---
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

### [Agent Seer: Synthesizing Scenarios from Specification Understanding](https://yomu.fyi/post/agent-seer-synthesizing-scenarios-from-specification-understanding.md)
- Company: [Apple](https://yomu.fyi/company/apple.md)
- Published: Aug 28, 2026
- Type: Benchmark

Agent Seer addresses the problem of creating realistic, scalable test scenarios for AI agents that use external tools, where hand-built cases require domain expertise and static benchmarks cannot track changing APIs. It synthesizes scenarios from a single Model Context Protocol (MCP) specification containing function names, natural-language descriptions, and typed parameter schemas, without examples, live tool access, or domain-specific tuning. The pipeline enriches raw schemas, generates graded scenarios with synthetic tool outputs, and expands them into mock-data-grounded, multi-turn dialogues, then evaluates tool-calling correctness and conversational coherence across seven MCP specifications spanning domains and tool-suite sizes. It achieves strong quality across all domains and complete tool coverage on small and medium specifications; parameter schema complexity correlates most strongly with quality variation, while tool-suite size has a smaller orthogonal role. Among imperfect scenarios, argument value accuracy is the dominant failure mode, which coarse-grained name-match metrics do not reveal.


### [Data Mesh at Grab (Part III): Operationalizing data reliability with automated DPIs](https://yomu.fyi/post/data-mesh-at-grab-part-iii-operationalizing-data-reliability-with-auto.md)
- Company: [Grab](https://yomu.fyi/company/grab.md)
- Author: Harvey Li
- Published: Aug 28, 2026
- Type: Problem & solution

Grab’s data mesh uses the Signals Marketplace to publish, discover, and reuse trusted data products, while certification establishes ownership, documentation, lineage, and quality controls but does not prevent runtime failures. Data Production Issues (DPIs) turn data-quality signals into an operational workflow that begins when Kinabalu detects unhealthy contract tests and closes when the underlying condition recovers. Triage promotes meaningful contract breaches, deduplicates repeated signals, groups related issues, and records affected assets, intervals, contracts, tests, and dependencies. For diagnosis, the Data Health API exposes a portable taxonomy of UPSTREAM\_ERROR, PLATFORM\_ERROR, JOB\_ERROR, and DATA\_ERROR, allowing ownership to follow the likely root cause while platforms retain detailed internal diagnostics. Combined with Hugo’s diagnosis and remediation patterns, the approach is presented as reducing resolution time and feeding operational reliability back into certification.


### [So, is ClickHouse winning the observability wars?](https://yomu.fyi/post/so-is-clickhouse-winning-the-observability-wars.md)
- Company: [Clickhouse](https://yomu.fyi/company/clickhouse.md)
- Published: Aug 27, 2026
- Type: Explainer

The article examines whether ClickHouse is “winning” observability by separating the storage-and-query question from the broader work of collection, schema design, correlation, visualization, alerting, and investigation. It argues that ClickHouse fits telemetry because its columnar layout stores fields separately, enabling compression and allowing analytical queries to read only selected columns across millions or billions of events; ordering keys, sparse primary indexes, and data-skipping indexes further reduce work. The discussion covers both rich, wide events and conventional logs, traces, and metrics, and presents ClickStack as a combination of ClickHouse, an OpenTelemetry Collector, and a UI that preserves familiar workflows. The conclusion is qualified: ClickHouse has become the de facto storage and query engine for many teams and a strong default for large telemetry workloads, but it has not won observability as a whole, while full-text indexes, PromQL support, and broader product layers remain important.


### [How we saved 100 terabytes of memory by optimizing 1.1.1.1’s DNS cache](https://yomu.fyi/post/how-we-saved-100-terabytes-of-memory-by-optimizing-1-1-1-1-s-dns-cache.md)
- Company: [Cloudflare](https://yomu.fyi/company/cloudflare.md)
- Author: Sebastiaan Neuteboom
- Published: Aug 27, 2026
- Type: Problem & solution

Cloudflare's Big Pineapple platform, which powers 1.1.1.1 and other DNS services, stores over 250 billion cache entries simultaneously. To decrease fleet-wide memory consumption, engineers implemented five changes to in-memory cache data structures. Key modifications included swapping growable types like Vec and String for Box slices to eliminate capacity overhead, combining distinct record sections into a single list indexed by offsets, omitting redundant owner domain names, and serializing record bytes into contiguous buffers. In benchmarks, these optimizations decreased per-entry memory footprints by 56% while improving cache lookup latency by 19% and insert throughput by 43%. Across production instances, the rollout reduced working-set memory by approximately 100 terabytes.


