---
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

### [MAPS: Netflix’s Multimodal Asset Personalization at Scale](https://yomu.fyi/post/maps-netflix-s-multimodal-asset-personalization-at-scale.md)
- Company: [Netflix](https://yomu.fyi/company/netflix.md)
- Author: Netflix Technology Blog
- Published: Aug 28, 2026
- Type: Explainer

Netflix treats artwork and video previews as personalized assets, but ID-based models struggle with newly launched titles and assets that have little interaction history. The systems encode artwork with 768-dimensional CLIP image embeddings and combine those vectors with learned asset IDs, allowing visual preferences to transfer across titles and enabling one model to serve five canvases. Reward-based weighting balances pooled canvas data by the long-term value of interactions, while inverse propensity scoring on randomized exploration traffic supports offline evaluation. For video previews, MediaFM combines visual, audio, and text signals and ranked ahead of SeqCLIP in a linear probe, offline IPS evaluation, and online A/B testing; the Netflix Embedding Store supplies consistent embeddings to production models.


### [Introducing Claude for Teachers](https://yomu.fyi/post/introducing-claude-for-teachers.md)
- Company: [Anthropic](https://yomu.fyi/company/anthropic.md)
- Published: Aug 28, 2026
- Type: Announcement

Claude for Teachers gives verified US K-12 educators free access to premium Claude capabilities, tailored teaching skills, Learning Commons curricula, and academic standards across all 50 states. It is designed to help teachers apply differentiation, mastery-based learning, and small-group instruction despite limited time, resources, and large classes, with an emphasis on protecting time with students. Once verified, users can ask Claude to draft standards-aligned lessons, differentiate materials, analyze de-identified classroom data, and schedule recurring tasks; Claude Code and Cowork support carrying work forward. The offering includes no model training on verified teacher conversations, FERPA-aligned protections, and district and state control over educational-data use. Anthropic is releasing connectors, open-source teaching skills, and an evaluation write-up, while planning a Detroit Public Schools Community District pilot focused on educator wellbeing and practice.


### [The State of Django 2026: Boring is so back](https://yomu.fyi/post/the-state-of-django-2026-boring-is-so-back.md)
- Company: [Jetbrains](https://yomu.fyi/company/jetbrains.md)
- Author: Will Vincent
- Published: Aug 28, 2026
- Type: Explainer

The fifth annual Django Developers Survey, produced by Django Software Foundation and PyCharm, draws on responses from nearly 3,500 developers in more than 40 countries. It reports that Django’s core remains stable, with PostgreSQL used by 76–79% of respondents across five years, templates around 80%, and 43% already on Django 6.0. At the same time, AI is routine but fragmented: 58% of AI users work with it daily, while 56% use it only for chat and advice and just 27% use autonomous multi-step tasks. The survey also describes rapid consolidation around uv and Ruff, growing adoption of type hints, and a testing and CI/CD base that can verify agent-generated changes, although 19% write no automated tests. Django supports both server-rendered and API/frontend patterns, while its stable foundation lets developers change surrounding tools and workflows without replacing the framework.


### [BotBase for Operators: A clearer path to joining Cloudflare's directory of bots and agents](https://yomu.fyi/post/botbase-for-operators-a-clearer-path-to-joining-cloudflare-s-directory.md)
- Company: [Cloudflare](https://yomu.fyi/company/cloudflare.md)
- Author: Julian Laxman
- Published: Aug 28, 2026
- Type: Announcement

Cloudflare launched BotBase for Operators to give bot operators a clearer way to submit, monitor, update, and cancel bot-directory submissions from the dashboard. The new area includes a bots directory, submission form, and history tab with Waiting for review, Accepted, and Rejected statuses, including rejection reasons and classification changes. Its intake model records what a bot does, how it uses content, and whether it is direct or intermediary, while allowing multiple behaviors and Content Signals declarations. Reviews now use automated checks for duplicates, user-agent specificity, and verification through IP lists, reverse DNS, or Web Bot Auth, routing exceptions to staff. The launch focuses on visibility, with ownership, observability, and conversation identified as future goals.


