 only supported a single GET request with no headers or status codes; just raw HTML from the server. It was simple and fast, but lacked structure and scalability. 𝗛𝗧𝗧𝗣/𝟭.𝟬 introduced headers, status codes, and content types, making the web more structured and extensible, but required a new TCP connection for every request, which didn’t scale well. 𝗛𝗧𝗧𝗣/𝟭.𝟭 added persistent connections, caching, and chunked transfer, making web interactions more efficient, but still suffered from head-of-line blocking and connection overhead at scale. 𝗛𝗧𝗧𝗣𝗦 is HTTP layered over TLS, providing encryption, integrity, and authentication, securing communication over the internet, but adds handshake overhead and operational complexity. 𝗛𝗧𝗧𝗣/𝟮 introduced multiplexing, binary framing, and header compression, allowing multiple requests over a single connection, improving performance, but still relies on TCP and suffers from transport-level blocking. 𝗛𝗧𝗧𝗣/𝟯 runs over QUIC (UDP) with built-in TLS 1.3, enabling independent streams, faster handshakes, and reduced latency, but introduces more complexity in infrastructure and network handling. From per-request connections → to multiplexing over TCP → to QUIC and independent streams. That’s the evolution of HTTP. 📌 If you want to learn more, dive deeper here: lucode.co/evolution-of-http-… What else would you add?