---
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

### [Introducing Sonnet 4.6](https://yomu.fyi/post/introducing-sonnet-4-6.md)
- Company: [Anthropic](https://yomu.fyi/company/anthropic.md)
- Published: Jul 21, 2026
- Type: Announcement

Claude Sonnet 4.6 is introduced as an upgrade across coding, computer use, long-context reasoning, agent planning, knowledge work, and design. It adds a 1M token context window in beta, becomes the default for Free and Pro users in claude.ai and Claude Cowork, and keeps Sonnet 4.5’s starting price of $3/$15 per million tokens. Early testing reports that users preferred it over Sonnet 4.5 in Claude Code roughly 70% of the time and over Opus 4.5 59% of the time. The release also reports gains on OSWorld computer use, prompt-injection resistance, OfficeQA document comprehension, and long-horizon planning, while adding adaptive and extended thinking, beta context compaction, and generally available code execution, memory, programmatic tool calling, tool search, and tool-use examples.


### [The last mile: why great first-party data still doesn't make great marketing](https://yomu.fyi/post/the-last-mile-why-great-first-party-data-still-doesn-t-make-great-mark.md)
- Company: [Databricks](https://yomu.fyi/company/databricks.md)
- Author: Michael Burton, Katy Yuan
- Published: Jul 21, 2026
- Type: Problem & solution

Modern data platforms and sophisticated marketing systems can still fail to produce timely customer experiences when no bridge connects first-party data to campaign activation. The post explains Scott Brinker’s “composable canvas,” a five-ring architecture centered on a unified data core, with semantic layer, CaaS, decisioning, and apps and agents operating on shared data without repeated movement. In contrast, batch files, dashboards, and disconnected teams can leave autonomous agents unable to trigger campaigns or propensity models unused. It recommends closing this last mile through activation-ready data architecture, self-service marketing analytics, and one narrowly scoped AI agent tied to a measurable campaign outcome. Examples in the post include a 4X model conversion rate, a four-week account scoring launch, and grocery offer flows that changed weekly preparation from hours to minutes.


### [Zone-Failure-Resilient OpenSearch® at Uber](https://yomu.fyi/post/zone-failure-resilient-opensearch-at-uber.md)
- Company: [Uber](https://yomu.fyi/company/uber.md)
- Published: Jul 21, 2026
- Type: Problem & solution

Uber describes how it made OpenSearch deployments zone-failure resilient so querying and ingestion remain available after a complete zone loss. The design combines OpenSearch’s shard-allocation awareness with Odin-based isolation groups that distribute nodes predictably across failure domains and balance roles. Clusters maintain at least three shard copies across three isolation groups, while forced awareness keeps shards for a missing group unassigned instead of triggering heavy rebalancing; this trades immediate full replication for stability in surviving zones. For manager resiliency, five cluster manager nodes and cluster.auto\_shrink\_voting\_configuration allow quorum after a zone failure plus one additional manager failure. Uber reports that isolation groups replaced uneven physical zone IDs, enabling complete shard assignment, preventing disk skew and hot nodes, and supporting green cluster health.


### [How Dow Built a Carbon Footprint Ledger on Databricks to Accelerate Sustainability at Scale](https://yomu.fyi/post/how-dow-built-a-carbon-footprint-ledger-on-databricks-to-accelerate-su.md)
- Company: [Databricks](https://yomu.fyi/company/databricks.md)
- Author: Jesse Grekowicz, Tim Licquia, Varun Mahajan
- Published: Jul 21, 2026
- Type: Problem & solution

Dow built the Carbon Footprint Ledger (CFL) on Databricks to provide a transparent, verifiable way to track and communicate product carbon footprints amid customer Scope 3 commitments and tighter environmental disclosure requirements. The system combines carbon-accounting methodology assured against ISO 14067 and the GHG Protocol Product Standard with a calculation engine and certificate-issuing ledger that maintains residual balances. Within Dow’s Integrated Data Hub, Apache Spark unifies supply-chain, sustainability, operations, and other source data, while Unity Catalog governs access and Delta tables provide ACID transactions, time travel, and efficient upserts. An advanced optimization model, developed and deployed with MLflow, identifies lowest-greenhouse-gas production pathways and sends versioned, monitored results through the pipeline with lineage tracking. The implementation reduced PCF processing from weeks to a fraction of that, enabled portfolio-scale calculations, and supports verifiable certificates and commercial use of decarbonization investments.


