--- title: "Latest reads" description: "The engineering internet, summarised so you can actually read it." --- # Latest reads > The engineering internet, summarised so you can actually read it. ## Articles ### [How we do product design at Intercom – Inside Intercom](https://yomu.fyi/post/how-we-do-product-design-at-intercom-inside-intercom.md) - Company: [Intercom](https://yomu.fyi/company/intercom.md) - Published: Oct 13, 2026 - Type: Explainer In a 2014 email to Intercom’s design team, Paul Adams shares a deliberate ten-step process focusing on deep problem understanding. Designers must first read all research and customer feedback, consult the Customer Success team for clarification, and use product instrumentation to avoid assumptions about user behavior. They are instructed to generate at least three or four different solutions, research how other products solve similar problems, and iterate through feedback from multiple designers. Work proceeds systematically through four design layers: outcome, structure, interaction, and visual, with the principle of not mixing levels and always optimizing for simplicity. After shipping, designers must proactively measure success against defined metrics, talk to the Customer Success team, and learn from mistakes, as truly great designers are dogged in understanding problems even after a solution ships. ### [7 things I wish every search box did - The Intercom Blog](https://yomu.fyi/post/7-things-i-wish-every-search-box-did-the-intercom-blog.md) - Company: [Intercom](https://yomu.fyi/company/intercom.md) - Published: Oct 13, 2026 - Type: Explainer Search in web applications often fails to deliver speed and relevance, yet it is critical for findability as users cannot browse effectively through large amounts of content. The post, by Algolia’s CEO, presents seven principles: show real results from the first keystroke, search across multiple attributes like name and company without field specification, rank based on object importance and context such as user activity or location, highlight matched terms to explain results, offer faceted filters instead of complex syntax, accept typos and synonyms while ranking approximate matches lower, and ensure consistent mobile and desktop experiences. Ranking should consider dimensions like recent activity or geo-location rather than text frequency, making users feel the search reads their minds. Results must explain why they matched to help users quickly identify the best result. These principles help users find results with minimal effort and make search feel intuitive, addressing the iceberg of search complexity. ### [Interview with Bob Moesta on Jobs-to-be-Done (Part 2) – Inside Intercom](https://yomu.fyi/post/interview-with-bob-moesta-on-jobs-to-be-done-part-2-inside-intercom.md) - Company: [Intercom](https://yomu.fyi/company/intercom.md) - Published: Oct 13, 2026 - Type: Explainer Bob Moesta explains how startups can apply Jobs-to-be-Done theory by identifying struggling moments and understanding the switching mechanism, including why people switch, the cost to switch, and how they choose alternatives. He advises focusing on people who have already switched rather than hypothetical buyers, because actual purchases embed a value code that reveals trade-offs and what they are willing to give up. Moesta argues that beta tests must include a payment component to reveal what customers truly value, otherwise feedback is unreliable and leads to over-engineering features no one will pay for. He illustrates with a home building example where reframing from selling homes to helping people move increased sales, and including a moving service or furniture for a divorced mom sold more homes. The interview also covers competing against non-consumption, as seen with Siri enabling texting in situations where typing was avoided, increasing texting by 300 percent. ### [Show, don't tell, with GIFs - The Intercom Blog](https://yomu.fyi/post/show-don-x27-t-tell-with-gifs-the-intercom-blog.md) - Company: [Intercom](https://yomu.fyi/company/intercom.md) - Published: Oct 13, 2026 - Type: Explainer Often dismissed as outdated, the GIF format has found a crucial new role in customer support for showing users how to perform tasks, replacing inefficient convoluted text instructions. Intercom uses GIFs to demonstrate actions, following the “Show, don’t tell” principle, which removes the need for step enumeration and annotations. They are particularly effective for reporting interface errors such as broken transitions or animations, removing the need for arrows, lines, and wordy descriptions. Recording a desktop area into a GIF is easy with free tools like Licecap for macOS, or GifGrabber and GifBrewery as