---
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

### [Claude in Microsoft Foundry and 365 Copilot](https://yomu.fyi/post/claude-in-microsoft-foundry-and-365-copilot.md)
- Company: [Anthropic](https://yomu.fyi/company/anthropic.md)
- Published: Sep 10, 2026
- Type: Announcement

Anthropic and Microsoft are expanding their partnership by making Claude Sonnet 4.5, Haiku 4.5, and Opus 4.1 available in public preview through Microsoft Foundry, where Azure customers can build applications and enterprise agents. The integration uses serverless deployment, supports Foundry APIs, tools, and workflows, and works with existing Azure agreements, including MACC, while offering Python, TypeScript, and C# SDKs with Microsoft Entra authentication. Claude models are also available through Claude Code in Foundry, and Microsoft 365 Copilot integrations include Researcher, Copilot Studio custom agents, and a preview Agent Mode option in Excel. In Excel, Claude can generate formulas, analyze data, identify errors, and iterate on solutions. Foundry's Global Standard deployment uses standard API pricing; US DataZone availability is described as coming soon, and the models support capabilities including code execution, web search, citations, vision, tool use, and prompt caching.


### [Snowflake and Anthropic announce $200M partnership](https://yomu.fyi/post/snowflake-and-anthropic-announce-200m-partnership.md)
- Company: [Anthropic](https://yomu.fyi/company/anthropic.md)
- Published: Sep 10, 2026
- Type: Announcement

Anthropic and Snowflake announced a multi-year, $200 million partnership that brings Claude models to Snowflake’s more than 12,600 customers across Amazon Bedrock, Google Cloud Vertex AI, and Microsoft Azure. The agreement also includes a joint go-to-market initiative for deploying AI agents in large enterprises. Claude can analyze structured and unstructured data through Snowflake Cortex AI, while Snowflake Intelligence and Cortex Agents support natural-language questions, multimodal SQL-based analysis, and custom multi-agent solutions under Snowflake Horizon Catalog governance. Snowflake’s internal benchmarks report greater than 90% accuracy on complex text-to-SQL tasks, and examples include Simon Data and Intercom. The partnership is positioned to help regulated industries move AI agents from pilots to production within governed data environments.


### [Anthropic acquires Bun as Claude Code hits $1B](https://yomu.fyi/post/anthropic-acquires-bun-as-claude-code-hits-1b.md)
- Company: [Anthropic](https://yomu.fyi/company/anthropic.md)
- Published: Sep 10, 2026
- Type: Announcement

Anthropic announced the acquisition of Bun, a JavaScript runtime, as Claude Code reached $1 billion in run-rate revenue six months after becoming publicly available. Bun is an all-in-one toolkit that combines a runtime, package manager, bundler, and test runner, and the post says its focus on speed, reliability, and developer experience has made it important to Claude Code’s infrastructure. The companies had collaborated for months, with Bun helping scale Claude Code and contributing to its native installer; Anthropic says the acquisition will bring faster performance, improved stability, and new capabilities. Bun will remain open source and MIT-licensed, while Anthropic continues investing in its JavaScript and TypeScript tooling, which has more than 7 million monthly downloads and over 82,000 GitHub stars.


### [Donating MCP to the Agentic AI Foundation](https://yomu.fyi/post/donating-mcp-to-the-agentic-ai-foundation.md)
- Company: [Anthropic](https://yomu.fyi/company/anthropic.md)
- Published: Sep 10, 2026
- Type: Announcement

Anthropic is donating the Model Context Protocol (MCP) to the Agentic AI Foundation (AAIF), a directed fund under the Linux Foundation co-founded by Anthropic, Block, and OpenAI, with support from Google, Microsoft, AWS, Cloudflare, and Bloomberg. Introduced as an open standard for connecting AI applications to external systems, MCP now has more than 10,000 active public servers, adoption across major AI products, enterprise deployment support, and official SDKs with 97M+ monthly downloads across Python and TypeScript. The donation places MCP alongside Block’s goose and OpenAI’s AGENTS.md as AAIF founding projects, while its existing governance model remains unchanged. MCP also has a community-driven Registry, and the November 25th specification release added asynchronous operations, statelessness, server identity, and official extensions.