### [Why R&D Data Belongs in the Lakehouse - and Why Agents Need It There](https://yomu.fyi/post/why-r-d-data-belongs-in-the-lakehouse-and-why-agents-need-it-there.md)
- Company: [Databricks](https://yomu.fyi/company/databricks.md)
- Author: Sebastian Eberhardt, Dominik Bentele, Jonathan Bräuer
- Published: Jul 21, 2026
- Type: Explainer

cellcentric describes how it built a governed lakehouse foundation for R&D data so industrial AI agents can reason over engineering context, not isolated source-system records. On Azure and Databricks, the Data Hub combines Unity Catalog governance, Lakehouse Federation, Delta Sharing, Declarative Automation Bundles, and an MCP interface for agents. Its Fuel Cell Passport unifies five enterprise systems, models seven hierarchy levels, refreshes daily, and uses a state-based temporal model for point-in-time configuration and rework-history questions. The platform treats context coverage as a data-quality metric, with 27 published products averaging 90% column-comment coverage, while authenticated user identity and Unity Catalog authorization constrain agent access. For many R&D and process-development workflows, work that once took weeks now ships in days, though complex multi-source investigations still require substantial domain review.


### [Announcing the Public Preview of Discover and Domains, powered by Unity Catalog](https://yomu.fyi/post/announcing-the-public-preview-of-discover-and-domains-powered-by-unity.md)
- Company: [Databricks](https://yomu.fyi/company/databricks.md)
- Author: Stef Bran Melendez, Kelly Albano
- Published: Jul 21, 2026
- Type: Announcement

Databricks announces the Public Preview of Domains and the Discover page, powered by Unity Catalog, to help people and agents identify relevant, high-quality, and safe data and AI assets. Domains organize tables, dashboards, notebooks, queries, metric views, Genie Agents, and apps by business structure, while Discover provides an internal marketplace with search, certification signals, popularity and trending indicators, and AI-powered recommendations. Data stewards can create domains and subdomains, certify assets, add descriptions and contacts, and curate page sections and pinned content. Domains extend Unity Catalog Semantics and feed Genie Ontology, giving agents business context for narrowing retrieval, prioritizing trusted assets, and interpreting metrics within each user’s existing permissions. The features are available in Public Preview through a Databricks workspace, where organizations can create domains and curate assets.


### [Setting a World Record for MoE Pre-Training on NVIDIA GB300 NVL72](https://yomu.fyi/post/setting-a-world-record-for-moe-pre-training-on-nvidia-gb300-nvl72.md)
- Company: [NVIDIA](https://yomu.fyi/company/nvidia-developer-blog.md)
- Author: Kirthi Devleker
- Published: Jul 21, 2026
- Type: Benchmark

Frontier AI pre-training increasingly relies on mixture of experts architectures, which lower compute per token while introducing severe all-to-all communication bottlenecks. To address these scaling limits, the NVIDIA GB300 NVL72 rack-scale system co-designs Blackwell Ultra GPUs, fifth-generation NVLink scale-up fabrics, and 800 Gbps SuperNIC scale-out networking. Testing on DeepSeek-V3 671B across 256 GPUs achieved a pre-training benchmark of 1,648 TFLOPs per GPU using Megatron Core. In addition to hardware co-design, ongoing software optimizations across Megatron Core, TorchTitan, and JAX contributed major throughput gains over time. Scaling from 256 to 1,024 GPUs sustained over 97% per-GPU performance efficiency across all evaluated frameworks by overlapping scale-out gradient traffic behind compute windows.


### [Inside NVIDIA Rubin GPU Architecture: Powering the Era of Agentic AI](https://yomu.fyi/post/inside-nvidia-rubin-gpu-architecture-powering-the-era-of-agentic-ai.md)
- Company: [NVIDIA](https://yomu.fyi/company/nvidia-developer-blog.md)
- Author: Eduardo Alvarez
- Published: Jul 21, 2026
- Type: Explainer

Agentic artificial intelligence workloads require sustained multistep inference, demanding low per-step latency, high decode throughput, and efficient long-context processing. NVIDIA designed the Rubin GPU to deliver up to ten times more agentic throughput per unit of energy than Blackwell. The hardware combines 336 billion transistors across two dies unified by NV-HBI, up to 288 GB of HBM4 memory providing 22 TB/s peak bandwidth, and the third-generation Transformer Engine reaching up to 50 petaflops of NVFP4 compute. Performance optimizations include inline descriptor updates for the Tensor Memory Accelerator, doubled K-dimension instruction throughput, 3-bit lookup-table weights, and structured 2:4 sparse attention compression. At the rack level, Vera Rubin NVL72 leverages Intelligent Power Smoothing and DSX MaxLPS to enable up to 40% more GPUs within a fixed power budget.