### [A new getting started experience for ClickHouse Managed Postgres](https://yomu.fyi/post/a-new-getting-started-experience-for-clickhouse-managed-postgres.md)
- Company: [Clickhouse](https://yomu.fyi/company/clickhouse.md)
- Published: Aug 28, 2026
- Type: Announcement

ClickHouse Managed Postgres introduces a four-step onboarding experience designed to move users from provisioning a service to working with operational data and ClickHouse analytics. Existing onboarding placed new users on an empty Overview page, while the documentation quickstart required users to track progress outside the product. The new flow guides users through creating a service, migrating or ingesting data, running a query in the SQL Console, and configuring Postgres CDC to ClickHouse. Data can arrive through a managed Postgres-to-Postgres ClickPipe, a language client, a sample dataset, or a table created from scratch; pg\_clickhouse provides a PostgreSQL query layer over ClickHouse foreign tables. The article reports less than a minute of replication lag and benchmarks showing five aggregations falling from 555 ms to 164 ms and a JOIN with aggregations from 1,246 ms to 170 ms on one million-row data set.


### [Expanding our support for scientists](https://yomu.fyi/post/expanding-our-support-for-scientists.md)
- Company: [Anthropic](https://yomu.fyi/company/anthropic.md)
- Published: Aug 28, 2026
- Type: Announcement

Anthropic is expanding support for scientific researchers through free and discounted Claude subscriptions, research credits, and products designed for scientific work. The new Claude team plan for scientists opens 10,000 seats worldwide for one year: standard seats are free, while premium seats with five-times usage limits cost $15 per month; the program is intended to grow beyond that initial allocation. Its AI for Science program, previously focused largely on biological sciences, will offer credits to researchers in additional fields, including compute-heavy work, with up to $50,000 available per project after lab usage exceeds standard or premium allowances. Access requires principal investigators or equivalent leaders at academic or nonprofit institutions to verify their labs, while biology and chemistry researchers remain limited to Opus-class models and Fable continues to block professional biology and drug-development queries because of dual-use risks.


### [Security Incident Affecting JetBrains Cadence](https://yomu.fyi/post/security-incident-affecting-jetbrains-cadence.md)
- Company: [Jetbrains](https://yomu.fyi/company/jetbrains.md)
- Author: Daniel Gallo
- Published: Aug 28, 2026
- Type: Incident / postmortem

JetBrains reports a security incident affecting Cadence, its hosted service that integrates with PyCharm through an optional plugin and runs projects on cloud compute resources. Cadence used TeamCity to orchestrate workloads, and attackers exploited the environment through critical vulnerability CVE-2026-63077. The attackers accessed personal data, compromised a 2024 server backup and Cadence-related AWS IAM credentials, and accessed files in JetBrains AWS S3 buckets; the scope remains under investigation. JetBrains discovered the exploitation on August 23, took the server offline on August 24, invalidated plugin access tokens, and urges users to rotate secrets, investigate connected systems, review repositories, and treat executions and synchronized files as untrusted.


### [From one switch to a control panel: meet \`dataCollection\`](https://yomu.fyi/post/from-one-switch-to-a-control-panel-meet-datacollection.md)
- Company: [Sentry](https://yomu.fyi/company/sentry.md)
- Author: Sigrid Huemer
- Published: Aug 28, 2026
- Type: Explainer

Sentry is replacing the JavaScript SDK boolean option sendDefaultPii with dataCollection, a category-level control for automatically collected data. In version 11, dataCollection becomes the default, sendDefaultPii is removed, and several categories—including user information, cookies, HTTP bodies, GenAI inputs and outputs, database query data, and stack-frame variables—are collected by default. The post distinguishes PII, which is collected by default, from sensitive credentials, whose values in collected headers are filtered automatically; manually attached data remains unaffected. Developers can preserve v10’s restrictive baseline or select true, false, allow lists, and deny lists for supported fields, while beforeSend and beforeSendSpan handle additional filtering. Because v11 also enables span streaming by default, span drops move to ignoreSpans, and existing withStreamedSpan wrappers should be removed.