### [One Write Path, Multiple Entries: How Rippling uses Unity Catalog-managed Apache Iceberg™](https://yomu.fyi/post/one-write-path-multiple-entries-how-rippling-uses-unity-catalog-manage.md)
- Company: [Databricks](https://yomu.fyi/company/databricks.md)
- Author: Tae Lee
- Published: Aug 27, 2026
- Type: Explainer

Rippling describes an architecture in which Databricks writes some tables while Trino, Snowflake, and other lakehouse engines read them, making catalog ownership a design decision rather than a metadata detail. Its rule is that the catalog follows the writer: AWS-native workloads use Glue, Databricks workloads use Unity Catalog, and Snowflake-produced shared data uses Horizon Catalog with Snowflake-managed Iceberg tables. For selected Databricks outputs, Unity Catalog-managed Apache Iceberg stores table data and metadata in Rippling-owned S3, exposes access through the Iceberg REST Catalog API or federation, and uses credential vending for scoped storage access. This lets downstream engines read Parquet directly without duplicate materialization or producer-side query execution. Databricks also provides Predictive Optimization for maintenance, while the trade-off is continued Unity Catalog dependence for metadata, permissions, lineage, and lifecycle.


### [Differential Privacy for Hugging Face Trainers – Without Rewriting Your Training Loop](https://yomu.fyi/post/differential-privacy-for-hugging-face-trainers-without-rewriting-your.md)
- Company: [Jetbrains](https://yomu.fyi/company/jetbrains.md)
- Author: Katie Fraser
- Published: Aug 27, 2026
- Type: Announcement

Training large language models on sensitive data risks memorization and membership inference attacks, yet standard high-level transformer tools lack native differential privacy capabilities. JetBrains Research created and open-sourced DPTrainer, a library that integrates PyTorch's Opacus framework with the Hugging Face Trainer ecosystem. The library provides differential privacy through differentially private stochastic gradient descent without requiring manual training loops or source code modifications. DPTrainer extends base training arguments with a PrivacyArguments dataclass, managing per-sample gradient calculation, gradient clipping, noise injection, and Poisson-sampled data loading. A dynamic patching mechanism injects differential privacy into specialized subclasses like DPOTrainer and SFTTrainer, while built-in privacy accountants track the cumulative privacy budget across training steps and checkpoints.


### [Building for the AI Era: Lakebase, Streaming, and Lakehouse Innovations  at VLDB 2026](https://yomu.fyi/post/building-for-the-ai-era-lakebase-streaming-and-lakehouse-innovations-a.md)
- Company: [Databricks](https://yomu.fyi/company/databricks.md)
- Author: Indrajit Roy, Ippokratis Pandis
- Published: Aug 27, 2026
- Type: Announcement

Databricks previews its VLDB 2026 keynote, four accepted papers, a demo, and a sponsor talk covering database architecture, streaming, and Lakehouse optimization. Reynold Xin will introduce Lakebase, which separates serverless PostgreSQL compute from storage, and LTAP, a transactional-analytical model intended for AI-agent workloads; Lakebase stores data and write-ahead logs in cloud object storage and supports sub-second cold starts and copy-on-write branching. The papers describe a decade of Spark Structured Streaming evolution, including microbatch pipelining that improved throughput by up to 3x, stateful APIs, and fine-grained access control, alongside AutoLiquid’s CLUSTER BY AUTO and Ultron’s history-based query optimization. AutoLiquid outperformed customer-selected clustering keys on over 95% of evaluated workloads, while Ultron improved median join latency by 25% in production workloads. The program also includes an Enzyme demonstration of incremental materialized-view maintenance and a sponsor presentation on LakehouseRT and the Reyden engine.


### [Introducing Anthropic's AI for Science Program](https://yomu.fyi/post/introducing-anthropic-s-ai-for-science-program.md)
- Company: [Anthropic](https://yomu.fyi/company/anthropic.md)
- Published: Aug 27, 2026
- Type: Announcement

Anthropic is launching Anthropic's AI for Science program, an initiative that provides free API credits to researchers pursuing high-impact scientific projects, with particular emphasis on biology and life sciences. Qualified researchers will be selected according to their contributions to science, the potential impact of proposed research, and whether AI can meaningfully accelerate the work. The program is intended to support uses including analyzing complex scientific and genetic data, generating hypotheses, designing experiments, communicating findings, understanding biological systems, accelerating drug discovery, and increasing agricultural productivity. Researchers attached to research institutions can apply through an application form, with reviews conducted by Anthropic's team and subject matter experts in relevant fields.


