---
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

### [Selection-Based Structured Reasoning: Toward Efficient Multimodal Search Agents](https://yomu.fyi/post/selection-based-structured-reasoning-toward-efficient-multimodal-searc.md)
- Company: [Apple](https://yomu.fyi/company/apple.md)
- Published: Oct 15, 2026
- Type: Problem & solution

Multimodal search agents typically rely on freeform reasoning before acting, but small models often produce lengthy reasoning that offers little guidance and increases inference cost. The authors propose Selection-Based Structured Reasoning (SSR), which reframes reasoning as selecting from pre-specified natural-language candidates rather than open-ended generation. SSR enables parallel scoring by using teacher-forced prefilling and a shared context KV cache to compute token likelihoods across candidates simultaneously. Evaluated on seven multimodal search benchmarks with 2B and 4B models, SSR maintains task performance under various reinforcement learning objectives and supervised fine-tuning. The framework achieves search success rates competitive with leading agents of the same scale while reducing per-turn reasoning latency by over 90% and total per-question inference latency by 28–54%.


### [Plan or Retrieve First? A Scale Study of Tool Discovery Architectures](https://yomu.fyi/post/plan-or-retrieve-first-a-scale-study-of-tool-discovery-architectures.md)
- Company: [Apple](https://yomu.fyi/company/apple.md)
- Published: Oct 15, 2026
- Type: Benchmark

Tool-augmented LLM agents must discover relevant APIs from large registries before planning, but finite context windows prevent injecting all tool definitions into a single prompt. Two competing architectures are Retrieve-then-Plan (RtP), which embeds the query and retrieves similar tools via nearest-neighbor search, and Plan-then-Retrieve (PtR), which first generates abstract tool stubs as retrieval queries. Evaluating on ToolBench with 46,984 API endpoints across 50 categories reveals a clear complexity-dependent crossover: RtP outperforms PtR on single-tool queries while PtR outperforms RtP on multi-tool queries. PtR’s stub-hallucination failure mode acts as a double-edged sword—on simple queries it abstracts away proper-noun signals, yet on complex queries it decomposes multi-faceted intent into functional sub-needs. PtR roughly doubles inference cost, which pays off on multi-tool queries but hurts on single-tool ones, motivating complexity-adaptive routing driven by RtP’s top-1 retrieval score.


### [Meta Muse Spark Open is live on Together AI](https://yomu.fyi/post/meta-muse-spark-open-is-live-on-together-ai.md)
- Company: [Together AI](https://yomu.fyi/company/together-ai.md)
- Published: Oct 15, 2026
- Type: Announcement

Meta Muse Spark Open, an open-weight model for agentic coding, is now available on Together AI as a Day 0 launch partner with a 1-million-token context window on serverless and dedicated infrastructure. The model is trained on real developer work including multi-file refactors, debugging sessions, and long-running tasks, leading to higher first-attempt accuracy and reliable tool calling. It features long-horizon autonomy with planning, observation, and self-correction, computer use for desktop and mobile, multimodal input accepting text, image, audio, and video, and intent commentary that states actions before execution for auditability. Early access customers Factory and Ollama tested the model pre-launch and are offering it to their users from day zero. Serverless pricing is $1.25 per million input tokens, $4.25 per million output tokens, and $0.15 per million cached input tokens.


### [Three layers, one repo: a tour of google/adk-recipes](https://yomu.fyi/post/three-layers-one-repo-a-tour-of-google-adk-recipes.md)
- Company: [Google](https://yomu.fyi/company/google.md)
- Author: Pier Paolo Ippolito, Shahin Saadati
- Published: Oct 15, 2026
- Type: Explainer

The google/adk-recipes repository serves as the single source of truth for ADK content across Google Cloud’s Agent Platform, including the Agents CLI, Agent Garden, and Gemini Enterprise app. It is organized into three top-level folders: core/ with reference patterns curated by the Agents CLI and DevRel teams, plugins/ with installable vertical skills packaged as Plugin Recipes conforming to the Agent Plugins v1.0.0 spec, and contrib/ with end-to-end deployable services in five languages. Every recipe ships a validated manifest.yaml and an AGENTS.md written primarily for AI coding assistants, and CI enforces the standards on every pull request. The core/ layer features Horizon, a reference harness for long-horizon memory agents that handles cross-session memory, per-user sandboxing, guardrails, and multi-step task orchestration. The retail Plugin Recipe includes product-search and virtual-tryon skills using Vector Search 2.0 and Gemini models, with deployable counterparts in contrib/ that can be surfaced on five deployment personas.