### [Accenture and Anthropic launch partnership](https://yomu.fyi/post/accenture-and-anthropic-launch-partnership.md)
- Company: [Anthropic](https://yomu.fyi/company/anthropic.md)
- Published: Sep 10, 2026
- Type: Announcement

Anthropic and Accenture announced an expanded partnership aimed at helping enterprises move AI projects from pilots to full-scale deployment. The companies are forming the Accenture Anthropic Business Group, a dedicated Claude-focused practice, and plan to train approximately 30,000 Accenture professionals, including engineers who embed Claude in client environments. Their first joint offering places Claude Code at the center of software development and combines productivity and ROI measurement, workflow redesign, and change management and training for CIOs. Initial industry solutions target financial services, life sciences, healthcare, and the public sector, where security and governance requirements are strict; the partnership also brings Claude to Accenture Innovation Hubs and establishes a Claude Center of Excellence for co-designing offerings.


### [Protecting the wellbeing of our users](https://yomu.fyi/post/protecting-the-wellbeing-of-our-users.md)
- Company: [Anthropic](https://yomu.fyi/company/anthropic.md)
- Published: Sep 10, 2026
- Type: Explainer

Claude’s safeguards for conversations involving emotional support focus on suicide and self-harm, sycophancy, and an 18+ access requirement. For suicide and self-harm, the approach combines system-prompt guidance, reinforcement learning informed by human preference data and expert input, and a classifier that can display country-specific crisis resources supplied by ThroughLine. The post reports that Claude Opus 4.5, Sonnet 4.5, and Haiku 4.5 respond appropriately to clear-risk single-turn prompts 98.6%, 98.7%, and 99.3% of the time, while multi-turn results are 86% and 78% for Opus 4.5 and Sonnet 4.5. Prefilled real-conversation stress tests produced 91% and 73% for those models, versus 36% for Opus 4.1. The work also describes efforts to reduce sycophancy, evaluate model behavior, identify underage users, and continue publishing methods and results.


### [Working with the US Department of Energy](https://yomu.fyi/post/working-with-the-us-department-of-energy.md)
- Company: [Anthropic](https://yomu.fyi/company/anthropic.md)
- Published: Sep 10, 2026
- Type: Announcement

Anthropic and the US Department of Energy (DOE) announce a multi-year partnership under the Genesis Mission, targeting American energy dominance, biological and life sciences, and scientific productivity across potentially all 17 national laboratories. Anthropic plans to give DOE researchers access to Claude and engineers who can build purpose-built AI agents, Model Context Protocol servers connecting Claude with scientific instruments and tools, and Claude Skills for specialized workflows. Proposed applications include speeding permitting reviews, supporting nuclear technology research and energy security, developing pandemic and biological-threat early-warning systems, accelerating drug discovery, and using fifty years of DOE research to surface ideas and patterns. The partnership extends prior DOE work, including a nuclear risk classifier co-developed with the National Nuclear Security Administration and Claude’s rollout at Lawrence Livermore National Laboratory. Anthropic says future arrangements could develop a model for collaboration between AI and human researchers.


### [(Re)introducing Developer Story](https://yomu.fyi/post/re-introducing-developer-story.md)
- Company: [Stack Overflow](https://yomu.fyi/company/stack-overflow.md)
- Author: Philippe Beaudette
- Published: Sep 10, 2026
- Type: Announcement

Stack Overflow is bringing back Developer Story, a profile feature that records a developer’s career and presents expertise alongside contributions. The relaunch introduces “specialties,” areas of proven expertise that users can choose to highlight, with planned integrations to show verified contributions from other sites and Stack Overflow’s network. The feature is intended to become the center of a fuller technical identity and the foundation of “Stack Identity,” described as a verified developer proof-of-work layer for showcasing skills and career. The release is privacy-centric, excludes the old Developer Story data because it was deleted for privacy compliance, and is being developed through frequent changes, including improvements to specialties and integrations.