### [Self-service backup for VMs and containers on OpenShift: No cluster-admin required](https://yomu.fyi/post/self-service-backup-for-vms-and-containers-on-openshift-no-cluster-adm.md)
- Company: [Red Hat](https://yomu.fyi/company/red-hat.md)
- Author: Michal Pryc, Shubham Dilip Pampattiwar
- Published: Aug 28, 2026
- Type: Problem & solution

In multi-tenant Red Hat OpenShift clusters, Velero Backup and Restore resources live in the OADP installation namespace, leaving non-admin application teams dependent on platform staff. OADP self-service addresses this with namespace-scoped NonAdminBackup and NonAdminRestore resources that proxy Velero resources, while a controller validates requests against DataProtectionApplication policies and synchronizes phase, progress, and queue status back to the tenant namespace. Administrators can enforce retention, snapshot and file-system backup settings, resource inclusion or exclusion, and approval for tenant-provided backup storage; cluster-scoped resources such as SCCs, ClusterRoles, CRDs, and KubeVirt cluster preferences are automatically excluded. OADP 1.6 supports the same workflow for containers, stateful applications, and OpenShift Virtualization VMs, but self-service has no native schedules, cross-namespace restores, or short-lived credentials for tenant-provided storage locations.


### [Project Loom in IntelliJ IDEA: Virtual Threads, Scoped Values, and Structured Concurrency](https://yomu.fyi/post/project-loom-in-intellij-idea-virtual-threads-scoped-values-and-struct.md)
- Company: [Jetbrains](https://yomu.fyi/company/jetbrains.md)
- Author: Marit van Dijk
- Published: Aug 28, 2026
- Type: Tutorial

Project Loom in IntelliJ IDEA introduces Java developers to three cohesive concurrency features: Virtual Threads (JEP 444, stable since Java 21), Scoped Values (JEP 506, stable since Java 25), and Structured Concurrency (JEP 533, seventh preview in Java 27). The post identifies pain points such as CompletableFuture ignoring interrupt flags, manual ExecutionException unwrapping chains, duplicated cancel() calls across catch blocks, mutable ThreadLocal pitfalls, and flat thread dumps lacking request hierarchy. A CustomerProfileService example is refactored from CompletableFuture.allOf to StructuredTaskScope.ShutdownOnFailure, where sibling tasks are cancelled automatically and a single timeout covers the whole scope. IntelliJ IDEA debugger support from version 2026.1 groups virtual threads forked within a StructuredTaskScope into containers representing their scopes, and the post walks through downloading JDK 27 (EA) and enabling the experimental language level.


### [Building hermetic notebook images for Open Data Hub and Red Hat OpenShift AI](https://yomu.fyi/post/building-hermetic-notebook-images-for-open-data-hub-and-red-hat-opensh.md)
- Company: [Red Hat](https://yomu.fyi/company/red-hat.md)
- Author: Vath Sok
- Published: Aug 28, 2026
- Type: Problem & solution

Open Data Hub and Red Hat OpenShift AI adopted hermetic container builds to eliminate network failures, satisfy Conforma compliance, and ensure reproducible notebook environments. Instead of pulling packages during image creation, dependencies are pinned in lockfiles, prefetched into a local cache using tools like Cachi2, and installed with the network disabled. The workflow relies on a unified Containerfile capable of running locally, within GitHub Actions, and on Konflux via Tekton tasks. To pass audit checks and avoid generic tarball downloads, the team packaged awkward binaries like ripgrep and Pandoc into multi-arch Python wheels while patching Node postinstall scripts that fetch files at install time. Upstream CentOS and downstream Red Hat Enterprise Linux dependencies are maintained in distinct lock trees to prevent subtle packaging mismatches.


### [Fast, fault-tolerant PyTorch training on AI Runtime](https://yomu.fyi/post/fast-fault-tolerant-pytorch-training-on-ai-runtime.md)
- Company: [Databricks](https://yomu.fyi/company/databricks.md)
- Author: Bruce Fontaine
- Published: Aug 28, 2026
- Type: Explainer

At scale, training jobs face frequent GPU interruptions, while remote-storage input pipelines can leave accelerators idle. Distributed Checkpoint (DCP) has each rank write a shard plus a .metadata file, allowing parallel saves and reloads onto a different number of GPUs; DCP is presented as useful for DDP as well as sharded training. Asynchronous saves stage data locally and upload in the background; cited measurements report 1.8x savings for a 2.8B-parameter DDP model and 58x for a 20B-parameter FSPD model on 32 H100s, though the comparison excludes torch.save network storage time. UCVolumeDataset and the accompanying DataLoader cache files on local NVMe and prefetch disjoint rank/worker slices while computation runs. Correct recovery also requires checkpointing data position plus seeds and RNG state so resumed sampling remains consistent.