### [Welcome Muse Spark Open: open-weight frontier multimodal mixture-of-experts model from Meta](https://yomu.fyi/post/welcome-muse-spark-open-open-weight-frontier-multimodal-mixture-of-exp.md)
- Company: [Hugging Face](https://yomu.fyi/company/hugging-face.md)
- Author: Pedro Cuenca, merve, ben burtenshaw, Guarin Flück, Quentin Gallouédec, Alvaro Bartolome, Serveurperso, Jun Kim
- Published: Oct 15, 2026
- Type: Announcement

Meta released Muse Spark Open, an open-weight multimodal mixture-of-experts model with approximately 816 billion total parameters and around 42 billion active per token. It processes text, images, audio, and video inputs while generating text output, supporting a context window of up to one million tokens. The model uses 256 experts per MoE layer with 8 selected per token, and the BF16 checkpoint requires about 1.5 TB of storage. Meta trained Muse Spark Open for agentic coding tasks including reading codebases, writing and debugging code, and calling tools. The post provides recipes for running the model using transformers, llama.cpp, vLLM, and SGLang, as well as fine-tuning with LoRA on 64 H100 GPUs.


### [Welcome FLUX.3 Image Klein in Diffusers](https://yomu.fyi/post/welcome-flux-3-image-klein-in-diffusers.md)
- Company: [Hugging Face](https://yomu.fyi/company/hugging-face.md)
- Author: Aritra Roy Gosthipaty, Sayak Paul, Linoy Tsaban
- Published: Oct 15, 2026
- Type: Announcement

The Black Forest Labs FLUX.3 Image Klein model is now available in Diffusers, combining text-to-image, multi-reference image-to-image, and bounding-box layout control into one pipeline. Two open checkpoints are offered: a distilled version using 16 inference steps without classifier-free guidance for fast generation, and a base version requiring 50 steps with guidance. The distilled checkpoint produces a 1024x1024 image in about two seconds on an H100, and quantized NVFP4 and FP8 weights can be loaded via from\_single\_file to reduce memory. The model is built as a Modular Diffusers pipeline, enabling users to remove, replace, or insert custom components, such as a block that captures intermediate clean-image previews. LoRA fine-tuning scripts for both text-to-image and image editing are included, along with a prompting guide for optimal scene descriptions.


### [Local and Distributed Large LLM Inference with MLX and RDMA on Mac Studio](https://yomu.fyi/post/local-and-distributed-large-llm-inference-with-mlx-and-rdma-on-mac-stu.md)
- Company: [Apple](https://yomu.fyi/company/apple.md)
- Published: Oct 15, 2026
- Type: Benchmark

MLX enables running large language models locally on Apple silicon by exploiting the unified memory architecture of Mac Studio. The latest MLX release supports distributed inference across multiple Macs using RDMA over Thunderbolt, with pipeline or tensor parallelization schemes, and uses the jaccl communication collective. Benchmarks compare M3 Ultra and M5 Ultra across three models: Qwen 3.8 (27B) and Mistral Medium 3.5 (128B) fit on a single 512GB Mac, while the newly released Muse Spark Open (816B) requires parallelization across up to four machines at its 4-bit NVFP4 quantization. The M5 Ultra’s GPU Neural Accelerator delivers up to 4.6 times faster time-to-first-token than the M3 Ultra, while its 1.2 TB/s memory bandwidth improves generation throughput by 1.6 times for dense models. On four M5 Ultra Mac Studios with tensor parallelism, Muse Spark Open achieves over 36 tokens per second.


### [Evaluating AI agents: moving past the vibe check](https://yomu.fyi/post/evaluating-ai-agents-moving-past-the-vibe-check.md)
- Company: [Google](https://yomu.fyi/company/google.md)
- Author: Hugo Selbie, Dani Zamora, Jesse Carah
- Published: Oct 15, 2026
- Type: Problem & solution

Manual vibe checks for AI agents miss silent failures like retry loops, token waste, and hallucinated parameters while plausible final answers mask them. The post outlines a structured evaluation pipeline using the open-source agent-eval toolkit, starting with critical user journeys (CUJs) that define success in concrete, testable clauses. Synthetic user simulations (via ADK) generate multi-turn scenarios for initial trace collection, which developers grade with a ‘red pen’ before automating checks. Four layers of verification are recommended: unit tests, deterministic metrics, LLM-as-a-judge with boolean rubrics (avoiding unreliable 1-to-5 scales), and integration tests, all standardized on OpenInference/OpenTelemetry traces. The rollout progresses from Day 0 manual review through Day 1 automated checks to Day 2 CI/CD gates and shared dashboards, with real production traces fed back into the golden dataset.