### [How Full-Stack NIM Optimizations Deliver 2.5x More Users on Nemotron 3 Ultra](https://yomu.fyi/post/how-full-stack-nim-optimizations-deliver-2-5x-more-users-on-nemotron-3.md)
- Company: [NVIDIA](https://yomu.fyi/company/nvidia-developer-blog.md)
- Author: Elizabeth Goodman
- Published: Sep 10, 2026
- Type: Benchmark

Production serving for agentic AI must maximize concurrent users on fixed GPU infrastructure while preserving latency, especially with long prompts, reused context, and streamed responses. The article presents NVIDIA NIM as a deployable microservice that packages validated model- and GPU-aware choices, then measures Nemotron 3 Ultra NIM 2.0.12 against an open-source baseline on four B200 GPUs using a 64K/400 workload, 76% KV reuse, and 50 TPS per user (20 ms ITL). The optimized stack combines autotuned MoE and Mamba kernels, tensor and expert-aware parallelism, prefix and Mamba state-cache reuse, scheduling, batching, memory tuning, and MTP speculative decoding. At the stated target, throughput rises from 718 to 1,997 output tokens per second—more than 2.5x—while the article recommends replaying representative traffic with NVIDIA AIPerf to select a Pareto point meeting the application’s latency constraint.


### [Building resilient real-time streaming workers with Amazon DynamoDB leases](https://yomu.fyi/post/building-resilient-real-time-streaming-workers-with-amazon-dynamodb-le.md)
- Company: [AWS](https://yomu.fyi/company/aws.md)
- Author: Siddhesh Tiwari
- Published: Sep 10, 2026
- Type: Tutorial

Managing hundreds of persistent outbound WebSocket connections requires explicit ownership because a failed worker can drop connections and stop data flow until they are reassigned. The described fleet-management pattern uses Amazon DynamoDB leases, conditional writes, and a global secondary index to ensure one worker owns each connection, renew ownership with heartbeats, and find expired leases for reconciliation. Amazon ECS on AWS Fargate workers poll Amazon SQS for notifications, maintain the WebSockets, and release or transfer ownership during graceful shutdown, failures, and scaling. The design also records desired state, the upstream URL, and the last processed sequence number in the lease item, while clock synchronization supports its time-based expiry. The post identifies heartbeat writes as the dominant DynamoDB cost and recommends provisioned capacity with Auto Scaling for sustained high connection counts, while noting that longer intervals reduce cost but extend failover.


### [Refreshing the Travel-Time Map Behind Lyft’s Marketplace: Rebuilding Neighborhood Reachability…](https://yomu.fyi/post/refreshing-the-travel-time-map-behind-lyft-s-marketplace-rebuilding-ne.md)
- Company: [Lyft](https://yomu.fyi/company/lyft.md)
- Author: Manjunath Shettar
- Published: Sep 10, 2026
- Type: Problem & solution

Lyft’s Neighborhood Reachability Signals dataset supplies precomputed travel-time estimates between geohash-6 cells and a vocabulary of usable cells for pricing, driver heatmaps, and marketplace features. Production ETAs had remained largely based on 2018–2019 data, systematically underestimating travel times and omitting genuinely matchable pairs, while consumers maintained separate allowlists and denylists. The FORTOP team redesigned the Airflow-generated files by rebuilding the geohash universe from demand and supply history, estimating dense pairs from trip history, using Routing Simulation System (RSS) fallbacks for sparse pairs, and blending partial observations. In the second half of 2025, the refreshed dataset was adopted by Pricing after validation, with reported improvements in accuracy and completeness, a six-month refresh cadence, and plans for nine time-category files per region to represent traffic patterns.


### [Introducing WalShadow: Sub-second Postgres replication to ClickHouse from physical WAL](https://yomu.fyi/post/introducing-walshadow-sub-second-postgres-replication-to-clickhouse-fr.md)
- Company: [Clickhouse](https://yomu.fyi/company/clickhouse.md)
- Published: Sep 10, 2026
- Type: Announcement

ClickHouse announces WalShadow, an open-source engine that replicates Postgres data to ClickHouse directly from physical WAL, targeting sub-second analytical availability without logical replication. It consumes the physical WAL stream outside the source database, tracks schema in a shadow Postgres instance, decodes heap records through parallel Rust-based decoders, builds ClickHouse-native blocks, and inserts them concurrently. Because parallel blocks can arrive out of order, each row carries the source WAL position (\_lsn), while barriers preserve ordering for schema changes and truncations. In benchmarks using modest c8i.2xlarge instances, WalShadow reached approximately 200 ms commit-to-visible latency and 289,000 rows per second, compared with PeerDB’s roughly 10 seconds and 120,000 rows per second; it supports initial load, continuous replication, schema evolution, restart recovery, and planned switchovers.