alternatives. Customers respond enthusiastically to these GIFs, often asking how they were made, and they can be used across emails, in-app messages, and documentation. ### [Interview with Bob Moesta on Jobs-to-be-Done – Inside Intercom](https://yomu.fyi/post/interview-with-bob-moesta-on-jobs-to-be-done-inside-intercom.md) - Company: [Intercom](https://yomu.fyi/company/intercom.md) - Published: Oct 13, 2026 - Type: Explainer Bob Moesta, who works with Clay Christensen to develop an online course on Jobs-to-be-Done (JTBD), applies the framework through his company The ReWired Group, which helps Fortune 500 companies find innovation whitespace and incubates technology startups. JTBD is gaining popularity because Christensen promotes it, markets are saturated, and traditional marketing methods target large markets rather than seeding niches from existing consumer struggles. The method starts with the consumer’s situational context and motivation, stopping abstraction when the consumer cannot think about the problem, as in the milkshake example of solving commute boredom. Products often perform multiple jobs; the decision to split products depends on conflicts in how consumers measure value, not just benefits. Branding can be structured with endorser, driver, descriptor, and ingredient brands to clearly communicate the job a product solves, as with IBM Thinkpad and Intel Inside. ### [Is machine learning hard? Not always - The Intercom Blog](https://yomu.fyi/post/is-machine-learning-hard-not-always-the-intercom-blog.md) - Company: [Intercom](https://yomu.fyi/company/intercom.md) - Published: Oct 13, 2026 - Type: Explainer The post argues that machine learning does not always require advanced mathematics and, as described by Intercom’s Fergal Reid, is about systems that improve with more training data. It contrasts a naive keyword-matching approach in Ruby for a ‘You might also like’ feature, which produced irrelevant results, with a basic machine learning method using K-Means clustering. The machine learning approach first represents each blog post as a binary vector indicating the presence of each word from the entire corpus. The K-Means algorithm then iteratively assigns each post to the nearest cluster center and recalculates centers as the average of assigned posts until assignments stabilize. The post demonstrates that roughly 40 lines of Ruby code can produce relevant recommendations and encourages readers to try libraries like Scikit Learn. ### [How to Educate & Persuade Customers - The Intercom Blog](https://yomu.fyi/post/how-to-educate-amp-persuade-customers-the-intercom-blog.md) - Company: [Intercom](https://yomu.fyi/company/intercom.md) - Published: Oct 13, 2026 - Type: Explainer The post presents a framework for product messaging that categorizes user tasks by ease of execution and user desire, using a quadrant to determine whether to educate, persuade, or nudge. Tasks that are desirable but daunting require education to simplify the concept, while easy but low-value tasks need persuasion to highlight benefits, and tasks in between may only need a nudge. The author warns against simplifying tasks users do not want, citing failed startups, and references Ev Williams’ formula of removing steps from long-standing human desires. The post also discusses the cost-benefit trade-off in user decisions, distinguishing honest growth hacks that alter the trade-off from dishonest ones that deceive users or move goalposts. It advises against building a business around complex unwanted tasks and notes that Intercom itself uses this approach by teaching customers about drip-feed marketing. ### [How to launch with a validated idea - The Intercom Blog](https://yomu.fyi/post/how-to-launch-with-a-validated-idea-the-intercom-blog.md) - Company: [Intercom](https://yomu.fyi/company/intercom.md) - Published: Oct 13, 2026 - Type: Tutorial The post describes a three-step process for launching a validated idea, called the three I’s—Ideas, Input, and Iteration—mirroring the lean startup approach. The author explains how his team validated their idea by drafting mockups, creating a sign-up page, and running AdWords, and they got lucky when blogs shared their landing page, giving them a base of potential users for product preview sessions. They then launched a minimum viable product (MVP) and used Intercom to send surveys and automated in-app messages, creating a continuous feedback loop. To iterate, they formed a CoPilots program of users highly likely to purchase who provided weekly feedback via Google+ Hangouts in exchange for free use. The entire team was involved in reviewing interviews, and the cycle is presented as an ongoing process for SaaS startups. ### [How to hire designers - The