---
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

### [Two years of vector search at Notion: 10x scale, 1/10th cost](https://yomu.fyi/post/two-years-of-vector-search-at-notion-10x-scale-1-10th-cost.md)
- Company: [Notion](https://yomu.fyi/company/notion.md)
- Author: Preeti Gondi, Mickey Liu, Nathan Louie, Calder Lund, Jacob Sager
- Published: Feb 19, 2026
- Type: Explainer

Notion describes two years of scaling vector search for AI Q&A, where semantic embeddings retrieve workspace and connected-tool content by meaning rather than exact keywords. The initial system combined Apache Spark batch indexing with Kafka-driven real-time updates, workspace-based sharding, and generation routing; rapid demand drove a 600x increase in daily onboarding capacity and cleared a waitlist of millions of workspaces by April 2024. A serverless migration cut vector database costs 50 percent from peak usage, while a later move to turbopuffer reduced search-engine spend 60 percent, lowered AWS EMR costs 35 percent, and improved p50 query latency from 70–100ms to 50–70ms. Further work uses xxHash-based page-state caching in DynamoDB to reprocess only changed spans and shifts embeddings generation from Spark and query embeddings away from a third-party API to Ray on Anyscale, with early rollout results pointing to a projected 90+ percent reduction in embeddings infrastructure costs.


### [Building a mental model for Stripe payments](https://yomu.fyi/post/building-a-mental-model-for-stripe-payments.md)
- Company: [Stripe](https://yomu.fyi/company/stripe.md)
- Author: Allison Farris
- Published: Feb 19, 2026
- Type: Tutorial

An online bookshop checkout must coordinate card networks, issuing banks, fraud checks, regulatory requirements, currencies, and varied payment methods while completing in roughly 2–3 seconds. Stripe abstracts these concerns through a unified API, with the PaymentIntent representing the payment lifecycle from amount and currency collection through tokenization, authorization, capture, and settlement. The Payment Element and client-side tokenization keep sensitive card details away from the shop’s servers, while Stripe may perform risk checks and additional 3D Secure authentication. Because payments can change state minutes, hours, or days later through refunds and disputes, webhook events should be treated as the source of truth for synchronizing order and downstream systems without polling.


### [Minions: Stripe’s one-shot, end-to-end coding agents—Part 2](https://yomu.fyi/post/minions-stripe-s-one-shot-end-to-end-coding-agents-part-2.md)
- Company: [Stripe](https://yomu.fyi/company/stripe.md)
- Author: Alistair Gray
- Published: Feb 19, 2026
- Type: Explainer

Stripe’s minions are an unattended agentic coding flow whose pull requests are human-reviewed but contain no human-written code; more than 1,300 merge weekly. The system runs agents in AWS EC2-based devboxes that are prewarmed to become usable within about 10 seconds, providing parallel, predictable, and isolated environments. Its custom harness uses code-defined blueprints that combine deterministic nodes, such as linting and pushing changes, with agent loops for tasks such as implementation and fixing CI failures. Minions receive automated feedback through local linting, one full CI iteration, and a possible second repair attempt, while devbox isolation and security controls limit destructive actions; Stripe presents the approach as an extension of long-running investments in human developer productivity.


### [First Block with Gabe Pereyra, Co-founder and President of Harvey](https://yomu.fyi/post/first-block-with-gabe-pereyra-co-founder-and-president-of-harvey.md)
- Company: [Notion](https://yomu.fyi/company/notion.md)
- Author: Anastasia Crew
- Published: Feb 18, 2026
- Type: Explainer

The episode presents Harvey’s founding and growth through Gabe Pereyra’s account of building an AI platform for large law firms and their clients. Two roommates—Winston, a securities litigator, and Gabe, an AI researcher—identified an opportunity in legal. Harvey chose a general-purpose platform over a narrow workflow such as NDA review, aiming for a comprehensive AI associate capable of handling varied work. The discussion also covers enterprise accuracy, infrastructure over models, and the difficulty of turning early seven-figure deals into a repeatable sales machine, with investors helping on that learning curve. Harvey’s trajectory includes a seed check from OpenAI and a valuation reported at $8 billion, while the episode frames domain expertise and scalable go-to-market execution as central founder lessons.