### [Italian Enterprises Leading in AI](https://yomu.fyi/post/italian-enterprises-leading-in-ai.md)
- Company: [MongoDB](https://yomu.fyi/company/mongodb.md)
- Published: Oct 15, 2026
- Type: Problem & solution

AI models depend on flexible, high-performance data infrastructure. Leading Italian enterprises including TeamSystem, altermAInd, Brembo Solutions, and Eni are using MongoDB Atlas to power their AI initiatives. TeamSystem built a centralized AI platform with hybrid search combining full-text and vector search that delivers precise, context-aware results while minimizing hallucinations, scaling from 50,000 to 500,000 users and cutting task times by 30–40%. altermAInd solved GDPR and EU AI Act compliance at the database layer, enabling secure innovation for over 200,000 customers. Brembo Solutions uses ALCHEMIX to slash R&D formulation from months to minutes by leveraging MongoDB’s document model for diverse material data. Eni uses Atlas Search and vector search to instantly analyze years of geoscience data and Atlas Charts for real-time interactive visualizations without hand-coded dashboards. These companies show that a robust, secure, and scalable data platform is foundational to enterprise AI success.


### [OLTP vs. OLAP: Choosing Between Transactional and Analytical Databases](https://yomu.fyi/post/oltp-vs-olap-choosing-between-transactional-and-analytical-databases.md)
- Company: [Databricks](https://yomu.fyi/company/databricks.md)
- Author: Databricks Staff
- Published: Oct 14, 2026
- Type: Explainer

OLTP handles high-volume, low-latency transactions on current operational data for online banking, e-commerce, and inventory, while OLAP handles complex analytical queries on large historical datasets for financial analysis and business intelligence. OLTP databases ensure data integrity through ACID compliance and use normalized row-oriented tables; OLAP systems use denormalized dimensional models and columnar storage for efficient scans and aggregations. Data pipelines such as ETL or CDC tools like Debezium connect OLTP and OLAP systems, but latency and pipeline complexity rise with analytical freshness requirements. For mixed workloads, HTAP or lakehouse architectures can reduce synchronization overhead and analytical lag. The choice between OLTP and OLAP depends on the workload’s read-write pattern, latency needs, and data volume, with TPC-C and TPC-H as reference benchmarks.


### [What is HTAP? A plain-English guide to hybrid transactional/analytical processing](https://yomu.fyi/post/what-is-htap-a-plain-english-guide-to-hybrid-transactional-analytical.md)
- Company: [Databricks](https://yomu.fyi/company/databricks.md)
- Author: Databricks Staff
- Published: Oct 14, 2026
- Type: Explainer

HTAP (Hybrid Transactional/Analytical Processing) brings transaction processing and analytics together so applications can run analytical queries on current operational data without waiting for batch data pipelines. This architecture addresses the delay caused by traditional separation of OLTP and OLAP systems, where data had to be copied into a data warehouse before analysis. HTAP designs vary: converged systems use a single database engine for both workloads, while composed systems use separate engines connected through shared storage or replication. The central challenge is providing analytical access to fresh operational data without compromising transactional performance and reliability. For workloads needing independent scaling, Lake Transactional/Analytical Processing (LTAP) offers a unified storage layer under one governance model, reducing the need for change data capture and replication pipelines.


### [Everyone's adopting AI agents. Almost no one can prove the ROI. Here's why](https://yomu.fyi/post/everyone-s-adopting-ai-agents-almost-no-one-can-prove-the-roi-here-s-w.md)
- Company: [Databricks](https://yomu.fyi/company/databricks.md)
- Author: Databricks Staff
- Published: Oct 14, 2026
- Type: Explainer

Many organizations are quickly adopting AI agents, yet almost none can convincingly demonstrate the return on investment due to a mismatch between purchasing decisions and actual usage. This gap arises because executives buy AI out of fear of being left behind, not from validated ROI studies, and teams measure productivity using speed metrics that miss the broader picture. True productivity from agents is about automating routine, schedulable tasks so human experts can devote energy to judgment calls and creative work, provided quality is maintained alongside efficiency. Without rich business context like institutional knowledge, local regulations, and historical data, an AI agent cannot deliver meaningful work, no matter how fast it executes generic instructions. The most frequent mistake is scaling prematurely trying to roll out broadly after a few promising demos, without ensuring the necessary iteration, infrastructure, and readiness for reliable, quality outcomes at scale.