### [GLM-5.3 vs. GLM-5.3 Flash on DeepSWE: Cost, Coding, and Routing](https://yomu.fyi/post/glm-5-3-vs-glm-5-3-flash-on-deepswe-cost-coding-and-routing.md)
- Company: [Together AI](https://yomu.fyi/company/together-ai.md)
- Published: Aug 28, 2026
- Type: Benchmark

The GLM-5.3 Flash, a distilled variant of the full GLM-5.3, achieves a pass@1 of 63.4% on the DeepSWE benchmark at $0.24 per rollout, 17 times cheaper than the full model’s 69.0% at $3.99. The pass@4 gap narrows to 2.6 points (87.6% versus 85.0%). Distillation removed consistency rather than capability: the Flash solves 93 of the 99 tasks the full model solves, but its ability to convert extra effort into a win is lost (46% win rate on flaky tasks versus 61%). A cascade that runs Flash first and escalates to the full model only when the answer is rejected solves 80.9% of tasks at $1.70 each, 57% lower cost than using the full model alone.


### [How to Approve Real Users and Block Fake Accounts at Sign Up with Lookup and Verify in C#](https://yomu.fyi/post/how-to-approve-real-users-and-block-fake-accounts-at-sign-up-with-look-csharp.md)
- Company: [Twilio](https://yomu.fyi/company/twilio.md)
- Author: Amanda Lange, Dhruv Patel, Kelley Robinson
- Published: Aug 28, 2026
- Type: Tutorial

Fake accounts and fraudulent sign-ups motivate a C# onboarding flow that screens phone numbers before sending an OTP. It combines Twilio Lookup line type intelligence, optional line status checks, and Identity Match with Twilio Verify phone verification. The pipeline rejects landline, nonfixed VoIP, toll-free, and pager types, requires first- and last-name results of exact\_match or high\_partial\_match, and can reject lookup failures or mismatches. The post supplies a .NET 9 project setup, environment variables, MVC files, and service code for starting and checking verification. It reports example logs for a mobile number with active status and exact name matches, demonstrates blocked toll-free or VoIP testing, and notes that Lookup and Verify pricing varies by package and country.


### [How to Use Svelte and Go to Build a Video Chat App](https://yomu.fyi/post/how-to-use-svelte-and-go-to-build-a-video-chat-app.md)
- Company: [Twilio](https://yomu.fyi/company/twilio.md)
- Author: Donal Toomey
- Published: Aug 28, 2026
- Type: Tutorial

The tutorial builds a video chat application with a Svelte single-page frontend and Go REST API, using Twilio Video for room-based calls. Users enter a room name, the backend validates it, mints a Twilio Access Token with a VideoGrant, and the frontend uses that token to connect and attach local and remote video tracks. The backend loads credentials from .env.local, assigns each participant a random crypto/rand identity, enables CORS, and serves requests on port 8000; room names shorter than six characters receive a 400 response. The frontend uses Vite, the Twilio Video JavaScript SDK, and iziToast, while testing involves running both servers and joining the same room in two browser tabs. The completed app handles video communication without custom permission, real-time communication, or data-streaming code.


### [How to Confirm Phone Number Ownership with Lookup Identity Match in Python](https://yomu.fyi/post/how-to-confirm-phone-number-ownership-with-lookup-identity-match-in-py.md)
- Company: [Twilio](https://yomu.fyi/company/twilio.md)
- Author: Kelley Robinson, Dylan Frankcom
- Published: Aug 28, 2026
- Type: Tutorial

The tutorial shows how to use Twilio Lookup Identity Match in Python to confirm that signup information is associated with a phone number and reduce signup fraud. It builds a Flask application that accepts an E.164-formatted number, first name, and last name, then calls the Lookup API with fields="identity\_match" and the supplied identity attributes. The response exposes statuses including exact\_match, high\_partial\_match, no\_match, and no\_data\_available; the example accepts a result only when the number is valid and both names are exact or high partial matches. The implementation covers project setup with Flask, twilio, and python-dotenv, environment-based credentials, an intl-tel-input form, and frontend handling for success and failure messages. It also notes that Identity Match can use combinations of name, address, national ID, or date of birth, and points to other Lookup packages such as Line Type Intelligence and SIM swap detection.