### [Advancing Claude for Education](https://yomu.fyi/post/advancing-claude-for-education.md)
- Company: [Anthropic](https://yomu.fyi/company/anthropic.md)
- Published: Aug 27, 2026
- Type: Announcement

Claude for Education is adding integrations with Canvas, Panopto, and Wiley to bring course materials, lecture transcripts, and peer-reviewed content into student conversations. Canvas LTI lets students use Claude inside courses, while Wiley and Panopto connect through pre-built MCP servers. Conversations are private by default, excluded from AI training by default, institutional data requests require formal approval, and self-serve data exports are limited by default. The announcement also covers new institutional partnerships, a tenfold expansion of the student ambassador program, and Claude Builder Clubs featuring hackathons, workshops, and demo nights. It frames responsible collaboration, ethics, privacy, and equitable access as necessary to avoid deepening existing divides while AI transforms education.


### [Claude for Life Sciences](https://yomu.fyi/post/claude-for-life-sciences.md)
- Company: [Anthropic](https://yomu.fyi/company/anthropic.md)
- Published: Aug 27, 2026
- Type: Announcement

Anthropic is expanding Claude into a life sciences partner intended to support work from early discovery through translation and commercialization, beyond earlier uses for coding and paper summarization. The rollout combines model improvements, scientific connectors, Agent Skills, a prompt library, and dedicated support for researchers, clinical coordinators, and regulatory affairs managers. On Protocol QA, Sonnet 4.5 scores 0.83 against a human baseline of 0.79 and Sonnet 4’s 0.74, while BixBench shows a similar improvement over its predecessor. Connectors include Benchling, BioRender, PubMed, Scholar Gateway, Synapse.org, and 10x Genomics, while a single-cell-RNA-QC skill applies scverse best practices to quality control and filtering; use cases also cover literature reviews, protocols, bioinformatics, and regulatory submissions. The offering is available through Claude.com and the AWS Marketplace, with Google Cloud Marketplace availability described as coming soon.


### [Anthropic and Iceland announce one of the world’s first national AI education pilots](https://yomu.fyi/post/anthropic-and-iceland-announce-one-of-the-world-s-first-national-ai-ed.md)
- Company: [Anthropic](https://yomu.fyi/company/anthropic.md)
- Published: Aug 27, 2026
- Type: Announcement

Anthropic and Iceland’s Ministry of Education and Children are launching a national AI education pilot that will give hundreds of teachers across every region access to Claude. The initiative combines the AI tool with educational resources, training materials, and a dedicated support network while examining how artificial intelligence can support instruction, lesson preparation, and student learning. Teachers will be able to analyze content ranging from complex texts to mathematical problems, create personalized lesson plans, adapt materials for different learners, and use Claude’s recognition of Icelandic and other languages to support students. Iceland’s education minister says the project will examine AI’s use across education with teachers’ needs guiding the work and with harm prevention in view, while Anthropic describes it as a practical model for responsible public-sector deployment.


### [Advancing Claude in healthcare and the life sciences](https://yomu.fyi/post/advancing-claude-in-healthcare-and-the-life-sciences.md)
- Company: [Anthropic](https://yomu.fyi/company/anthropic.md)
- Published: Aug 27, 2026
- Type: Announcement

Anthropic announces Claude for Healthcare and expanded Claude for Life Sciences capabilities, adding HIPAA-ready tools for providers, payers, health-tech companies, and individuals managing personal health data. Healthcare connectors cover CMS coverage determinations, ICD-10 codes, the National Provider Identifier Registry, PubMed, and beta integrations with HealthEx, Function, Apple Health, and Android Health Connect; new Agent Skills support FHIR development and prior-authorization review. Life sciences additions connect Claude to Medidata and ClinicalTrials.gov while extending support from preclinical research toward clinical trial operations and regulatory work. The announcement presents Claude Opus 4.5 as a major step in agentic performance on detailed medical and scientific simulations and says its extended-thinking mode improves correctness on honesty evaluations, while emphasizing opt-in permissions, user-controlled access, no training on health data, contextual disclaimers, and referral to healthcare professionals.