### [IBM and UC Berkeley Diagnose Why Enterprise Agents Fail Using IT-Bench and MAST](https://yomu.fyi/post/ibm-and-uc-berkeley-diagnose-why-enterprise-agents-fail-using-it-bench.md)
- Company: [Hugging Face](https://yomu.fyi/company/hugging-face.md)
- Author: Ayhan Sebin, Rohan Arora, Saurabh Jha
- Published: Feb 18, 2026
- Type: Benchmark

IBM Research and UC Berkeley analyze how agentic LLM systems fail in enterprise IT automation, including incident triage, logs and metrics queries, and Kubernetes actions across long-horizon tool loops. To move beyond benchmark success rates, the analysis applies MAST (Multi-Agent System Failure Taxonomy) to execution traces, converting failures into structured signatures across system design, inter-agent misalignment, and task verification. It covers 310 ITBench SRE traces from Gemini-3-Flash, Kimi-K2, and GPT-OSS-120B, whose mean recalls were 75.5%, 28.6%, and 12.4%, respectively. Gemini averaged 2.6 failure modes per failed trace, compared with 4.7 for Kimi and 5.3 for GPT-OSS-120B, while FM-3.3, Incorrect Verification, was identified as the strongest predictor across models; the analysis recommends external verification, explicit termination controls, and clarification branches.


### [Implementing a secure sandbox for local agents](https://yomu.fyi/post/implementing-a-secure-sandbox-for-local-agents.md)
- Company: [Cursor](https://yomu.fyi/company/cursor.md)
- Author: Ani Betts, Yash Gaitonde & Alex Haugland
- Published: Feb 18, 2026
- Type: Problem & solution

Auto-approved terminal commands make coding agents more capable but increase the risk of destructive changes and secret leakage, while repeated approval prompts create approval fatigue. Over three months, the team rolled out a uniform sandbox API across macOS, Linux, and Windows, allowing agents to run freely inside controlled environments and request approval mainly for internet access. macOS uses dynamically generated Seatbelt profiles; Linux combines Landlock, seccomp, and overlay filesystems to enforce workspace and ignored-file restrictions; Windows runs the Linux sandbox through WSL2. Changes to tool descriptions and result rendering help agents recognize constraints and escalate instead of retrying failed commands. Production data reports that sandboxed agents stop 40% less often, and the sandbox now handles a third of requests on supported platforms, although native Windows isolation remains under development.


### [One-Shot Any Web App with Gradio's gr.HTML](https://yomu.fyi/post/one-shot-any-web-app-with-gradio-s-gr-html.md)
- Company: [Hugging Face](https://yomu.fyi/company/hugging-face.md)
- Author: yuvraj sharma, hysts, Freddy Boulton
- Published: Feb 18, 2026
- Type: Tutorial

The post presents Gradio 6’s gr.HTML as a way to build interactive web components in a single Python file, combining custom HTML templates, scoped CSS, and JavaScript interactivity without a separate build step. It describes applications including a Pomodoro timer, contribution heatmap, Kanban board, spinning wheel, and machine-learning viewers for detection, 3D camera control, and speech transcription. Within the component API, ${value} injects Python state into the template, props.value updates that state from JavaScript, and trigger('change') synchronizes it back to Python; CSS keyframes and native HTML5 drag events supply richer behavior. For reuse, a gr.HTML subclass packages the templates and script so the result can be used like built-in Gradio components, while reload mode and Spaces deployment support rapid iteration from generated code to a live app.


### [Building a Movie Recommendation Engine with Hugging Face and Voyage AI](https://yomu.fyi/post/building-a-movie-recommendation-engine-with-hugging-face-and-voyage-ai.md)
- Company: [MongoDB](https://yomu.fyi/company/mongodb.md)
- Author: Arek Borucki
- Published: Feb 17, 2026
- Type: Tutorial

Traditional movie search mechanisms depend on coarse filters such as genre, actor, or title, which fail to capture emotional context and narrative nuances. This guide demonstrates building a mood-based semantic recommendation system by combining the voyage-4-nano open-weights embedding model, Hugging Face datasets, and MongoDB Atlas Vector Search. The architecture uses Sentence Transformers to generate embeddings and exposes an endpoint through FastAPI to match user moods against movie plots. Truncating the Matryoshka embeddings from 2048 to 1024 dimensions balances semantic retrieval quality, storage footprint, and query latency. Testing shows that while abstract emotional queries return moderate similarity scores around 0.62 to 0.67, concrete descriptive queries exceed 0.75 without requiring exact keyword matches.