### [Why 78% of customers try to bypass your AI agent](https://yomu.fyi/post/why-78-of-customers-try-to-bypass-your-ai-agent.md)
- Company: [Twilio](https://yomu.fyi/company/twilio.md)
- Author: Jesse Sumrak
- Published: Aug 28, 2026
- Type: Explainer

Twilio research cited in the post finds that 78% of consumers have tried to bypass an AI agent to reach a human, while 63% want easy escalation and only 47% of brands provide it. The reported causes are unclear AI disclosure, discomfort with high-stakes tasks, and prior failures: 90% of tested consumers misidentify text-based AI, 80% report a negative AI experience, and failed attempts add 81 seconds before escalation. Consumers are most comfortable with scheduling and reservations, returns and exchanges, and concert seat selection, but pull back on medical inquiries, payments, and loan applications. The proposed operating model combines routine automation with clear disclosure, obvious escalation, persistent customer memory, and real-time detection of sentiment and escalation risk. It argues that trust, rather than wholesale replacement, should determine how humans and AI divide customer-service work.


### [AI didn't write our SDK. It changed how we built it.](https://yomu.fyi/post/ai-didn-t-write-our-sdk-it-changed-how-we-built-it.md)
- Company: [Stripe](https://yomu.fyi/company/stripe.md)
- Author: Mike North
- Published: Aug 28, 2026
- Type: Explainer

Stripe's extensibility SDK was built with AI involved across prototyping, architecture, implementation, and refinement, but the main change was to the development loop rather than code-generation speed. For a custom-object design, one engineer produced a working prototype, demo, and documentation site in roughly two days instead of an estimated two to three weeks, allowing the team to evaluate boundaries and reprioritize the feature for the next major launch. The team then used reviewed, detailed specifications and persistent domain guidance to delegate bounded work, including about twenty ESLint rules covering authoring constraints such as sandbox boundaries, imports, and type safety. AI also helped keep scripts, workflows, SDK generation, runtime enforcement, tooling, documentation, examples, and tests aligned, while human judgment remained responsible for abstractions and trade-offs. The post concludes that AI's leverage came from making more artifacts concrete earlier, enabling earlier feedback and a tighter deliverable.


### [The Open ASR Leaderboard Adds Its First Global South Language](https://yomu.fyi/post/the-open-asr-leaderboard-adds-its-first-global-south-language.md)
- Company: [Hugging Face](https://yomu.fyi/company/hugging-face.md)
- Author: Eric Bezzam, Shobhit Banga, Manas Dhir, Bhaskar Singh, Manmeet Kaur, Aaditya Pareek, Walecha, Sagar Jain, Hanuman Sidh, Vanshika Chhabra
- Published: Aug 28, 2026
- Type: Announcement

Voice Arena and Hugging Face have added two Monsoon evaluation sets, en-IN and hi-IN, to the Open ASR Leaderboard, introducing Indian English and Hindi coverage. The initiative addresses a limitation of aggregate WER: existing test sets record what was said but little about who spoke, making disparities across geography, age, gender, accent, devices, and other conditions difficult to see. The four speaker-disjoint public and private splits contain 4,888 speakers and vary across nine collection axes, using unscripted dual-channel conversations, broad district and device coverage, and 12 speaker metadata attributes. Because Hindi has extensive valid orthographic variation, its references use manually curated lattices and are scored with Orthographically-Informed Word Error Rate (OIWER); comparisons with flattened references show higher, non-uniform error rates and changed system rankings.


### [LLMs Are Not (Consistently) Bayesian: Quantifying Internal (In)consistencies of LLMs’ Probabilistic Beliefs](https://yomu.fyi/post/llms-are-not-consistently-bayesian-quantifying-internal-in-consistenci.md)
- Company: [Apple](https://yomu.fyi/company/apple.md)
- Published: Aug 28, 2026
- Type: Explainer

LLMs are deployed in medicine, science, and law, where observed evidence may not determine a single correct answer and systems must represent and update uncertain beliefs. The work treats LLMs as information processing rules and measures an information processing gap, defined as deviation from Bayes updates, to quantify internal consistency in probabilistic belief updates. Extensive experiments compare multiple ways of incorporating evidence: some yield nearly Bayesian updates, while others rely on learned heuristics. Although exact Bayesian updating is optimal information processing, the non-Bayesian heuristic updates often achieve better downstream task performance, suggesting that the LLMs’ probabilistic models of the world are misspecified. The measure also serves as a diagnostic for issues in LLM-powered inferential systems.


[Newer posts](https://yomu.fyi/posts/page/62.md) · [Older posts](https://yomu.fyi/posts/page/64.md)