Intercom Blog](https://yomu.fyi/post/how-to-hire-designers-the-intercom-blog.md) - Company: [Intercom](https://yomu.fyi/company/intercom.md) - Published: Oct 13, 2026 - Type: Explainer The post argues that design is fundamentally about how something works, not how it looks, and that hiring based on visual portfolios like Dribbble perpetuates a misunderstanding. The author presents a four-layer design framework: intended outcome, component design, detailed interactions, and visual layer. He asserts that these layers are interdependent and that the best designers excel at one or two while being very good at the others. He advises evaluating candidates against the specific layers needed for the role, noting that a popular Dribbble portfolio only proves graphics software skill and trend awareness, not design ability. The goal is to break the cycle of misunderstanding design education and ensure companies hire designers who can solve real problems. ### [How to send good email – opens, clicks, conversions - The Intercom Blog](https://yomu.fyi/post/how-to-send-good-email-opens-clicks-conversions-the-intercom-blog.md) - Company: [Intercom](https://yomu.fyi/company/intercom.md) - Published: Oct 13, 2026 - Type: Explainer Email open rate, which relies on a tracking pixel being loaded, is an unreliable metric because modern clients let users read without opening. Reader fatigue is inevitable as a list grows, and the only cure is staying relevant. Subject lines and sender fields have the strongest effect on opens; simple, personal lines like “Dinner?” performed well for a political campaign. Click-through rates are more accurate, and the first two links get most of the clicks, so the call to action must be prominent. Links should be visually obvious and placed predictably, as weak styling halves their click-through rate. The core advice is to write about what the reader wants to know, not about the product, and to test the email on a stranger before sending. ### [Make every call count: The evolution of phone support - The Intercom Blog](https://yomu.fyi/post/make-every-call-count-the-evolution-of-phone-support-the-intercom-blog.md) - Company: [Intercom](https://yomu.fyi/company/intercom.md) - Published: Oct 13, 2026 - Type: Announcement The podcast transcript traces the evolution of phone support from Alexander Graham Bell to modern call centers and the science of hold music. It argues phone support has not kept pace with other technologies, remaining expensive and unscalable despite being essential for complex issues. A customer’s three-hour hold time that ended in a dropped call illustrates the problem with traditional phone support. Intercom Phone is introduced as a native solution that integrates phone support into the Intercom platform, allowing agents to handle calls alongside other channels without switching tools or sharing links. Key features include customizable IVR trees and the ability to escalate a live chat to a voice or video call with screen sharing directly in the Messenger, with seamless switching back to messaging mid-call, and the post concludes that phone support must evolve to meet customer expectations. ### [Stopping business email compromise with AI-driven email security: How VIPRE protects modern enterprises on AWS](https://yomu.fyi/post/stopping-business-email-compromise-with-ai-driven-email-security-how-v.md) - Company: [AWS](https://yomu.fyi/company/aws.md) - Author: Pedram Jahangiri - Published: Oct 13, 2026 - Type: Problem & solution Business email compromise (BEC) uses social engineering and impersonation rather than malicious links, bypassing traditional secure email gateways that are blind to internal traffic and lateral movement. VIPRE developed Integrated Email Security (IES), a cloud-native solution on AWS that integrates directly with Microsoft 365 via APIs, eliminating the need for MX record changes. The solution uses multiple machine learning layers, including semantic similarity models, contextual language models, and dynamic behavioral analysis of links and attachments, running on AWS services such as Amazon ECS with NVIDIA GPUs. In production, VIPRE IES processes over half a million messages daily with P95 latency between 1.4 and 78 seconds depending on sandbox analysis, detecting approximately 1% of messages as BEC attempts. Deployment can be completed in minutes rather than days, providing immediate visibility into internal risks that traditional gateways miss. ### [Bulk Executor for Amazon DynamoDB, Part 2: Lessons we learned the hard way](https://yomu.fyi/post/bulk-executor-for-amazon-dynamodb-part-2-lessons-we-learned-the-hard-w.md) - Company: [AWS](https://yomu.fyi/company/aws.md) - Author: Jason Hunter - Published: Oct 13, 