### [Power real-time applications and agents with Document Intelligence](https://yomu.fyi/post/power-real-time-applications-and-agents-with-document-intelligence.md)
- Company: [Databricks](https://yomu.fyi/company/databricks.md)
- Author: Archika Dogra, Jane Zhang, Zachary Wang, Nihit Desai
- Published: Oct 14, 2026
- Type: Announcement

Databricks announces General Availability of Document Intelligence REST APIs for real-time document parsing, extraction, and classification, with the ai\_decide API for routing in beta. These synchronous APIs eliminate the manual bottleneck of reading documents by transforming unstructured contracts, invoices, and insurance claims into structured insights in real time. The ai\_parse\_document API converts tables and figures into structured text with citation bounding boxes, while ai\_extract pulls specified fields with optional citations and confidence scores for verification. The ai\_classify API assigns labels using a business taxonomy with optional rationales, and ai\_decide uses content to route or flag documents for review. Developers can integrate these APIs into Databricks Apps for human-in-the-loop workflows, such as a contract intake flow where a lawyer uploads an agreement and sees key terms prefilled, or into custom agents on Agent Bricks.


### [Preparing for the 2026 US midterms](https://yomu.fyi/post/preparing-for-the-2026-us-midterms.md)
- Company: [Anthropic](https://yomu.fyi/company/anthropic.md)
- Author: Anthropic
- Published: Oct 14, 2026
- Type: Announcement

Anthropic detailed strengthened election safeguards for Claude ahead of the 2026 US midterms, based on a week-long analysis of 250,000 conversations to understand usage patterns and prevent misuse. About 5.75% of US Claude.ai conversations touch on politics, 0.51% on elections, and 0.01% ask for voting recommendations, with violations at 0.002%; election-specific queries are expected to increase closer to Election Day. New features include TurboVote banners, state-specific voting information from Democracy Works, and live election-night results. Claude Opus 5.5, Sonnet 5.5, and Haiku 5.5 scored 99% on political evenhandedness, but independent reviews found uneven hedging and unsolicited caveats in some responses. Anthropic partnered with ISD and McCain Institute to pressure-test defenses, added classifiers for multi-step misuse, and banned accounts used to produce content favoring adversarial narratives.


### [Yandex outage: Data shows that instead of going dark, malicious bots went on the move](https://yomu.fyi/post/yandex-outage-data-shows-that-instead-of-going-dark-malicious-bots-wen.md)
- Company: [Cloudflare](https://yomu.fyi/company/cloudflare.md)
- Author: Lai Yi Ohlsen
- Published: Oct 14, 2026
- Type: Incident / postmortem

Reported drone strikes on Yandex data centers from October 7 to 11 caused network traffic to drop 99% below baseline for Yandex Search and 90% for Yandex.Cloud. Using Cloudflare Radar and Bot Management data, the analysis compared pre-event automation baselines with post-strike patterns on both ASNs. The Yandex Search ASN drop consisted almost entirely of verified search crawlers: 93.6% of its bot requests were verified, and 78.1% belonged to YandexBot. On Yandex.Cloud, vulnerability scanning fell 90% and plugin probing dropped 93%, but high-volume operations migrated to other Russian hosting providers, preserving identical JA4 fingerprints. Global bot traffic remained stable at 39.5%, demonstrating that malicious automation moves rather than disappears and that request behavior analysis is necessary to understand threats.


### [Why AI at Work Needs a Rethink](https://yomu.fyi/post/why-ai-at-work-needs-a-rethink.md)
- Company: [Notion](https://yomu.fyi/company/notion.md)
- Author: Andrew McCarthy
- Published: Oct 14, 2026
- Type: Problem & solution

A survey of over 6,000 knowledge workers across the UK, France, and Germany, commissioned by Notion, reveals that daily AI users report significantly more benefits than occasional users. For example, 59% of daily users in the UK say AI agents are changing team collaboration, versus 37% overall, and 48% say AI helps them demonstrate more value, nearly double the overall 25%; similar patterns hold in France and Germany. Gen Z workers are especially likely to use AI to expand their responsibilities, but they also express higher concern about job security, notably in Germany where concern was 21 points above the national average. However, AI use remains largely individual: only about 20% of respondents embed AI agents in team-wide workflows. The post argues that the next step is rebuilding AI around shared workflows and company context, as demonstrated by companies like Pennylane and Brainlabs.