### [How Stripe rolled out a consistent Cursor experience for 3,000 engineers](https://yomu.fyi/post/how-stripe-rolled-out-a-consistent-cursor-experience-for-3-000-enginee.md)
- Company: [Cursor](https://yomu.fyi/company/cursor.md)
- Author: Cursor
- Published: Feb 17, 2026
- Type: Explainer

Stripe rolled out Cursor to more than 3,000 engineers while preserving its standards for developer experience and API uptime. The company preloaded Cursor on every machine, taught it in onboarding labs alongside VS Code and IntelliJ, and used Cursor Rules to supply codebase and coding-standard context, with teams able to add local rules. Adoption was reinforced by power users, while code review evolved to use LLMs to flag complex methods and risky files so reviewers could focus attention without sacrificing quality. An internal survey found developer sentiment at a five-year high, and MacVicar observed the largest productivity gains among longest-tenured engineers rather than junior developers. That finding has prompted exploration of letting new hires learn the codebase before receiving Cursor access, reinforcing MacVicar’s advice to start before every AI-workflow question is settled.


### [Extend Cursor with plugins](https://yomu.fyi/post/extend-cursor-with-plugins.md)
- Company: [Cursor](https://yomu.fyi/company/cursor.md)
- Author: Cursor
- Published: Feb 17, 2026
- Type: Announcement

Cursor now supports plugins, allowing agents to connect to external tools and acquire new knowledge. Plugins bundle MCP servers, skills, subagents, rules, and hooks to add custom functionality. The initial curated set includes Amplitude, AWS, Figma, Linear, Stripe, Cloudflare, Vercel, Databricks, Snowflake, and Hex, covering planning, design, payments, infrastructure, data, and analytics workflows. Users can install prebuilt plugins from the Cursor Marketplace or create and share their own, while plugin primitives let agents discover domain-specific skills, run subagents in parallel, connect through MCP servers, observe behavior with hooks, and follow rules. The announcement also introduces the Cursor Team Kit for internal CI, code review, and testing workflows, and says private team marketplaces with governance and security controls are coming.


### [Anthropic opens Bengaluru office](https://yomu.fyi/post/anthropic-opens-bengaluru-office.md)
- Company: [Anthropic](https://yomu.fyi/company/anthropic.md)
- Published: Feb 16, 2026
- Type: Announcement

Anthropic has officially opened a Bengaluru office, marking its second in Asia after Tokyo, and announced partnerships spanning enterprise, education, agriculture, and public services in India. India is Claude.ai's second-largest market, with nearly half of usage involving computer and mathematical tasks; Anthropic says its run-rate revenue there has doubled since its October 2025 expansion. The company is improving support for Indic languages through training data in ten languages and is working with Karya and the Collective Intelligence Project on evaluations for locally relevant tasks in agriculture and law. Other initiatives include Claude-powered education pilots, public-sector projects such as an Adalat AI WhatsApp helpline, and adoption of the open-source Model Context Protocol, while the office will hire local talent and provide applied AI expertise to customers.


### [Building AI-Powered Observability with Heroku Managed Inference and Agents](https://yomu.fyi/post/building-ai-powered-observability-with-heroku-managed-inference-and-ag.md)
- Company: [Heroku](https://yomu.fyi/company/heroku.md)
- Author: Karunasri (Karuna) Garigipati
- Published: Feb 13, 2026
- Type: Tutorial

Production incident debugging often requires switching among an IDE, Splunk, Sentry, and PagerDuty while translating natural-language questions into platform-specific queries. The described system connects AI coding assistants to observability platforms through the Model Context Protocol (MCP), exposing typed tools for Splunk searches, Honeycomb queries, and Sentry event lookups. Heroku Managed Inference and Agents provides an MCP gateway for SSE transport, allowing servers to run as STDIO processes and be deployed with a Heroku app, an inference add-on, credentials, and an inference token configured in tools such as Cursor or Claude. The post reports ranked production error results, processing more than 175,000 events in minutes for a suspicious-account review, and using natural-language questions to reduce syntax barriers and keep telemetry in the IDE.