### [NVIDIA Vera CPU: Olympus Cores Built for Maximum Single-Thread Performance in Agentic AI](https://yomu.fyi/post/nvidia-vera-cpu-olympus-cores-built-for-maximum-single-thread-performa.md)
- Company: [NVIDIA](https://yomu.fyi/company/nvidia-developer-blog.md)
- Author: Praveen Menon
- Published: Jul 21, 2026
- Type: Explainer

Agentic AI workloads shift critical execution paths onto CPUs through sandboxed environments that execute code, invoke tools, and manage runtime states. The NVIDIA Vera CPU addresses these compute requirements via Olympus cores built to maximize single-thread performance and instructions per cycle under full socket load. To handle irregular control flows and pointer-heavy structures, Olympus combines a ten-wide decode engine, neural branch prediction supporting two taken branches per cycle, dependency-breaking acceleration, and dedicated graph prefetchers. Across scale-up systems, dual-socket Vera configurations leverage second-generation NVLink-C2C interconnects to maintain a clean two-NUMA-node topology alongside PCIe 6.4, CXL 3.1, and Arm CCA confidential computing. These architectural choices provide predictable latency and memory bandwidth across highly concurrent reinforcement learning and agent loops.


### [Personalizing Airbnb search by learning from the guest journey](https://yomu.fyi/post/personalizing-airbnb-search-by-learning-from-the-guest-journey.md)
- Company: [Airbnb](https://yomu.fyi/company/airbnb.md)
- Author: Daochen Zha
- Published: Jul 21, 2026

Airbnb replaced hundreds of hand-crafted ranking features with a Transformer-based sequence model that captures both long-term booking history and short-term browsing behavior. To keep latency low and throughput high, sequence representations are generated via daily batch jobs and combined with real-time queries using a co-trained setwise ranker.


### [How to structure a log](https://yomu.fyi/post/how-to-structure-a-log.md)
- Company: [Sentry](https://yomu.fyi/company/sentry.md)
- Author: Kyle Tryon
- Published: Jul 21, 2026
- Type: Tutorial

Logging arbitrary JSON objects often produces unstable messages and deeply nested data that complicate searching, aggregating, and debugging production incidents. Structured logging treats log output as real application data by establishing predictable, consistent schemas across services. Engineers should define stable event names using a domain.action pattern and capture dynamic dimensions with flattened, snake\_case attributes formatted in dot notation. Attribute payloads must contain only primitive values or primitive arrays rather than arbitrary objects, utilizing low-cardinality result identifiers and explicit units for numeric measurements. Teams can enforce these structural conventions automatically with linting tools such as ESLint, while routing execution durations and timing measurements to tracing spans.


### [Supabase Pipelines is now in Public Alpha](https://yomu.fyi/post/supabase-pipelines-is-now-in-public-alpha.md)
- Company: [Supabase](https://yomu.fyi/company/supabase.md)
- Author: riccardo\_busetti
- Published: Jul 21, 2026
- Type: Announcement

Supabase Pipelines has moved into public alpha as a service for replicating Supabase Postgres data to external analytical systems in near real time. The release adds automatic application of supported schema changes, faster initial synchronization through parallelism, and a destination request form for ClickHouse, Snowflake, and DuckLake; Google BigQuery is the first destination available to everyone in public alpha. Powered by the open-source Rust-based Supabase ETL, Pipelines uses Postgres logical replication to copy existing tables, stream inserts, updates, deletes, and truncates, and resume from the last acknowledged position after interruptions. It provides at-least-once delivery and managed pipeline controls, while remaining an alpha product with ongoing work on throughput, batching, backpressure, broader schema support, and additional destinations.