### [How scientists are using Claude to accelerate research and discovery](https://yomu.fyi/post/how-scientists-are-using-claude-to-accelerate-research-and-discovery.md)
- Company: [Anthropic](https://yomu.fyi/company/anthropic.md)
- Published: Aug 27, 2026
- Type: Explainer

Claude is being used by scientific research groups as a collaborator across experiment planning, data analysis, and discovery, rather than only for literature reviews or coding. The post describes Biomni, a Stanford platform that gives a Claude-powered agent access to hundreds of biological tools and databases; in an early trial, it completed a genome-wide association study analysis in 20 minutes, while reported validation cases included wearable data and 336,000-cell gene-activity data. Other systems encode expert workflows, including MozzareLLM for interpreting gene-knockout clusters, while a Lundberg lab project uses molecular relationship maps to propose genes for under-studied biological processes. These systems remain imperfect and require guardrails, expert-defined skills, or prospective testing, but the reported examples suggest AI can reduce research bottlenecks and support new approaches to scientific investigation.


### [Anthropic and Teach For All launch global AI training initiative for educators](https://yomu.fyi/post/anthropic-and-teach-for-all-launch-global-ai-training-initiative-for-e.md)
- Company: [Anthropic](https://yomu.fyi/company/anthropic.md)
- Published: Aug 27, 2026
- Type: Announcement

Anthropic is partnering with Teach For All to bring AI tools and training to educators in 63 countries. Through the AI Literacy & Creator Collective (LCC), more than 100,000 teachers and alumni in Teach For All’s network, serving over 1.5 million students, can develop AI fluency and adapt Claude for classroom needs. Teachers act as co-architects, giving feedback while using Claude through a six-episode fluency series, the Claude Connect peer-learning hub, and Claude Lab, which includes Claude Pro, monthly office hours, and product-roadmap input. Examples include an interactive climate curriculum built in Liberia with Claude Artifacts and a gamified mathematics app built in Bangladesh with boss battles, a leaderboard, and XP rewards.


### [Anthropic partners with Rwandan Government and ALX to bring AI education to hundreds of thousands of learners across Africa](https://yomu.fyi/post/anthropic-partners-with-rwandan-government-and-alx-to-bring-ai-educati.md)
- Company: [Anthropic](https://yomu.fyi/company/anthropic.md)
- Published: Aug 27, 2026
- Type: Announcement

Anthropic is partnering with Rwanda’s government and African technology-training provider ALX to bring Chidi, a learning companion built on Claude, to hundreds of thousands of learners across Africa. Rwanda’s ICT & Innovation and Education ministries are deploying Chidi in the national education system, with training for up to 2,000 teachers and civil servants; ALX will provide it to students through its programs, where it acts as a “Socratic mentor” rather than giving direct answers. The initiative supports skills development in areas including data analytics and cloud computing, while teachers can use Claude to plan lessons and support students. Since its November 4 rollout, Chidi has logged more than 1,100 conversations and nearly 4,000 learning sessions, with nine in ten users reporting positive experiences.


### [Anthropic partners with Allen Institute and Howard Hughes Medical Institute to accelerate scientific discovery](https://yomu.fyi/post/anthropic-partners-with-allen-institute-and-howard-hughes-medical-inst.md)
- Company: [Anthropic](https://yomu.fyi/company/anthropic.md)
- Published: Aug 27, 2026
- Type: Announcement

Modern biological research generates data at unprecedented scale, but transforming it into validated biological insights remains a fundamental bottleneck. Knowledge synthesis, hypothesis generation, and experimental interpretation still depend on manual processes, while scientific AI systems must provide reasoning researchers can evaluate, trace, and build upon. Anthropic announces partnerships with the Allen Institute and Howard Hughes Medical Institute (HHMI) to extend Claude’s capabilities into frontier research and help scientists plan and execute experiments. HHMI will develop specialized lab agents integrated with scientific instruments and analysis pipelines, while the Allen Institute will build coordinated multi-agent systems for multimodal analysis, including multi-omic integration, knowledge graphs, temporal dynamics, and experimental design. Together, the collaborations aim to compress manual analysis, surface overlooked patterns, and create transparent systems that augment rather than replace scientific judgment while keeping researchers in control.