2026 - Type: Problem & solution The post describes design tradeoffs and workarounds encountered while building the open-source Bulk Executor tool for Amazon DynamoDB on AWS Glue. Core requirements included real-time local-script feel, scaling to hundreds of parallel workers without user infrastructure management, and easy command extensibility. AWS Glue was chosen for serverless parallelism and a built-in DynamoDB connector, but the connector simplified data types, losing distinctions like string set versus list of strings. Challenges included estimating costs for provisioned tables, handling job termination with ^C, and streaming output via CloudWatch Live Tail which samples events and delivers out-of-order. Solutions included a two-phase cost estimate for deletes, programmatic interrupt handling, writing large output to S3 with a preview, mandating point-in-time recovery for mutating commands, and using AI for testing and linting; the team also migrated from AWS Glue 4 to 5 and 6 for faster startup and 30 percent lower DPU costs. ### [NVIDIA Run:ai Is Now Open Source as Part of NVIDIA DSX AI Factory Platform](https://yomu.fyi/post/nvidia-run-ai-is-now-open-source-as-part-of-nvidia-dsx-ai-factory-plat.md) - Company: [NVIDIA](https://yomu.fyi/company/nvidia-developer-blog.md) - Author: Tanya Lenz - Published: Oct 13, 2026 - Type: Announcement NVIDIA has released the complete Run:ai platform as open source software, expanding the NVIDIA DSX AI Factory Platform and giving customers, developers, and ecosystem partners production-grade AI workload and GPU orchestration capabilities. Run:ai has dynamically allocated GPU resources and managed workloads on Kubernetes across data processing, training, and inference. The release includes the full platform plus four independently adoptable components: KAI Scheduler, GPU Fractioning Operator, KAI Resource Management, and Workload Map. Teams can deploy the stack from the NVIDIA Run:ai repository in the DSX AI Factory GitHub organization or via an NGC Community Product Helm chart, without a paid subscription or enterprise support entitlement. Ecosystem partners including Crusoe, Mirantis, Nebius, OpenNebula Systems, Rafay, Saturn Cloud, Spectro Cloud, and vCluster are integrating the platform or its components into their offerings. ### [Test and deploy bootable containers with bootc virtualization kit (bcvk)](https://yomu.fyi/post/test-and-deploy-bootable-containers-with-bootc-virtualization-kit-bcvk.md) - Company: [Red Hat](https://yomu.fyi/company/red-hat.md) - Author: Andre Luiz Facina - Published: Oct 13, 2026 - Type: Tutorial Managing a traditional operating system across multiple hosts often leads to configuration drift, slow updates, and painful rollbacks. Bootable containers solve this by packaging the entire OS as an OCI container image, including the kernel, initrd, bootloader, and applications. Red Hat Enterprise Linux 10 introduced image mode deployment, and the bootc container virtualization kit (bcvk) enables end-to-end testing of these images on a laptop using familiar Podman and libvirt packages. The tool offers three deployment modes: ephemeral for rapid testing, persistent for long-term development, and disk image for creating bootable artifacts. Users build a custom bootc image via a Containerfile with the containers.bootc=1 label, then deploy it identically across virtual and physical infrastructure, reducing environmental drift. ### [Red Hat AI Inference 3.5 MaaS governance on AKS](https://yomu.fyi/post/red-hat-ai-inference-3-5-maas-governance-on-aks.md) - Company: [Red Hat](https://yomu.fyi/company/red-hat.md) - Author: clobner - Published: Oct 13, 2026 - Type: Tutorial Red Hat AI Inference 3.5 extends its Model-as-a-Service (MaaS) governance layer to non-OpenShift Kubernetes distributions, including Azure Kubernetes Service (AKS). MaaS adds multi-tenant API governance on top of the inference platform, controlling authentication, rate limiting, and model access via Authorino and Limitador from Red Hat Connectivity Link (Kuadrant). A model is first deployed with a single LLMInferenceService YAML, creating a vLLM server and router-scheduler, but the endpoint is open. Applying a MaaSSubscription and MaaSAuthPolicy activates governance: Authorino validates API keys or JWTs, and Limitador enforces token rate limits. The post demonstrates three enforcement scenarios on an AKS cluster with a T4 GPU running Qwen3-0.6B: 401 for missing or invalid keys, 200 for authenticated requests, and 429 when rate limits are exceeded, blocking requests before they reach vLLM. ### [Sharding etcd on hosted control planes and what the numbers actually say](https://yomu.fyi/post/sharding-etcd-on-hosted-control-planes-and-what-the-numbers-actually-s.md) - Company: [Red Hat](https://yomu.fyi/company/red-hat.md) - Author: Vishnu Challa - Published: Oct 13, 2026 - Type: Benchmark Sharding etcd on hosted control planes splits the key-value store into multiple instances by resource kind, a tech preview feature tested with a main shard for cluster state and a dedicated events shard on Red Hat OpenShift 4.21 with HyperShift. Under event pressure from synthetic storms and realistic crashlooping pods, the sharded configuration showed up to 96% fewer slow applies, faster compaction and raft snapshot durations, and 60% more WAL fsyncs and 90% more backend commits combined. The sharded arm never failed the event storm workload, while the standard arm occasionally did. Pod-ready latencies and workload finish times remained comparable across both configurations. The trade-offs are roughly 1.8 to 2.3 times the etcd container memory and a compaction regression under large custom resource quota pressure that is still under investigation. ### [CI/CD Pipeline Observability Tools: How to Choose the Right Setup](https://yomu.fyi/post/ci-cd-pipeline-observability-tools-how-to-choose-the-right-setup.md) - Company: [Jetbrains](https://yomu.fyi/company/jetbrains.md) - Author: Olga Bedrina - Published: Oct 13, 2026 - Type: Explainer CI/CD pipeline observability tools help teams understand why builds fail or slow down by connecting data from version control, build agents, tests, and deployment tools. The JetBrains Developer Ecosystem Survey 2026 found that 44% of 20,416 respondents cited debugging cloud infrastructure as a top challenge, while 16% named observability. Pipeline observability covers queue time, duration, success rates, flaky tests, cache performance, and platform-level metrics like agent utilization and server health. Built-in CI/CD dashboards are used by 76% of respondents, but many teams also use Prometheus, Grafana, Elastic, or OpenTelemetry for metrics, logs, and traces. The post compares built-in dashboards, Prometheus and Grafana, Elastic and OpenSearch, OpenTelemetry, commercial platforms like Datadog, and custom in-house solutions, recommending teams start with the questions their current setup cannot answer. ### [Python vs Go: A Practical Comparison for Developers](https://yomu.fyi/post/python-vs-go-a-practical-comparison-for-developers.md) - Company: [Jetbrains](https://yomu.fyi/company/jetbrains.md) - Author: Dominika Stankiewicz - Published: Oct 13, 2026 - Type: Explainer The post compares Python and Go across language design, performance, tooling, and developer profiles, drawing on quantitative data from the JetBrains Developer Ecosystem 2026 survey of 28,945 respondents and qualitative interviews with developers who migrated production workloads from Python to Go. Python favors expressive code and rapid iteration, while Go emphasizes explicit, standardized code that is easy to maintain across teams. Go excels in raw performance and concurrency and is widely used in cloud infrastructure, whereas Python remains stronger for machine learning, data science, and scientific computing. Survey data shows that Go developers are typically more infrastructure-focused and senior, and that Go is rarely a first language, while Python is beginner-friendly with 10% of its developers coding for less than a year. Most teams use both languages, with Go for services and infrastructure and Python for AI and data workloads, and moving to Go rarely means abandoning Python. ### [Chasing that 32 ms tail latency in ClickHouse](https://yomu.fyi/post/chasing-that-32-ms-tail-latency-in-clickhouse.md) - Company: [Clickhouse](https://yomu.fyi/company/clickhouse.md) - Published: Oct 13, 2026 - Type: Problem & solution A customer observed an extra ~32 ms tail latency on simple SELECT queries after idle periods in ClickHouse’s Native TCP protocol. The root cause was that after a TCP idle period (RTO), the congestion window resets to initcwnd (8 MSS), and with TCP\_NODELAY enabled, each metadata packet (ProfileEvents, Logs, Progress) becomes a separate TCP segment, exceeding the small window and forcing an additional round trip. The ClickHouse Native protocol sends these metadata packets between Data and EndOfStream, and the unbatched writes caused the problem. The team reproduced the issue deterministically using Linux network namespaces and tc to simulate 32 ms RTT and initcwnd 8. They fixed it by batching those metadata packets into a single buffer write at the application layer, reducing tail latency from ~32 ms to ~1 ms, as confirmed by a Go client program that measured the time between the last Data packet and EndOfStream. [Newer posts](https://yomu.fyi/posts/page/6.md) · [Older posts](https://yomu.fyi/posts/page/8.md)