### [Everyone wants AI. That isn't enough.](https://yomu.fyi/post/everyone-wants-ai-that-isn-t-enough.md)
- Company: [Notion](https://yomu.fyi/company/notion.md)
- Author: Makenna Lowe
- Published: Feb 13, 2026
- Type: Explainer

Survey data from more than 3,000 global knowledge workers and 85-plus interviews examines why strong demand for workplace AI has not translated into broad adoption. Although 93 percent call AI essential to modern productivity software and 80 percent are likely to switch to a platform offering it, only 20 percent say AI is deeply integrated into their company’s tools. Respondents prioritize recurring workflows such as automated project updates, continuously current knowledge hubs, and goal-based templates, while decision-makers identify integration as the leading implementation challenge alongside security and data quality, unclear strategy, resource constraints, and compatibility. The proposed pattern is to place AI inside existing work, illustrated by Faire’s reported results with Notion AI: smoother meetings for 81 percent of surveyed staff and workflows running 30 percent faster.


### [CX Score: How we built a metric support leaders can defend](https://yomu.fyi/post/cx-score-how-we-built-a-metric-support-leaders-can-defend.md)
- Company: [Intercom](https://yomu.fyi/company/intercom.md)
- Author: Mark O'Donovan
- Published: Feb 13, 2026
- Type: Explainer

CX Score was built to give support leaders a defensible view of customer experience across conversations, addressing whether an automated metric can be trusted. The team grounded its criteria in how experienced support professionals assess strong, mixed, and poor interactions, then created a dataset of thousands of conversations reviewed by experts, with two reviewers where possible and disagreement resolution. It operates as a production system using specialized prompt engineering, support-focused model configuration, neutral defaults for ambiguous evidence, and multi-stage evaluation across LLM models. Precision, recall, F1, regression, and bias checks set a required F1 above 0.8, while field testing with real customers examined varied tickets, languages, agent types, and conversation lengths. The stated conclusion is that CX Score is designed to be inspected, explained, and defended rather than treated as a black-box output.


### [Box chooses Cursor for enterprise-grade quality, security, and control](https://yomu.fyi/post/box-chooses-cursor-for-enterprise-grade-quality-security-and-control.md)
- Company: [Cursor](https://yomu.fyi/company/cursor.md)
- Author: Cursor
- Published: Feb 13, 2026
- Type: Announcement

Box selected Cursor as its enterprise agentic coding platform after evaluating data privacy, security, and agent-quality controls. More than 85% of Box developers, or over 800 people, now use Cursor daily, while the company reports a 30–50% increase in product-roadmap throughput and migrations completed 80–90% faster. Box encodes agent guardrails in custom Cursor rules and commands, including a frontend AI toolkit, automated test-coverage work, and rules for React and design-system migrations. A mentorship program pairing power users with newer adopters increased Cursor usage by 75% and power users by 25% in six weeks, while Box says it can move faster without compromising product quality, security, or accessibility.


### [Supabase incident on February 12, 2026](https://yomu.fyi/post/supabase-incident-on-february-12-2026.md)
- Company: [Supabase](https://yomu.fyi/company/supabase.md)
- Author: paul\_copplestone
- Published: Feb 13, 2026
- Type: Incident / postmortem

Supabase reports a 3-hour-and-42-minute outage on February 12–13, 2026, affecting every service in its us-east-2 region, including Postgres, Auth, Data APIs, Edge Functions, Storage, and Realtime. A deployment of an internal monitoring service enabled AWS VPC Block Public Access in block-bidirectional mode, blocking internet gateway traffic across production VPCs in the region. The incident response was delayed because alerts pointed toward shared services, the relevant CloudTrail event was inconspicuous, pre-production lacked us-east-2, and infrastructure specialists were not involved initially. Supabase restored connectivity by rolling back and removing the monitoring stack, then audited regions, added AWS Organizations policies and IaC safeguards, and planned account isolation, external connectivity probes, production-parity testing, and improved escalation.