### [Anthropic partners with CodePath to bring Claude to the US’s largest collegiate computer science program](https://yomu.fyi/post/anthropic-partners-with-codepath-to-bring-claude-to-the-us-s-largest-c.md)
- Company: [Anthropic](https://yomu.fyi/company/anthropic.md)
- Published: Aug 27, 2026
- Type: Announcement

Anthropic is partnering with CodePath to redesign collegiate computer science education as AI reshapes software development, placing Claude and Claude Code at the center of courses and career programs. More than 20,000 students at community colleges, state schools, and HBCUs will receive access to the tools, while CodePath addresses unequal access to industry-vetted courses and career networks. Claude is being integrated into AI courses including Foundations of AI Engineering, Applications of AI Engineering, and AI Open-Source Capstone; in fall 2025, more than 100 students piloted Claude Code on open-source projects such as GitLab, Puter, and Dokploy. The partners will also study AI’s effects on coding education and economic opportunity, while Howard University’s credited introductory AI course provides an early academic implementation of the applied curriculum.


### [Anthropic and the Government of Rwanda sign MOU for AI in health and education](https://yomu.fyi/post/anthropic-and-the-government-of-rwanda-sign-mou-for-ai-in-health-and-e.md)
- Company: [Anthropic](https://yomu.fyi/company/anthropic.md)
- Published: Aug 27, 2026
- Type: Announcement

Anthropic and the Government of Rwanda have signed a three-year Memorandum of Understanding to expand AI use across education, health, and public-sector systems. The agreement formalizes the partners’ earlier education collaboration and creates programs supporting Rwanda’s health goals, including efforts to eliminate cervical cancer and reduce malaria and maternal mortality. Government developer teams will receive access to Claude and Claude Code, alongside hands-on training, capacity building, and API credits, while the education work includes 2,000 Claude Pro licenses for educators, AI literacy training for public servants, and a Claude-powered learning companion deployed across eight African countries. The partnership is presented as a long-term commitment to responsible deployment, local autonomy, skills, infrastructure, and institutions, with Anthropic’s Beneficial Deployments team working with Rwanda’s Ministry of ICT and Innovation.


### [Introducing Claude for Small Business](https://yomu.fyi/post/introducing-claude-for-small-business.md)
- Company: [Anthropic](https://yomu.fyi/company/anthropic.md)
- Published: Aug 27, 2026
- Type: Announcement

Anthropic is launching Claude for Small Business, a package of connectors and ready-to-run workflows intended to put Claude inside small businesses’ existing tools. Through Claude Cowork, owners can connect Intuit QuickBooks, PayPal, HubSpot, Canva, Docusign, Google Workspace, and Microsoft 365, choose a job, and approve actions before anything sends, posts, or pays. The package includes 15 agentic workflows and 15 skills covering finance, operations, sales, marketing, HR, and customer service, including payroll planning, month-end close, invoice chasing, campaign preparation, and contract review. Trust controls retain existing permissions, keep users in the approval loop, and state that Team and Enterprise Plans do not train on customer data by default. The launch also includes a free AI Fluency for Small Business course, a multi-city Claude SMB Tour beginning May 14 in Chicago, and partnerships supporting solopreneurs and community development financial institutions.


### [Anthropic partners with the Gates Foundation](https://yomu.fyi/post/anthropic-partners-with-the-gates-foundation.md)
- Company: [Anthropic](https://yomu.fyi/company/anthropic.md)
- Published: Aug 27, 2026
- Type: Announcement

Anthropic and the Gates Foundation are committing $200 million in grants, Claude usage credits, and technical support over four years for global health, life sciences, education, and economic mobility programs in the US and around the world. The Beneficial Deployments team will provide engineering support and develop public goods, including datasets, knowledge graphs, connectors, benchmarks, and evaluation frameworks, while offering discounted Claude access to nonprofits and educational institutions. Health initiatives target low- and middle-income countries through vaccine and therapy research, health-data decision support, frontline-worker tools, and improved disease-transmission forecasts; initial disease work includes polio, HPV, and eclampsia/preeclampsia. Education work includes K-12 tutoring, career guidance, literacy and numeracy apps, while economic-mobility projects cover agricultural datasets and benchmarks, portable skills records, and links between training and employment outcomes. Anthropic says it will publish its thinking, decisions, and lessons as the partnership and broader beneficial-deployments work scale.


[Newer posts](https://yomu.fyi/posts/page/63.md) · [Older posts](https://yomu.fyi/posts/page/65.md)