### [Devin Outposts on Modal](https://yomu.fyi/post/devin-outposts-on-modal.md)
- Company: [Modal](https://yomu.fyi/company/modal.md)
- Published: Jul 21, 2026
- Type: Announcement

Devin, an AI software engineer by Cognition, plans, writes, tests, and ships code semi-autonomously. Cognition now extends Devin with Outposts, moving execution into user-controlled environments while reasoning stays in Cognition’s cloud. Modal is a launch partner with modal-devin, an open-source integration running Outpost sessions in Modal Sandboxes with pre-built custom images, fast cold-starts, snapshotting, and cost-efficient bursting. For example, a task that requires a GPU fails immediately in the default sandbox but succeeds on a Modal-backed Outpost that spins up a GPU on demand and scales to zero when done. One command, uvx modal-devin init, sets up secure infrastructure with a scheduler and worker functions, and each team can have its own outpost with separate credentials.


### [Scaling Agentic RL: High-Throughput Agentic Training with Tunix](https://yomu.fyi/post/scaling-agentic-rl-high-throughput-agentic-training-with-tunix.md)
- Company: [Google](https://yomu.fyi/company/google.md)
- Author: Haoyu Gao, Lance Wang, Shadi Noghabi, Tianshu Bao, Weiren Yu
- Published: Jul 21, 2026
- Type: Explainer

Tunix is presented as a post-training library for scaling agentic reinforcement learning, where multi-turn tool use and environment latency can leave TPUs idle. Its design combines an asyncio-based, high-concurrency trajectory collector with asynchronous vLLM-TPU and SGLang-Jax integration, then streams variable-length trajectories through a producer-consumer queue rather than waiting for a complete synchronous batch. Composable agent and environment abstractions expose APIs for custom tools, benchmarks, and external systems while preserving multi-turn token handling and lifecycle management. Lightweight RL-specific metrics correlate rollout, training, weight synchronization, model-call, and environment stages with TPU timelines, complementing targeted XProf traces. The stated outcome is a more observable pipeline intended to reduce stalls, keep accelerators utilized, and support agentic RL natively in the JAX/TPU ecosystem.


### [Grabette: an open system to record robot-manipulation data](https://yomu.fyi/post/grabette-an-open-system-to-record-robot-manipulation-data.md)
- Company: [Hugging Face](https://yomu.fyi/company/hugging-face.md)
- Author: Steve Nguyen, Claire Houziel, Gaelle Lannuzel, Simon Le Goff, Jeremy Laville, Étienne
- Published: Jul 21, 2026
- Type: Announcement

Grabette is an open, low-cost system for recording robot-manipulation data without requiring a robot, lab, or teleoperation rig. It uses a handheld gripper with a wide fisheye observation camera, an RGBD tracking camera, an IMU, and gripper encoders to capture synchronized demonstrations and recover camera-local 6-DoF Cartesian pose plus gripper state. Users save episodes on a Raspberry Pi, then select them in a browser dashboard for upload, RTAB-MAP SLAM, trajectory verification, and conversion into a LeRobot dataset on the Hugging Face Hub. The release also includes a motorized Gripette gripper and an example that trains a Diffusion Policy on 200 demonstrations and evaluates it on an OpenArm 7-DoF arm via gRPC, demonstrating the recording-to-training workflow.


### [Branching databases like code: a CI/CD pattern for Lakebase, in production at Glaspoort](https://yomu.fyi/post/branching-databases-like-code-a-ci-cd-pattern-for-lakebase-in-producti.md)
- Company: [Databricks](https://yomu.fyi/company/databricks.md)
- Author: Hadi Farhat, Gideon Spierings, Ricardo de Vries, Raymon Veldman
- Published: Jul 20, 2026
- Type: Problem & solution

Glaspoort replaced a sprawl of outdated, one-off BI reports with a custom application that combines Databricks analytical tools with Lakebase, a serverless Postgres OLTP database. The application keeps curated lakehouse data in a read-only schema while application state occupies a separate schema, and supports development, acceptance, and production environments. Its CI/CD pattern treats database branches like code: long-lived environments branch directly from production, each pull request receives an ephemeral production-shaped branch, migrations are replayed, and a staging app image runs the full test suite against it. This topology makes periodic resets cheap, avoids deleting dependent branches, and preserves fresh test data; the team chose stacked promotion for velocity, retaining a separate crisis pipeline for urgent fixes.