### [Custom Kernels for All from Codex and Claude](https://yomu.fyi/post/custom-kernels-for-all-from-codex-and-claude.md)
- Company: [Hugging Face](https://yomu.fyi/company/hugging-face.md)
- Author: ben burtenshaw, Sayak Paul, Aritra Roy Gosthipaty, shaun smith
- Published: Feb 13, 2026
- Type: Announcement

Writing CUDA kernels that integrate with diffusers and transformers involves GPU-specific optimization, library integration, PyTorch registration, and environment compatibility. It presents a cuda-kernels agent skill that packages architecture guides, integration patterns, kernel templates, troubleshooting material, benchmarks, and Kernel Hub examples for coding agents. Claude and Codex used it to generate RMSNorm, RoPE 3D, GEGLU, and AdaLN kernels for LTX-Video in diffusers, and an RMSNorm kernel for Qwen3-8B in transformers, including CUDA source, PyTorch bindings, build configuration, and benchmark scripts. On H100 80GB HBM3, the LTX-Video RMSNorm tests averaged 1.88x speedup and the optimized pipeline with torch.compile reached 1.43x the listed baseline, while Qwen3-8B RMSNorm averaged 1.94x and reached 2.47x at 8192 tokens. The skill supports packaging kernels for precompiled distribution through the Hub, where get\_kernel selects a matching binary across Python, PyTorch, and CUDA variants.


### [The evolution of support starts now](https://yomu.fyi/post/the-evolution-of-support-starts-now.md)
- Company: [Intercom](https://yomu.fyi/company/intercom.md)
- Author: Yohanna Andom
- Published: Feb 12, 2026
- Type: Announcement

Intercom introduces Resolution Bot, a machine-learning support chatbot designed to resolve customer issues rather than merely direct people to help-center articles. In the Intercom Messenger, it provides tailored resolutions, can use integrations such as Shopify to surface order status, and proactively suggests answers while customers type. The update adds six languages—Spanish, French, Italian, Brazilian Portuguese, Russian, and German—along with demographic targeting based on factors such as spend, business type, and location. New reporting shows answer-level sends and resolution rates, identifies underperforming answers, and lets teams filter conversation metrics for bot interactions. Intercom reports that the bot automatically resolves 33% of common support questions and saved businesses an average of 109 hours of customers’ time over the last year, positioning the release as a way to provide faster, more consistent support at global scale.


### [Trusting the Untestable: Validation and Diagnostics for the Doubly Robust Models](https://yomu.fyi/post/trusting-the-untestable-validation-and-diagnostics-for-the-doubly-robu.md)
- Company: [Lyft](https://yomu.fyi/company/lyft.md)
- Author: Shima Nassiri
- Published: Feb 12, 2026
- Type: Problem & solution

When randomized A/B tests are infeasible for evaluating partnerships or long-term effects, Lyft applies Augmented Inverse Propensity Weighting (AIPW) doubly robust models to estimate treatment effects from observational data. The quasi-experimentation platform mandates rigorous confounder selection prior to treatment exposure and adjusts for downsampling bias via propensity score conversions and outcome reweighting. When comparing AIPW against experimental ground truths, initial estimates understated effects because trimming extreme propensity scores removed key users and unobserved confounders persisted. To evaluate model health and external validity, the platform incorporates diagnostic scorecards featuring common support checks, Marginal Sensitivity Models, and covariate comparisons between trimmed and untrimmed cohorts.


### [Balancing cost and reliability for Spark on Kubernetes](https://yomu.fyi/post/balancing-cost-and-reliability-for-spark-on-kubernetes.md)
- Company: [Notion](https://yomu.fyi/company/notion.md)
- Author: Justin Lee
- Published: Feb 12, 2026
- Type: Problem & solution

Notion describes how it reduced Spark compute costs on Kubernetes while maintaining reliability across production workloads. After moving from fixed EC2 clusters on Amazon EMR to EMR on EKS, it used Karpenter, EKS Auto Mode, and the MostAllocated scheduler for dynamic provisioning and bin-packing. Spot Instances then exposed a failure mode: one interruption could remove many bin-packed executors, causing shuffle loss and expensive recomputation. Spot Balancer addresses this by labeling executor creation through a Kubernetes webhook and maintaining each job’s configured spot-to-on-demand ratio, with per-job counts stored in Redis. Named stability weights make the cost-reliability trade-off explicit, and reported savings reached 60–90 percent while jobs that previously failed from interruptions finished reliably.


[Newer posts](https://yomu.fyi/posts/page/156.md) · [Older posts](https://yomu.fyi/posts/page/158.md)