### [Three principles for building a vector platform at Thumbtack](https://yomu.fyi/post/three-principles-for-building-a-vector-platform-at-thumbtack.md)
- Company: [Thumbtack](https://yomu.fyi/company/thumbtack.md)
- Author: John Zhu
- Published: Sep 10, 2026
- Type: Explainer

Thumbtack describes a vector platform that lets ML engineers stand up production vector search without separately negotiating database access, building ETL, or writing a query service. The design uses an existing Aurora Postgres cluster with pgvector, while Airflow jobs generate embeddings into BigQuery and managed reverse-ETL syncs them into vector tables; a query API validates inputs and returns ranked results. Three principles guide the layers: reuse trusted infrastructure, treat embeddings as ordinary data flowing through established pipelines, and lower adoption costs for subsequent teams. Production examples include semantic customer-support escalation using cosine similarity and an internal knowledge assistant combining vector retrieval, often with keyword search, before a language model responds; planned directions include self-serve provisioning, vector-backed feature retrieval, and framework-neutral agent memory.


### [ClickHouse is a launch partner for the Data agent in ChatGPT Work](https://yomu.fyi/post/clickhouse-is-a-launch-partner-for-the-data-agent-in-chatgpt-work.md)
- Company: [Clickhouse](https://yomu.fyi/company/clickhouse.md)
- Published: Sep 10, 2026
- Type: Announcement

ClickHouse is a launch partner for the Data agent in ChatGPT Work, with a plugin that connects ClickHouse Cloud to natural-language questions, reports, and interactive dashboards. The listing combines a fully managed ClickHouse Remote MCP server with open-source ClickHouse Agent Skills, and OAuth scopes access to the organizations and services visible to the connected user without requiring an API key. The MCP server exposes 13 read-only tools for metadata, SELECT queries, and service, backup, ClickPipes, and billing information, while the skills guide SQL and code around schema design, optimization, and ingestion. ChatGPT Work can use query results to build branded dashboards, with fresh answers coming from the ClickHouse Cloud service; availability depends on ChatGPT plan and workspace settings.


### [Unifying governance across engines and catalogs in the Open Lakehouse](https://yomu.fyi/post/unifying-governance-across-engines-and-catalogs-in-the-open-lakehouse.md)
- Company: [Databricks](https://yomu.fyi/company/databricks.md)
- Author: Daniel Weeks, Ryan Blue, Andrei Tserakhau
- Published: Sep 10, 2026
- Type: Explainer

Apache Iceberg’s REST Catalog additions—read restrictions and catalog labels—address different cross-platform governance problems: delegating enforcement to trusted engines and carrying governance context across federated catalogs. Read restrictions let a catalog evaluate policy for a principal and return bounded column-projection actions and row-filter expressions that a trusted engine must enforce, while leaving trust establishment outside the protocol. Catalog labels exchange opaque key-value metadata at table and column level, allowing a consuming catalog to map classifications locally, preserve native policy expressiveness, and maintain its own audit trail. Together with centralized enforcement through Cross-Engine ABAC, they provide distinct options for direct access and catalog federation, although read restrictions cannot express policies outside their standardized vocabulary.


### [Why don’t machine learning research agents overfit?](https://yomu.fyi/post/why-don-8217-t-machine-learning-research-agents-overfit.md)
- Company: [AWS](https://yomu.fyi/company/aws.md)
- Author: Martin Bertran Lopez, Aaron Roth
- Published: Sep 10, 2026
- Type: Explainer

Repeatedly optimizing against long-lived machine-learning benchmarks appears to invite overfitting, yet the post notes that gains on heavily reused benchmarks have largely transferred to newly created test sets. It describes a paper that resets LLM-based research agents, gives an explorer repeated access to validation feedback, and passes the resulting strategy through a token bottleneck to a fresh reproducer. Across eight datasets, strategies from hundreds of experiments were reproducible from prompts as short as 16 to 32 tokens; at shorter budgets, performance diverged, showing that the compressed message carried data-dependent choices rather than only prior knowledge. Deliberate overfitting occurred in 38 of 102 runs, with validation accuracy more than 10% above held-out accuracy, but those gains disappeared after compression; possible pretraining side channels remain unresolved.