### [Building a Transformer-Based Category Recommender at Thumbtack](https://yomu.fyi/post/building-a-transformer-based-category-recommender-at-thumbtack.md)
- Company: [Thumbtack](https://yomu.fyi/company/thumbtack.md)
- Author: Andrew Morss
- Published: Jul 20, 2026
- Type: Problem & solution

Thumbtack developed a transformer-based category recommender to personalize email and push notification suggestions across roughly 500 home service categories. Because user activity is sparse while contextual signals like location and climate are rich, the architecture represents search history and context features as sequences of tokens evaluated against candidate category tokens using cross-attention. Initial training with naive cross-entropy suffered from severe popularity bias, skewing recommendations toward head categories and yielding a low tail hit rate. To resolve this issue, the team implemented adaptive mixed sampling, combining in-batch and random negative candidates and keeping only top-scoring hard negatives for cross-entropy loss computation. This method raised tail hit rate from 2.1% to 28.4%, improved overall HR@10 to 54.6%, and yielded a statistically significant 5% lift in requests per user in production A/B testing.


### [Scaling document classification to 100k+ labels](https://yomu.fyi/post/scaling-document-classification-to-100k-labels.md)
- Company: [Databricks](https://yomu.fyi/company/databricks.md)
- Author: Jane Zhang, Arnav Singhvi
- Published: Jul 20, 2026
- Type: Benchmark

Large-taxonomy classification must map freeform text to 35,000–100,000 labels while meeting quality, cost, and throughput requirements. The benchmark compares vector search, a two-step vector-search-plus-AI-Classify workflow, and direct frontier-model calls across transaction, company, and biomedical datasets, using hybrid semantic and BM25 retrieval with Reciprocal Rank Fusion for the search stages. The combined workflow performed best, reaching 0.81 average accuracy versus 0.76 for Gemini 3.5 Flash, while costing roughly one-hundredth as much per document; its strongest average setting shortlisted the top 20 labels. Vector search alone was nearly 100 times cheaper than the workflow but scored more than 20 points lower, while full-taxonomy frontier calls could exceed context limits. Taxonomy updates can be handled by embedding and ingesting changed labels rather than retraining or redeploying classifiers.


### [Apply for Anthropic’s AI for Science rare disease research grants](https://yomu.fyi/post/apply-for-anthropic-s-ai-for-science-rare-disease-research-grants.md)
- Company: [Anthropic](https://yomu.fyi/company/anthropic.md)
- Published: Jul 20, 2026
- Type: Announcement

Anthropic’s AI for Science program is accepting applications for a rare genetic disease research call, offering accepted applicants up to $50,000 in Claude credits over six months across two tracks: basic science and early-stage biotechnology. The basic-science track focuses on collaboration and mechanism discovery, including work with the Monarch Initiative’s Mondo Disease Ontology, Monarch Knowledge Graph, and agent-friendly DisMech library. The biotech track targets faster therapeutic development through regulatory documentation, modality and target analysis, biomarker discovery, and potential basket-trial strategies. Applications are open through August 2, 2026, at 11:59 PM PST, and projects may use Claude Opus or other approved generally available biology models.


### [Online Learning for Cost-Efficient LLM Routing](https://yomu.fyi/post/online-learning-for-cost-efficient-llm-routing.md)
- Company: [Ramp](https://yomu.fyi/company/ramp.md)
- Author: Kedar Thakkar
- Published: Jul 20, 2026
- Type: Problem & solution

Ramp describes an internal LLM gateway that processes trillions of tokens daily and uses dynamic failure-aware routing to reduce AI spend. It combines an EWMA of genuine provider-side failure rates with Thompson Sampling over a Normal-Inverse-Gamma posterior for the mean and variance of log-latency. For each request, the router estimates the probability of outright failure or missing the caller's deadline, incorporates model cost, and orders options while preserving caller-defined model preferences and fallbacks. In a reranker with a six-second deadline, the strategy reacted to Gemini 3.1 Flash Lite rate limiting by switching to GPT 5.4 Nano Flex, whose latency matched Standard at half the price. Larger experiments reported more than 25% cost savings without degradation, while a streaming-equivalent deployment for Ramp Inspect produced 30% savings.


[Newer posts](https://yomu.fyi/posts/page/90.md) · [Older posts](https://yomu.fyi/posts/page/92.md)