### [From 2,500 requests to 24: Rebuilding star history without the stargazers](https://yomu.fyi/post/from-2-500-requests-to-24-rebuilding-star-history-without-the-stargaze.md)
- Company: [Github](https://yomu.fyi/company/github.md)
- Author: Taylor Cuilty
- Published: Oct 14, 2026
- Type: Problem & solution

The post explains GitHub’s redesign of the stargazers API to provide star history without exposing user identities. The original endpoint returned paginated lists of stargazers, which tools used to reconstruct growth curves by fetching all pages, leading to privacy risks and expensive database queries that counted all stars on each request due to a pagination bug. GitHub first restricted access to stargazer lists to repository collaborators, but this broke tools that needed star history. Collaborating with maintainers and the Star History project, GitHub built a new endpoint that returns up to 30 weeks of daily star counts in a single request with no identities attached. The new endpoint reduces database work by filtering stars by time before aggregation, cutting requests for repositories like react/react from about 2,511 to 24.


### [How Ridgeline unified Iceberg and ClickHouse into one analytics layer for investment management](https://yomu.fyi/post/how-ridgeline-unified-iceberg-and-clickhouse-into-one-analytics-layer.md)
- Company: [Clickhouse](https://yomu.fyi/company/clickhouse.md)
- Published: Oct 14, 2026
- Type: Problem & solution

Ridgeline, a cloud-native investment management platform, rebuilt its performance analytics layer after Postgres could not handle the scale of larger customers. The team evaluated Redshift, SingleStore, Trino over Iceberg, and a custom in-memory engine before choosing ClickHouse for its analytical performance and lower cost. They adopted ClickHouse Cloud on AWS, using separate organizations for staging and production and per-customer services, with assistance from ClickHouse engineers. Historical data lands in Apache Iceberg via a Spark pipeline and is queried through ClickHouse, while recent data is written directly into ClickHouse. The architecture enabled Ridgeline to serve customers 10x larger than before and more than double its live customer count.


### [Your next job doesn’t have a title yet: Emerging role skills Q&A with Corey Hymel of Andela](https://yomu.fyi/post/your-next-job-doesn-t-have-a-title-yet-emerging-role-skills-q-a-with-c.md)
- Company: [Stack Overflow](https://yomu.fyi/company/stack-overflow.md)
- Author: Ryan Donovan
- Published: Oct 14, 2026
- Type: Explainer

Corey Hymel, head of research at Andela, analyzed nearly 50,000 job postings to identify emerging skill bundles that precede formal job titles. The research found AI is creating niche skill sets, not turning everyone into generalists, with roles like LLM application engineer building on top of APIs, FinOps reliability engineer managing consumption-based costs, and docs-as-code engineer constructing context for AI models. Hymel described “bleed roles” where skills merge between adjacent specialties, such as front-end product engineers stretching into product management. The data also showed developer attitudes toward AI tooling vary, with some hand-coding for accountability while others ship code that passes tests without concern. Hymel advises engineers to understand where trends are heading and invest effort to prepare for roles that are morphing but not yet commonly titled, emphasizing the importance of strategic upskilling.


### [HiRO-ACEv2: Simulating extreme weather at 3 km resolution for a fraction of the cost](https://yomu.fyi/post/hiro-acev2-simulating-extreme-weather-at-3-km-resolution-for-a-fractio.md)
- Company: [Hugging Face](https://yomu.fyi/company/hugging-face.md)
- Author: Kyle Wiggers
- Published: Oct 14, 2026
- Type: Announcement

HiRO-ACEv2 generates global 3 km climate data for extreme weather at roughly 1/300th the cost of the X-SHiELD model, with even larger savings for single-region runs. The system pairs a coarse 100 km AI model (ACE) with the new HiROv2 downscaler that simultaneously predicts rain, wind, and sea-level pressure instead of just rainfall, and has improved skill at reproducing extreme events. When fed coarsened X-SHiELD data HiROv2 matches the target’s high-resolution statistics within 3% for events as rare as one in a million data points, and accurately estimates tropical cyclone energy for storms never seen during training. Over a 10-year simulation the full system’s hurricane energy and extreme event snapshots compare well with X-SHiELD. However, HiRO cannot yet handle other climates because it depends on X-SHiELD-specific ACE tuning; future work includes training on warmer-world datasets and adding temperature and snow outputs.


[Newer posts](https://yomu.fyi/posts/page/2.md) · [Older posts](https://yomu.fyi/posts/page/4.md)