### [High-Throughput Structure Prediction with BioNeMo Inference Runtime](https://yomu.fyi/post/high-throughput-structure-prediction-with-bionemo-inference-runtime.md)
- Company: [NVIDIA](https://yomu.fyi/company/nvidia-developer-blog.md)
- Author: Elizabeth Goodman
- Published: Sep 10, 2026
- Type: Tutorial

Proteome-scale biomolecular structure prediction requires moving large worklists efficiently, and NVIDIA BioNeMo Inference Runtime (BioIR) accelerates supported models on NVIDIA GPUs while retaining a PyTorch workflow. Its end-to-end processor passes InputRequest objects through parsing, tokenization, feature generation, GPU inference, and PDB or mmCIF writing; a Ray backend places complete model replicas on GPUs for concurrent independent inputs, while direct PyTorch integration supports custom code. In a matched 1,000-target human-dimer benchmark, BioIR completed all targets at 58.5K successfully folded residues per allocated GPU-hour versus 20.2K for the tested public implementation, a 2.90× improvement, which ran out of memory on 29 targets. Rated-power-equivalent estimates for one million comparable targets were 11 versus 35 MWh at 8-GPU TDP and 21 versus 64 MWh at full-node maximum power, excluding data-center overhead and several pipeline costs.


### [Announcing On-Demand Compute: Instant compute for your most intensive workloads](https://yomu.fyi/post/announcing-on-demand-compute-instant-compute-for-your-most-intensive-w.md)
- Company: [Clickhouse](https://yomu.fyi/company/clickhouse.md)
- Published: Sep 10, 2026
- Type: Announcement

ClickHouse Cloud’s On-Demand Compute is announced in private preview as a way to run eligible queries on a shared pool of workers outside a service’s primary cluster, avoiding production contention and reactive autoscaling delays. It uses ClickHouse’s multistage distributed query execution and cost-based optimizer to distribute complex workloads, with queries requesting worker counts through settings including make\_distributed\_plan and distributed\_plan\_workers\_num. Workers are leased for at least 60 seconds, reused during the lease, and released when it expires; concurrent queries share the pool, which grows to the largest request. In a stated benchmark, On-Demand Compute was faster for the tested workload and at 25 concurrent queries delivered similar performance while using 50% less compute, though results vary by query shape.


### [Improving Lakebase Postgres compute cache](https://yomu.fyi/post/improving-lakebase-postgres-compute-cache.md)
- Company: [Databricks](https://yomu.fyi/company/databricks.md)
- Author: David Wein, Sunil Kamath, Haoyu Huang
- Published: Sep 10, 2026
- Type: Problem & solution

Lakebase Postgres uses disaggregated storage backed by object storage, so compute-side caching must deliver low latency and high throughput while avoiding inefficient memory use. Its original design paired a conservatively sized PostgreSQL shared buffer with a local file cache (LFC), but the roughly 1 GB shared-buffer cap pushed many hits to slower NVMe storage and could create double buffering. For fixed-size computes with CU >= 80, the new configuration disables the LFC and sets shared buffers to 75% of DRAM, using explicit 2 MB HugeTLB pages to control page-table and TLB overhead. Benchmark tests found huge pages reduced tail read latency by up to ~40% and CPU use by up to ~30%, while production examples reported doubled throughput, 5× fewer storage reads, or 5× lower CPU use. Work is underway to dynamically resize shared buffers and huge pages for autoscaling computes.


### [From token consumer to token provider: Building your org's AI API](https://yomu.fyi/post/from-token-consumer-to-token-provider-building-your-org-s-ai-api.md)
- Company: [Red Hat](https://yomu.fyi/company/red-hat.md)
- Author: Markell Rawls
- Published: Sep 10, 2026
- Type: Explainer

Enterprise AI usage can become costly and fragmented when teams rely on separate provider accounts, personal keys, and independently operated inference stacks, leaving spending, workloads, and governance difficult to see. The post presents Model-as-a-Service (MaaS) in Red Hat OpenShift AI 3.4 as a shared service in which a platform team deploys approved models and exposes them through a standard, OpenAI-compatible API. Administrators can use the model catalog, distributed inference through llm-d on NVIDIA L40s, scoped API keys, rate limits, and dashboards for token use, latency, errors, GPU utilization, and queue length. Developers obtain self-service endpoints and can connect tools such as Claude Code, OpenCode, or Continue with existing clients. The architecture combines OpenShift, vLLM, an AI gateway, Prometheus, Grafana, and Jaeger to centralize inference, governance, and cost attribution across local and external model providers.


[Newer posts](https://yomu.fyi/posts/page/48.md) · [Older posts](https://yomu.fyi/posts/page/50.md)